Artificientis in the Autonomous Vehicles Industry

💡From 🇩🇪autonomous driving research to AI-powered smartphone📱video telematics solution to empower safer driving 🚗, decreased CO2 emission 🌱 and lower insurance costs 🤑 We make driving safer ✅ We reduce carbon emissions 🌍 🎓Artificient is proud to have RWTH Aachen & WHU alumni lead our international & female-empowered founding team.

Artificient

artificient

Hey there! Let me introduce you to LangChain, an awesome library that empowers developers to build powerful applications using (LLMs) and other computational resources. In this guide, I'll give you a quick rundown on how LangChain works and explore some cool use cases, like question-answering, chatbots, and agents. I'll also walk you through a quick-start guide to help you get going. Let's begin! large language models What is LangChain? is a powerful framework designed to help developers build end-to-end applications using language models. It offers a suite of tools, components, and interfaces that simplify the process of creating applications powered by large language models (LLMs) and chat models. LangChain makes it easy to manage interactions with language models, chain together multiple components, and integrate additional resources, such as APIs and databases. LangChain LangChain has a number of core concepts. Let's go through them one by one: 1. Components and Chains In LangChain, components are modular building blocks that can be combined to create powerful applications. A chain is a sequence of components (or other chains) put together to accomplish a specific task. For example, a chain might include a prompt template, a language model, and an output parser, all working together to handle user input, generate a response, and process the output. 2. Prompt Templates and Values A Prompt Template is responsible for creating a PromptValue, which is what's eventually passed to the language model. Prompt Templates help convert user input and other dynamic information into a format suitable for the language model. PromptValues are classes with methods that can be converted to the exact input types each model type expects (like text or chat messages). 3. Example Selectors Example Selectors are useful when you want to include examples in a prompt dynamically. They take user input and return a list of examples to use in the prompt, making it more powerful and context-specific. 4. Output Parsers Output Parsers are responsible for structuring language model responses into a more usable format. They implement two main methods: one for providing formatting instructions and another for parsing the language model's response into a structured format. This makes it easier to work with the output data in your application. 5. Indexes and Retrievers Indexes are a way to organize documents to make it easier for language models to interact with them. Retrievers are interfaces for fetching relevant documents and combining them with language models. LangChain provides tools and functionality for working with different types of indexes and retrievers, like vector databases and text splitters. 6. Chat Message History LangChain primarily interacts with language models through a chat interface. The ChatMessageHistory class is responsible for remembering all previous chat interactions, which can then be passed back to the model, summarized, or combined in other ways. This helps maintain context and improves the model's understanding of the conversation. 7. Agents and Toolkits Agents are entities that drive decision-making in LangChain. They have access to a suite of tools and can decide which tool to call based on user input. Toolkits are sets of tools that, when used together, can accomplish a specific task. The Agent Executor is responsible for running agents with the appropriate tools. By understanding and leveraging these core concepts, you can harness the power of LangChain to build advanced language model applications that are adaptable, efficient, and capable of handling complex use cases. What is a LangChain Agent? A LangChain Agent is an entity that drives decision-making in the framework. It has access to a set of tools and can decide which tool to call based on the user's input. Agents help build complex applications that require adaptive and context-specific responses. They are especially useful when there's an unknown chain of interactions that depend on the user's input and other factors. How would someone use LangChain? To use LangChain, a developer would start by importing the necessary components and tools, such as LLMs, chat models, agents, chains, and memory features. These components are combined to create an application that can understand, process, and respond to user inputs. LangChain provides a variety of components for specific use cases, such as personal assistants, question answering over documents, chatbots, querying tabular data, interacting with APIs, extraction, evaluation, and summarization. What's a LangChain model? A LangChain model is an abstraction that represents different types of models used in the framework. There are three main types of models in LangChain: : These models take a text string as input and return a text string as output. They are the backbone of many language model applications. LLMs (Large Language Models) : Chat Models are backed by a language model but have a more structured API. They take a list of Chat Messages as input and return a Chat Message. This makes it easy to manage conversation history and maintain context. Chat Models : These models take text as input and return a list of floats representing the text's embeddings. These embeddings can be used for tasks like document retrieval, clustering, and similarity comparisons. Text Embedding Models Developers can choose the appropriate LangChain model for their use case and leverage the provided components to build their applications. Key Features of LangChain LangChain is designed to support developers in six main areas: : LangChain makes it easy to manage prompts, optimize them, and create a universal interface for all LLMs. Plus, it includes some handy utilities for working with LLMs. LLMs and Prompts : These are sequences of calls to LLMs or other utilities. LangChain provides a standard interface for chains, integrates with various tools, and offers end-to-end chains for popular applications. Chains : LangChain enables chains to interact with external data sources to gather data for the generation step. For example, it can help with summarizing long texts or answering questions using specific data sources. Data Augmented Generation : An agent lets an LLM make decisions about actions, take those actions, check the results, and keep going until the job's done. LangChain provides a standard interface for agents, a variety of agents to choose from, and examples of end-to-end agents. Agents : LangChain has a standard interface for memory, which helps maintain state between chain or agent calls. It also offers a range of memory implementations and examples of chains or agents that use memory. Memory : It's tough to evaluate generative models with traditional metrics. That's why LangChain provides prompts and chains to help developers assess their models using LLMs themselves. Evaluation Use Cases LangChain supports a bunch of use cases, like: : Answer questions based on given documents, using the info in those documents to create answers. Question Answering over specific documents : Build chatbots that can produce text by leveraging LLMs' capabilities. Chatbots : Develop agents that can decide on actions, take those actions, observe the results, and keep going until they're done. Agents Quickstart Guide: Building an End-to-End Language Model Application with LangChain Installation First, let's get LangChain installed. Just run the following command: pip install langchain Environment Setup Now, since LangChain often needs to integrate with model providers, data stores, APIs, and more, we'll set up our environment. In this example, we're going to use OpenAI's APIs, so we need to install their SDK: pip install openai Next, let's set up the environment variable in the terminal: export OPENAI_API_KEY="..." Or, if you prefer to work inside a Jupyter notebook or Python script, you can set the environment variable like this: import os
os.environ["OPENAI_API_KEY"] = "..." Building a Language Model Application: LLMs With LangChain installed and the environment set up, we're ready to start building our language model application. LangChain provides a bunch of modules that you can use to create language model applications. You can combine these modules for more complex applications or use them individually for simpler ones. Building a Language Model Application: Chat Models In addition to LLMs, you can also work with chat models. These are a variation of language models that use language models under the hood but have a different interface. Instead of a "text in, text out" API, chat models work with chat messages as inputs and outputs. Chat model APIs are pretty new, so everyone is still figuring out the best abstractions. To get chat completions, you'll need to pass one or more messages to the chat model. LangChain currently supports AIMessage, HumanMessage, SystemMessage, and ChatMessage types. You'll mostly work with HumanMessage, AIMessage, and SystemMessage. Here's an example of using chat models: from langchain.chat_models import ChatOpenAI
from langchain.schema import (
    AIMessage,
    HumanMessage,
    SystemMessage
)

chat = ChatOpenAI(temperature=0) You can get completions by passing in a single message: chat([HumanMessage(content="Translate this sentence from English to French. I love programming.")])
# -> AIMessage(content="J'aime programmer.", additional_kwargs={}) Or pass in multiple messages for OpenAI's gpt-3.5-turbo and gpt-4 models: messages = [
    SystemMessage(content="You are a helpful assistant that translates English to French."),
    HumanMessage(content="Translate this sentence from English to French. I love programming.")
]
chat(messages)
# -> AIMessage(content="J'aime programmer.", additional_kwargs={}) You can also generate completions for multiple sets of messages using . This returns an LLMResult with an additional message parameter: generate batch_messages = [
    [
        SystemMessage(content="You are a helpful assistant that translates English to French."),
        HumanMessage(content="Translate this sentence from English to French. I love programming.")
    ],
    [
        SystemMessage(content="You are a helpful assistant that translates English to French."),
        HumanMessage(content="Translate this sentence from English to French. I love artificial intelligence.")
    ],
]
result = chat.generate(batch_messages)
result
# -> LLMResult(generations=[[ChatGeneration(text="J'aime programmer.", generation_info=None, message=AIMessage(content="J'aime programmer.", additional_kwargs={}))], [ChatGeneration(text="J'aime l'intelligence artificielle.", generation_info=None, message=AIMessage(content="J'aime l'intelligence artificielle.", additional_kwargs={}))]], llm_output={'token_usage': {'prompt_tokens': 71, 'completion_tokens': 18, 'total_tokens': 89}}) And you can extract information like token usage from the LLMResult: result.llm_output['token_usage']
# -> {'prompt_tokens': 71, 'completion_tokens': 18, 'total_tokens': 89} With chat models, you can also use templating by employing a MessagePromptTemplate. You can create a ChatPromptTemplate from one or more MessagePromptTemplates. The method of ChatPromptTemplate returns a PromptValue, which you can convert to a string or Message object, depending on whether you want to use the formatted value as input to an LLM or chat model. format_prompt Here's an example: from langchain.chat_models import ChatOpenAI
from langchain.prompts.chat import (
    ChatPromptTemplate,
    SystemMessagePromptTemplate,
    HumanMessagePromptTemplate,
)

chat = ChatOpenAI(temperature=0)

template="You are a helpful assistant that translates {input_language} to {output_language}."
system_message_prompt = SystemMessagePromptTemplate.from_template(template)
human_template="{text}"
human_message_prompt = HumanMessagePromptTemplate.from_template(human_template)

chat_prompt = ChatPromptTemplate.from_messages([system_message_prompt, human_message_prompt])

# get a chat completion from the formatted messages
chat(chat_prompt.format_prompt(input_language="English", output_language="French", text="I love programming.").to_messages())
# -> AIMessage(content="J'aime programmer.", additional_kwargs={}) You can use the LLMChain with chat models as well: from langchain.chat_models import ChatOpenAI
from langchain import LLMChain
from langchain.prompts.chat import (
    ChatPromptTemplate,
    SystemMessagePromptTemplate,
    HumanMessagePromptTemplate,
)

chat = ChatOpenAI(temperature=0)

template="You are a helpful assistant that translates {input_language} to {output_language}."
system_message_prompt = SystemMessagePromptTemplate.from_template(template)
human_template="{text}"
human_message_prompt = HumanMessagePromptTemplate.from_template(human_template)
chat_prompt = ChatPromptTemplate.from_messages([system_message_prompt, human_message_prompt])

chain = LLMChain(llm=chat, prompt=chat_prompt)
chain.run(input_language="English", output_language="French", text="I love programming.")
# -> "J'aime programmer." You can also use agents with chat models. Initialize an agent using as the agent type: AgentType.CHAT_ZERO_SHOT_REACT_DESCRIPTION from langchain.agents import load_tools
from langchain.agents import initialize_agent
from langchain.agents import AgentType
from langchain.chat_models import ChatOpenAI
from langchain.llms import OpenAI

# First, let's load the language model we're going to use to control the agent.
chat = ChatOpenAI(temperature=0)

# Next, let's load some tools to use. Note that the `llm-math` tool uses an LLM, so we need to pass that in.
llm = OpenAI(temperature=0)
tools = load_tools(["serpapi", "llm-math"], llm=llm)

# Finally, let's initialize an agent with the tools, the language model, and the type of agent we want to use.
agent = initialize_agent(tools, chat, agent=AgentType.CHAT_ZERO_SHOT_REACT_DESCRIPTION, verbose=True)

# Now let's test it out!
agent.run("Who is Olivia Wilde's boyfriend? What is his current age raised to the 0.23 power?") In this example, the agent will interactively perform a search and calculation to provide the final answer. Finally, let's explore using Memory with chains and agents initialized with chat models. The main difference between this and Memory for LLMs is that we can keep previous messages as their own unique memory objects, rather than condensing them into a single string. Here's an example of using a : ConversationChain from langchain.prompts import (
    ChatPromptTemplate, 
    MessagesPlaceholder, 
    SystemMessagePromptTemplate, 
    HumanMessagePromptTemplate
)
from langchain.chains import ConversationChain
from langchain.chat_models import ChatOpenAI
from langchain.memory import ConversationBufferMemory

prompt = ChatPromptTemplate.from_messages([
    SystemMessagePromptTemplate.from_template("The following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know."),
    MessagesPlaceholder(variable_name="history"),
    HumanMessagePromptTemplate.from_template("{input}")
])

llm = ChatOpenAI(temperature=0)
memory = ConversationBufferMemory(return_messages=True)
conversation = ConversationChain(memory=memory, prompt=prompt, llm=llm)

conversation.predict(input="Hi there!")
# -> 'Hello! How can I assist you today?'

conversation.predict(input="I'm doing well! Just having a conversation with an AI.")
# -> "That sounds like fun! I'm happy to chat with you. Is there anything specific you'd like to talk about?"

conversation.predict(input="Tell me about yourself.")
# -> "Sure! I am an AI language model created by OpenAI. I was trained on a large dataset of text from the internet, which allows me to understand and generate human-like language. I can answer questions, provide information, and even have conversations like this one. Is there anything else you'd like to know about me?" In this example, we used a to maintain the context of the conversation across multiple interactions with the AI. ConversationChain That's it! Now you have a solid understanding of how to use LangChain to build end-to-end language model applications. By following these examples, you can develop powerful language model applications using LLMs, chat models, agents, chains, and memory features. Conclusion In conclusion, LangChain is a powerful framework that simplifies the process of building advanced language model applications by providing a modular and flexible approach. By understanding the core concepts, such as components, chains, prompt templates, output parsers, indexes, retrievers, chat message history, and agents, you can create custom solutions tailored to your specific needs. LangChain's adaptability and ease of use make it an invaluable tool for developers, enabling them to unlock the full potential of language models and create intelligent, context-aware applications across a wide range of use cases. Also published . here

A Comprehensive Guide to LangChain: Building Powerful Applications with Large Language Models

While at the I managed to catch some time with , author of , along with a number of game books, role playing games, video games, and the creator and script-writer for . Festival International des Jeux FibreTigre Intelligences Artificielles: Miroirs de nos vies Game of Roles Both his video games and other projects heavily lean on interactive narrative, so given the chance to pick his brains on the topic I couldn’t resist. We started off with a question on procedurally generated fiction, running through to the latest approaches of systemic game worlds. Interactive Fiction to Systemic Worlds The creators of interactive fiction (IF) are thinking about branching narratives (one of the options in my earlier article on ). Players step through the story, get a choice, and continue down a different branch. IF has been around a while in both analogue and digital form, with being widely recognised as the first book in 1930, and and following a few decades later in the 70s. crisis simulations here Consider the Consequences! Zork Colossal Cave Adventure This is where analogies with programming come in. IF provides a decision tree which players walk (or iterate, if you’re the type to save the previous page in the books) through to try and find the optimum solution. “In the 80s, especially around the middle of the 90s, people were experimenting with interactive fiction in object oriented environments.” - FibreTigre Naturally, this then developed further on the digital side as object oriented programming (OOP) languages and concepts became more common. With OOP, instead of creating a decision tree to move from scene to scene, an object could be defined. The object would have a collection of attributes and methods allowing different interactions, such as a wooden chair being able to burn, or a key unlocking a door. At this point the developers are having to create every object (or at least templates) manually, and define their interactions. We have another step before we get into the third generation of IF, where we move to systemic games. In a systemic game, rather than defining each object, challenge, and puzzle, we define a world. This world comes with its own set of rules, physics, and even entities with motivations and feelings. They’re then left to interact with each other, and the player. The best two examples of this, and the classics, are and , with Nintendo’s providing a more structured example with a storyline and plot build-in. Dwarf Fortress Minecraft Zelda: Breath of the Wild Systemic games can be surprising, and bring challenges. Dwarf Fortress is famously complex and open-ended, with failure always an option in fire, flood, plague, invasion, starvation, suffocation, magic, apocalypse, and just about any other possible way your horde of dwarves could die horrible. https://www.youtube.com/watch?v=N5Ct81vsjiM&embedable=true To take one of FibreTigre’s examples, for a mystery game instead of working with a team of writers to build up the mystery, characters with opposing motivations might be released. They might never encounter each other, which raises problems for your game, or they might resolve differences. Superficially at least the world is outside the control of the developers, which is why many systemic games trend towards open world sandbox approaches. To shape a story the way the developers want the world itself needs to be tweaked, with events baked in to drive the narrative forwards. Where does ChatGPT come in? Since it’s currently at the peak of the hype cycle, and given his previous book on AI, I had to ask FibreTigre his thoughts on bringing in a technology like ChatGPT or another large language model (LLM) to enhance the storytelling and conversation capabilities. This is far from a new idea. In 1998 The Digital Village released a game that proudly touted an AI chatbot that could have conversations with the player. While not up to the complexity of ChatGPT, the SpookiTalk natural language parser was surprisingly effective at building immersion. The story was set, but the option for open conversations made it feel otherwise. Starship Titanic Naturally I had to ask about FibreTigre’s thoughts on ChatGPT for storytelling within games. While there has been a lot of hype about the immersion and creativity an LLM can bring to a game, it’s important to note that it’s currently untried and untested in this context. More importantly, we don’t know if there’s a glass ceiling for this approach which will prevent it from maturing. Currently FibreTigre describes it as a child telling stories - simplistic, not necessarily consistent, maybe completely unrelated to the world. Whether it will grow and be developed to the point where these become consistent and usable remains to be seen. Even implementing ChatGPT in games may not be helpful, given the difficulty of generating prompts. While SpookiTalk’s limited capacity made it fairly straightforward to work through conversations anyone who has tried to generate output from ChatGPT they want will know the frustration of trying prompts over and over with minor variations. On this, he believes the problem is more down to the user experience (UX) rather than the output. It’s a problem that will need to be solved if LLMs are going to be the key to more interactive, immersive games and storytelling. Why is Storytelling So Important? Given my own interest in tabletop games, scenarios, and narrative structure, I have to ask FibreTigre whether an aim is to move towards something with the same freedoms as a good tabletop roleplaying game. After a brief segue into worldbuilding, running tabletops, and writing novels, we cross topics with the Metaverse, and what it might mean if it’s ever realised. The Metaverse The metaverse is still poorly defined (or at least debated). FibreTigre has very clear ideas about what a true metaverse will be which calls back to the original source - from Neal Stephenson. In the book, as in FibreTigre’s description, the metaverse is a fully realised world with its own laws of physics, it’s own constraints. Games, events, areas, can be built automatically incorporating these rules - similar in many ways to many of today’s more powerful game engines. Snow Crash The game built within the metaverse doesn’t matter, as those laws can be tweaked or overriden as needed, but ultimately the world is consistent and fully realised separately from anything implemented within it. Only a little way from FibreTigre’s stand with he points me towards another exhibitor . They are working on a hybrid of metaverse and tabletop roleplaying which he recommends I look at. Input from those running games is required, but I can quickly see how the capabilities are rapidly approaching what he describes as the metaverse. Elder Craft Menyr If you’d like to follow more of my ramblings, then you can find almost everything I do by visiting my , or of course subscribe to my writings on . LinkTree HackerNoon

Games, ChatGPT, Interactive Fiction and the Metaverse with FibreTigre

Self-Supervised Learning is getting attention because it has the potential to solve a significant limitation of supervised machine learning, viz. requiring lots of external training samples or supervisory data consisting of inputs and corresponding outputs. Yann LeCun¹ recently in a Science and Future Magazine interview presented self-supervised learning as a significant challenge of AI for the next decade. Humans, for instance, can determine the semantic meaning of the word “orange” from context when it appears near “t-shirt,” “fridge,” “county,” or “mobile.” Similarly, in machine learning, the predicts the semantic context of a word based on surrounding words. The research behind self-supervised learning follows the same principle of automatically identifying, extracting and using supervisory signals. Word2Vec algorithm What is Self-Supervised Learning? Self-supervised learning is autonomous supervised learning. It is a representation learning approach that eliminates the pre-requisite requiring humans to label data. Self-supervised learning systems extract and use the naturally available relevant context and embedded metadata as supervisory signals. Cats continue to play an essential role in everything significant in machine learning. Self-supervised research “ ” predicts the positional location of one rectangular section of an image relative to another by using spatial context as a supervisory signal for training a rich visual representation. For instance, the right ear of a cat would be in the top-right position relative to the eyes of a cat. This approach allows learning about cats, dogs, or buses without prior explicit semantic labeling. Unsupervised Visual Representation Learning by Context Prediction Self-supervised learning in images (Photo by ) Quino Al Self-supervised learning, fortunately, is not limited to learning from visual cues or associated meta-data in cat images or videos and has use cases beyond computer vision. Self-supervised vs. supervised learning Self-supervised Learning is supervised Learning because its goal is to learn a function from pairs of inputs and labeled outputs. Explicit use of labeled input-outputs pairs in self-supervised learning is not needed. Instead, correlations, embedded metadata, or domain knowledge available within the input is implicitly and autonomously extracted from the data and used as supervisory signals. Like supervised learning, self-supervised learning has use cases in regression and classification. Self-supervised vs. unsupervised learning Self-supervised learning is like unsupervised Learning because the system learns without using explicitly-provided labels. It is different from unsupervised learning because we are not learning the inherent structure of data. Self-supervised learning, unlike unsupervised learning, is not centered around clustering and grouping, dimensionality reduction, recommendation engines, density estimation, or anomaly detection. Self-Supervised vs. semi-supervised learning Combination of labeled and unlabeled data is used to train semi-supervised learning algorithms, where smaller amounts of labeled data in conjunction with large amounts of unlabeled data can speed up learning tasks. Self-supervised learning is different as systems learn entirely without using explicitly-provided labels. Why is self-supervised learning relevant? Self-supervised learning is essential for many reasons but mainly because of shortcomings in both approach and scalability of supervised learning. Supervised learning is an arduous process, requiring collecting massive amounts of data, cleaning it up, manually labeling it, training and perfecting a model purpose-built for the classification or regression use case you are solving for, and then using it to predict labels for unknown data. For instance, with images, we collect a large image data set, label the objects in images manually, learn the network and then use it for one specific use case. This way is very different from the approach of learning in humans. Human learning is trial-based, perpetual, multi-sourced, and simultaneous for multiple tasks. We learn mostly in an unsupervised manner, using experiments and curiosity. We also learn in a supervised manner but we can learn from much fewer samples, and we generalize exceptionally well. For supervised learning, we have spent years collecting and professionally annotating tens of millions of labeled bounding boxes or polygons and image level annotations, but these datasets , , , and Microsoft collectively pale in comparison to billions of images generated on a daily basis on social media, or millions of videos requiring object detection or depth perception in autonomous driving. Similar scalability arguments exist for common sense knowledge. Open Images PASCAL Visual Object Classes Image Net COCO Self-Supervised Reinforcement Learning A dog trainer can reward a dog for positive behavior and punish for negative behavior. Over time the dog figures out and learns actions it took to get a reward. Similarly, in Reinforcement learning, a navigating robot learns how to navigate a course when rewarded for staying on course and punished when it collides with something in the environment. In both cases, this reward and punishment feedback reinforces which actions to perform and which to avoid. Reinforcement Learning works well in the presence of a feedback system for rewards. It also requires a comprehensive set of training data and may be impractical when considering the cost, time, and the number of iterations required before succeeding. In the absence of rewards-based feedback system, a dog or a navigating robot may learn on its own by curiously exploring the environment. Researchers from BAIR² created an “Intrinsic Curiosity Model,” a self-supervised reinforcement learning system that can work even in the absence of explicit feedback. It uses curiosity as a natural reward signal to enable the agent to explore its environment and learn skills for use later in its life. See reseach at: and . Curiosity-driven Exploration by Self-supervised Prediction the accompanying writeup Using self-supervised learning machines can predict through natural evolution and consequences of its actions, similar to how newborns can learn incredible amounts of information in their first weeks/months of life by observing and being and curious. Self-supervised learning has the potential to scale learning to levels required by new use cases including but not limited to use cases in medicine, autonomous driving, robotics, language understanding, and image recognition. Self-supervised learning brings us closer to human-like autonomous learning. Self-Supervised Learning has found success in , by using motion segmentation techniques to determine relative depth from geometric constraints between the scene’s motion field and camera motion. estimating relative scene depths without human supervision In medicine, it has found use cases and in . robotic surgery dense depth estimation in monocular endoscopy In autonomous driving, self-supervised learning is useful for . It has a use case in . estimating the roughness of the terrain when off-roading depth completion using LiDAR and monocular cameras To explore latest self-supervised learning use cases, see this search . link ¹ Yann LeCun — Professor NYU, Chief AI Scientist for Facebook AI Research (FAIR), and Director AI Research at Facebook. July 2018: https://www.sciencesetavenir.fr/videos/yann-lecun-explique-lintelligence-artificielle-et-ses-defis-a-venir_kzrzpf ² BAIR — Berkeley Artificial Intelligence Research: http://bair.berkeley.edu

Self-supervised learning gets us closer to autonomous learning

Libraries were not among the earliest adopters of artificial intelligence (AI), but more are increasingly using the technology in various ways. Librarians have found it can improve people’s experience and make their jobs a bit easier. What changes has AI made so far? 1. Streamlining Librarians’ Workflows Some institutions quickly provided resources explaining how library professionals and others could use artificial intelligence. Northwestern University offers a its librarians and faculty. It’s an excellent starting point for people open to trying AI but still need to figure out how to get started. continually updated resource list for Elsewhere, an Iowa school district complied with recently enacted state laws about banned books using AI to determine which titles to pull. Officials removed to the books’ sexual content. Authorities pointed out that classroom and school libraries have huge collections sourced in various ways. It’s unfeasible to read every page and look for content outside of what the law permits. 19 titles from school libraries due AI tools from photographs. Those are convenient for librarians who want to edit images before publishing them on social media or elsewhere. can also remove unwanted objects 2. Improving Interactivity for Visitors Some libraries also use AI to enhance the visitor experience and make it more interactive. Singapore’s Central Public Library has an Immersive Room where people to create interactive stories. They select the characters, genre and location while working within the parameters of six popular tales. The story is then displayed across a wall of six curved panels. Patrons can take them home to enjoy later by scanning QR codes and downloading them to their devices. can use a generative AI-based tool China’s Changning Library has capitalized on interactivity, too. A human avatar equipped with a recommendation algorithm scans people’s library cards or faces to get their reading histories and suggest relevant titles. Robots then deliver those books to patrons. The library introduced these AI features in the autumn of 2022. Between then and August 2023, library visitors have , indicating their popularity. Besides using artificial intelligence, staff members can rely on internal data in numerous ways customers will appreciate. Digging into data to learn more about community demographics or favorite titles can show library professionals where and how to spend their resources. used them more than 20,000 times 3. Providing Librarians With Opportunities to Teach Critical Thinking Although artificial intelligence can do some incredible things, it’s not perfect. Generative AI tools that users might treat as factual. sometimes offer wholly incorrect information The information sounds accurate, but this problem can become dangerous when people don’t know better. Librarians are well-positioned to help them recognize and work with AI’s shortcomings. With its nearly century-long history, Sharjah Library is the United Arab Emirates’ oldest. It harnessed digital technologies in 2020 by offering all its online resources to the public. The collection includes 5 million academic titles, 160,000 e-books and 30,000 videos. That digital transition in library membership. caused a 70% increase Library director Eman Bushulaibi believes that although AI and other technologies have brought changes, libraries will always have value as locations for exchanging and gaining knowledge. Bushulaibi and her staff create safe spaces where people can use and develop tools to think critically about the information they come across. 4. Exploring the Technology’s Future Role in Libraries Most career paths require professionals to adapt. Those working in libraries have recently experienced that as AI tools became more accessible. Technology in libraries, ranging from architectural and aesthetic changes to radical shifts in education and research capabilities. has spurred many changes A University of Texas at Austin program aims to reduce the learning curve. The grant-backed initiative will train to help them bring artificial intelligence and data science to libraries. nine doctoral students for three years This effort came about due to feedback from librarians and library science students getting assistance from tech experts. They said those with technical know-how often lacked the experience and knowledge of working in libraries. This three-year program will put those with AI and data science backgrounds in public, academic, and school libraries to give them much-needed perspectives. It relies on a rotation-based approach inspired by medical students' training in real-world environments. It’s too early to say how libraries might use artificial intelligence in five years. However, program participants could gain information to shape that future. 5. Updating Class Curriculums Many libraries offer free or low-cost classes to community members, which are taught by librarians, visiting scholars, and others. Whether someone wants to improve their language skills, learn fitness tips, or prepare for college, there’s a good chance libraries have themed events to help. Some librarians have examined how they could use AI tools to update their teaching methods. Lauren Todd has over a decade of experience as an engineering subject librarian at Washington University in St. Louis. Her job often allows her to go into engineering classes on campus and help students realize how the library’s resources could assist with their assignments. When teaching the students of a mechanical engineering and material design class, Todd quickly locate the standards engineers must follow. Some learners created prompts to determine the relevancy of specific standards to their design projects. found ChatGPT helped learners Todd acknowledges that ChatGPT is imperfect but can make library research less frustrating for people needing highly specific information. There’s no centralized database of engineering standards, so even seasoned librarians need significant time to find ones that may relate to students’ coursework. Librarians Can Identify the Best Use Cases These are some of the most appropriate ways to use artificial intelligence in libraries. However, more will undoubtedly emerge as the technology becomes more advanced and accessible. Library professionals can apply their skills and knowledge to decide which applications best suit their work and library visitors’ needs and which need further improvements to make them worthwhile. People who work at, use, or have other relevant associations with libraries must keep their minds open during this period of significant change. AI will allow people to do things differently, but that doesn’t necessarily mean they should. Individuals must use their knowledge and the advice of trusted individuals to determine whether artificial intelligence enhances certain library-related tasks.

5 Ways Artificial Intelligence Is (Quietly) Changing Libraries

5 উপায় কৃত্রিম বুদ্ধিমত্তা হল (নিভৃতে) লাইব্রেরি পরিবর্তন করা

5 Wege, wie künstliche Intelligenz Bibliotheken (still und leise) verändert

Cinco formas en que la inteligencia artificial está cambiando (silenciosamente) las bibliotecas

5 façons dont l’intelligence artificielle change (tranquillement) les bibliothèques

5 तरीके जिनसे आर्टिफिशियल इंटेलिजेंस (चुपचाप) लाइब्रेरीज़ को बदल रहा है

人工知能が図書館を（静かに）変える 5 つの方法

인공지능이 (조용히) 도서관을 변화시키는 5가지 방법

5 maneiras pelas quais a inteligência artificial está mudando (silenciosamente) as bibliotecas

5 способов, которыми искусственный интеллект (тихо) меняет библиотеки

Yapay Zekanın Kütüphaneleri (Sessizce) Değiştirmesinin 5 Yolu

5 cách trí tuệ nhân tạo đang (lặng lẽ) thay đổi thư viện

人工智能正在悄然改变图书馆的 5 种方式

In 2017, a got tired of its job and committed . Then, in 2022, a , after interacting with LaMDA, the company's chatbot, that technology had become conscious. Recently, it was reported that in 2021, a Tesla software engineer was attacked by an uncontrolled robot. This is without considering the fear that machines will steal jobs, or the classic fiction that machines will take over the world. robot suicide in a water fountain Google engineer declared Although we don't see machines as replicants in Blade Runner, people are predisposed to anthropomorphizing or attributing human qualities to non-humans [1]. It's a way of being able to analyze a scenario and bring it into our understanding: we give names to our boats and tornadoes, we talk to our pets, and say that the pleasant feeling they give us is love. It's easier to understand a system that declares in the first person "I'm sorry, I don't know how to help you" than to accept that the model we're talking to may be nothing more than a stochastic parrot. [2] The anthropomorphized interaction, therefore, relates to the phenomenon of "illusion of agency", a psychological concept that refers to the tendency to attribute autonomous action or control to entities that, in reality, do not possess such capabilities. In other words, it's the mistaken perception that an agent (be it a person, an animal, or a machine) is acting of its own free will, when in fact its actions are being determined by external factors (in the computational case, development rules). In situations where AI systems make flawed decisions or actions, there is a tendency to see these errors as "choices" of the entity rather than results of its programming and design by humans, not least because it has been reported that the selfish bias of a human being that leads someone to see the cause and responsibility for a negative outcome as not attributable to themselves can be observed even in human-machine interactions [3]. This change in perception is dangerously inclined to absolve human creators, operators, and regulators of their responsibility, not as a matter of judicial regulation (which remains a gap and has challenges in its realization not only because of the complexity of the subject but because Artificial Intelligence is often understood only as machine learning and the premise is not properly structured. Do we need stricter regulation? Do we need to take more risks?), but on a techno-ethical issue. Let's take a more extreme but true event from 2023: a user who was sentimentally attached to a chatbot committed suicide after sharing his thoughts with the bot and r to "turn your words into actions", as well as other messages. Would a court conviction of the developers of this product cause another user to behave in the same way with another chatbot, assuming this one has been deactivated, if the content of the messages and the affection are the same? It's not just a legal situation. It's a social, psychological, and technological education issue. eceiving the response The concept of humanizing AI is ambiguous, and a significant challenge lies in the absence of a universally accepted approach that dictates the best practices for designing and utilizing AI. While an interface that mimics human behavior can be more approachable, no clear boundaries are defining what should or shouldn't be done in a product. Ultimately, user rejection becomes the only limiting factor, although potential harm may manifest before the interface becomes too unfamiliar. A user-friendly interface is a reduction in the complexity of the system that operates behind it. But as long as there is no education about how systems work, the user will not be able to think critically about what they use. This doesn't mean that everyone should become a programmer, but at least understand that the output on their screen comes from a path of data collection, model development, and design. There is a set of rules taking place. Since humanization is an almost unconscious act on the part of us users, let's at least limit it with a little knowledge. And it's not easy to avoid anthropomorphism when communicating about AI, especially considering the standard language of the industry, the media, and everyday life itself: machine , computer , generative . Perceptions of Artificial Intelligence can be influenced by the specific use of language. learning vision modelling How AI is presented has "concrete impacts", particularly on the way people distribute responsibility and recognition for the work done. When described merely as a tool in the hands of humans, it tends to attribute greater responsibility and merit to certain individuals - such as the person operating the code. On the other hand, if AI is characterized with human traits - such as the ability to create - then it is seen as deserving of greater credit and responsibility, like an agent with independent thought and mental capacity. [4] When we read news stories about incidents or atypical events involving Artificial Intelligence, we often come across these terms. Also, the attribution of "intelligence" means that more blame or credit for the outcome of a task was attributed to a robot with autonomous behavior than to a non-autonomous robot, even if the autonomous behavior did not directly contribute to the task. [3] These studies therefore suggest that humans can assign responsibility to a computer or robot based on the machines' anthropomorphised mental capacities. The humanization of machines not only shifts the distinction between the responsibility of the device and that of its human creator, but the attribution of intentions or consciousness to a machine at the same time blurs the boundaries of what constitutes true autonomy and consciousness. However, the difficulty in imputing humanity and sentience to a machine lies not only in the fact that Artificial Intelligence is not capable of having it and when it says, for example, that it feels fear, it is actually emulating what it has learned, repeating that phrase without any kind of essence behind it. Even today, there is a heated debate about how to define consciousness. Our consciousness, as humans. Our understanding of how the brain works is still quite limited. We have considerable knowledge about fundamental chemistry: the way neurons activate, and the transmission of chemical signals. We also have a good understanding of the main functions of various brain areas. However, we have very little knowledge about how these functions orchestrate us. To some extent, theoretical speculation has supplanted detailed neurophysiological studies of what  happens in the brain. But what about beyond that? [5] Why do we have this magical notion of ourselves? Why does the same experience affect us differently? Is the same feeling felt in the same way by all of us? If being a human is something that, although we experience it, we still don't have a full understanding of what it is as a whole, how can we say that a machine also experiences this complexity? By elevating machines to human capabilities, we diminish the special character of people. For the 2023 Jabuti Prize, one of the highest honors in Brazilian literature, the Brazilian Book Chamber (CBL) decided to disqualify Frankenstein, an edition of the 1818 classic, from the Best Illustration category, since the artist reported having used AI solutions to develop the art. Ironically, one of the books recognized by the Prize in the Non-Fiction category deals with the impact of Artificial Intelligence on human beings (“Humanamente digital: Inteligência artificial centrada no humano”, something like "Humanly Digital: Human-Centered Artificial Intelligence" in English, by Cassio Pantaleone). On the one hand, we recognize the intertwining of the human experience with machines. On the other hand, we still haven't managed to validate whether an algorithm used as an art tool can be considered a valid method of creation or not, even though the artistic process, even if carried out by machine learning, requires the action (and appreciation of beauty and aesthetics) of a human being. Machines don't steal jobs if they aren't used indiscriminately to do so. Machines don't kill unless they are used as weapons. Nor do machines suffer or empathize, although their texts emulate this since they have been trained with data from us, loaded with feelings that only we can truly feel. They are almost the modern version of the Golem myth. How can humans relate to non-human intelligence? Anthropomorphism is a valid answer, but not the only one. And when it is used, it cannot absolve us of the real responsibility for its consequences, whether appropriate or not: us. Artificial Intelligence is, in the end, a mirror of ourselves. And if we're afraid of where it's going, it's actually because we're afraid of the path we're going to create. REFERENCES [1] Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: A three-factor theory of anthropomorphism. Psychological Review, 114(4), 864–886. https://doi.org/10.1037/0033-295X.114.4.864 [2] Shneiderman, B. & Muller, M. (2023) . On AI Antropomorphism https://medium.com/human-centered-ai/on-ai-anthropomorphism-abff4cecc5ae [3] Kawai, Y., Miyake, T., Park, J. et al. Anthropomorphism-based causal and responsibility attributions to robots. Sci Rep 13, 12234 (2023). https://doi.org/10.1038/s41598-023-39435-5 [4] Epstein, Z., Levine, S., Rand, D. G., & Rahwan, I. (2020). Who Gets Credit for AI-Generated Art? In iScience (Vol. 23, Issue 9, p. 101515). Elsevier BV. https://doi.org/10.1016/j.isci.2020.101515 [5] Goff, P. (2023). Understanding Consciousness Goes Beyond Exploring Brain Chemistry. Scientific American https://www.scientificamerican.com/article/understanding-consciousness-goes-beyond-exploring-brain-chemistry/ The following prompt generated the image used in this article: “Create a man looking in a mirror, but the mirror is shaped like a robot.” Sofia × DALL·E

No, You Are Not Afraid of Artificial Intelligence. You’re Afraid of Other People

না, আপনি কৃত্রিম বুদ্ধিমত্তাকে ভয় পান না। আপনি অন্য লোকেদের ভয় পান

Nein, Sie haben keine Angst vor künstlicher Intelligenz. Du hast Angst vor anderen Menschen

No, no le tienes miedo a la inteligencia artificial. Tienes miedo de otras personas

Non, vous n'avez pas peur de l'intelligence artificielle. Vous avez peur des autres

नहीं, आप आर्टिफिशियल इंटेलिजेंस से नहीं डरते। आप अन्य लोगों से डरते हैं

いいえ、人工知能を恐れることはありません。他人が怖い

아니요, 당신은 인공지능을 두려워하지 않습니다. 당신은 다른 사람을 두려워합니다

Não, você não tem medo da inteligência artificial. Você tem medo de outras pessoas

Нет, вы не боитесь искусственного интеллекта. Вы боитесь других людей

Hayır, Yapay Zekadan Korkmuyorsunuz. Başkalarından Korkuyorsunuz

Không, bạn không sợ trí tuệ nhân tạo. Bạn sợ người khác

不，你并不害怕人工智能。你害怕别人

In the of the analysis, we will bring back the that play on the ZKP stage. The part is dedicated to showing the “ ” of the ZPK scene. And , we will appeal of “possible” luck from the intruder. Now, let’s present the nature of the abstract and let’s unfold its mysteries. first part actors second magic thirdly rare probabilities Abstract We’ll be using an as the on how to open a mystical door. The nature of the abstract is to turn (such as left/right) into . The mechanism movers are unique to each instance of opening. The the of the structure to the shared . The abstract works when tangible and intangible . The abstract is an which ties them both together to form the . If you play with and have of the ones, you will have chances of the key. Imagine guessing the position of an respective to an . abstract shared knowledge decision gates mechanism movers abstract connects tangible rules intangible knowledge only both exist upper system tangible rules no knowledge intangible close to 0 guessing unknown number x unknown group of y numbers Part 1: The actors As the curtain rises, it reveals to us 3 actors. The is the and , is the and and is . When describing them, the aim is to provide a clear and structural analysis of them, any information regarding the . I will only refer to it but not provide any information until the second part. door giver of access holder of the lock Lisa opener bearer of the key, Bob the verifier outside structure of the lock The door The door is the of the abstract . In its internal knowledge, we find the logic and of the lock. The refers to the required to open each to the by the . Those locks after each opening, ensuring that any given by Lisa can be used for the (which will be passed after a failed/successful attempt). bearer lock hard-coded structure abstract system comparing internal steps individual lock information presented opener iterate successful information strictly current lock If a is reached, because Lisa made a and the , will begin and Lisa will have to provide a based on her knowledge of the . Based on its , the door will depict whether Lisa holds or not the key to enter. critical point mistake passed opening point another lock iteration new key overall (abstract) system internal gears Lisa Lisa, as the bearer of the key, holds knowledge of the and the structure. She knows the of the lock and will her knowledge of that in the form of to the door. She will the lock as if it in front of her. Stating movements such as l which will be the the door will receive and the of the lock accordingly. Once Lisa is sure the lock is in the position, she will state the and . internal system current lock’s limits present logical reasoning imagine opening sits eft/right input adjust internal gears open current key end communication value Bob Bob, as the verifier, can form the of a . Let’s say that Lisa comes and a conversation with the door. She will further state: Left, Right, Left, Left, Right, Right, Right, . After which, she will . Bob now knows how to open of the door. However, has a distinct route to be taken. image potential lock initiates Key end communication iteration 1 iteration 2 As the day passes Bob gathers the information about how to open all . Now, the comes into play and all the 500 iterations. The next day, becomes of the . Now, the will play a crucial role in of guessing. 500 iterations system shuffles Iteration 1 any other 499 logic of shuffling reducing the probability Remember when I said “One could always check the security by checking the memory. Then, questions about gibberish memory come into play to scare off the intruders who may want to bypass. Alluding later to the sense of courage and self-esteem as you broke down 1TB of key complexity.”? For this system to work, it requires that and the have a kind of inside them. of the Door allows for each lock to be and later . Now, the shuffling can come each day or based on . So the door also has to have some kind of (which I will further call the reason of the system) of . both Lisa Door computer Memory uniquely stored shuffled manually algorithmic logic logic shuffling Lisa and all other authorized personnel could very well gather their knowledge of from a “Headquarters”. today’s lock system Or they could have as the Door inside their heads in the case we to create a headquarters. similar shuffling reasoning do not want This allows for both and storage and interactions. A headquarters will allow to gain access. No headquarters would mean that only would be able to open the door. centralized decentralized new personnel limited predefined personnel Both avenues open deeper implementation as a as it could allow to be granted access by providing her with a . decentralized system, newcomers by Lisa unique encrypted key But now I may get too far ahead of myself. Let’s see how the is made. abstract lock Part 2: The lock Maybe is the simplest part to understand. proving it is, however, out of my reach and expertise, but I will try to provide a clear and structural analysis that could allow for mathematical understanding and implementation. the lock Mathematically 1. The prime-nonprime lock system. 1.1 Prime-nonprime lock cipher The cipher at play is made out of . two sets A set of only numbers and a set of only . prime non-primes Both sets have an . equal number of elements Each set can start at an number as long as the . arbitrary previous rule holds true , the set will obtain ‘ ’ as its , whereas the set will obtain ‘ ’ as its . After they are formed prime 3 last element non-prime 2 first element So we have: Prime=[prime, next prime, next next prime…(30 times),3] Non Prime= [2, NP, next NP, next next NP… (30 times)] And pretty much this is the lock. 1.2 Cipher interaction Let’s suppose Lisa comes and tries to unlock the door. We assume that she knows the and (except 3) are in the set. As well as non primes. She knows (except 2) and we find out she how many non-primes there are (we do not want to waste information on what is locally known from the primes set). first prime how many primes the first non prime already knows So the of each or will require knowledge of the from and the from (All the primes and non-primes that exist in math). To later it to the required which to the set of the (which only the authorized personnel will know of). internal knowledge opener intruder set of primes x to y set of non-primes x to y adjust parts correspond actual cipher So Lisa knows supposedly all primes and non-primes. Both up to infinity and divided in the required by the . respective set current lock She, as the authorized personnel, will know that of the starts with the and has (+3 at the end) and the starts with the (+2 at the beginning) and holds 100 (+1) non-primes in total. iteration 1 prime set 30th prime 100 primes non prime set 100th non-prime For a very basic interaction, I will present to you and their interactions with . The text is what know of. The is what is , what Bob can hear. both sets Lisa bolded only the door and the authorized personnel italicized text given outside Iteration 1: Prime set = [5,7,11,13,17,19,(3)] Non Prime set = [(2),4,6,8,9,10,12] Lisa comes to the door and states: ‘ ’. This is the . Now, the door knows that Lisa is aware of . 2,3 initiation both locks inside The process of opening starts. The ‘selected’ prime is 3. The ‘selected’ non-prime is 2. (This is a rule of the internal logic) Lisa says: ‘3, left’. Now, the gears of the prime selection turn to 19. ‘2, right’ The gears of the non prime point now 4. ‘3, left’ (17) ‘2, right’ (6) ‘3, left’ (13) ‘2, right’ (8) Lisa says: Now, the lock has reached a middle place, 13 is selected as the prime and 8 as the non-prime. Lisa has to state their difference (5) and end communication with the door for the door to run its own rules and check. ‘5; 3,2’. Now the door will run internally and check the difference between the middle numbers of each set, if it corresponds to the number specified by Lisa, access is granted. If Lisa however fails, another iteration begins and Lisa has to take a different route. This new deeper layer of failure requires both centralized and decentralized systems. If Lisa was sent from the headquarters and only knew how to open iteration 1, she would have to return to regain information about the second. If the iteration is based on a rule of background iterations (in the case of repeated failed opening), then the iteration will turn instead of 2 to 1.1 iteration. Assuming that Lisa also gained knowledge from the headquarters on how the lock will change in case she fails to open it. (Assuming the cycle doesn’t repeat) At iteration 1.3 (after 3 failed attempts) the pace of the “left/right” decision could turn from 1 step to 3 steps and the newly introduced sets will have a total number of 3n+1. Ensuring that no matter how hard you try, you can never select the middle point of the sets. Who knows however if this is necessary. If we allow the locks to repeat themselves we could envision such a protection: After you state “left” when you are positioning the prime selection and you are located at the first prime, you will be sent back at 3. This would assume you are also positioned at the last position of non-primes and thus, stating “right” would turn you to 2. (Or we could never tell the opener that they failed so they will infinitely try to guess, but that of course, would give of no knowledge of whether someone tried to enter and failed or not.) The ‘3,2’ state was achieved and no key was mentioned so the input was wrong, access was denied and the iteration moved to the next lock. 2. Metaphysics (Creativity/Imagination/ ) vs Mathematics (Present state) Possibility The and seem to . What does that mean? It means that we could use and to stop any “ ” or avenues that aim to break the system. It’s not a problem of , but one of of the set limits. Even if you know you can do, you can never know if that 1TB holds 1000 sets of 20 to 1000 numbers of 100 sets of 10.000 numbers. And thus, Especially since those 10.000 numbers could very well be sets that follow one another. possibilities of interaction security measures stop at the power of imagination (endless routes could be taken) reason hidden information creative newly taken computational problem prior knowledge all the actions you will never know your starting point. I will present a view on “cipher interaction” depicted with the help of ChatGPT. (I will lend my personal view in bold and italicized) “The mention of "the power of imagination" in the context of interaction and security measures implies that the system's robustness relies on the ability to anticipate and counteract potential creative or innovative attempts to breach its security . (Any route taken, no matter of creative or smart it is, is not tied to the “ground-truth” knowledge of the system) It suggests that the system designers have incorporated elements that make it resistant to unconventional or unexpected approaches by leveraging reasoning and hidden information. In practical terms, this means that the system is designed in such a way that even if an attacker tries to think outside the box or employ novel strategies to break the system, there are safeguards in place to mitigate such attempts. These safeguards may include: : The system's structure and rules are sufficiently complex and dynamic, making it difficult for an attacker to predict or exploit weaknesses effectively. Complexity : Certain critical information about the system's configuration or operation is kept hidden or encrypted, preventing attackers from gaining a comprehensive understanding of the system's vulnerabilities. Hidden Information : The system can adapt or evolve in response to emerging threats or attack strategies, ensuring that it remains resilient over time. Adaptive Security Measures : Attackers are constrained by their prior knowledge of the system's set limits, preventing them from exploiting vulnerabilities that they are not aware of or cannot anticipate.”- ChatGPT Limits of Prior Knowledge Imagine that you are playing chess. However, you can’t see your opponent’s pieces in exchange for him not seeing yours. This play state could be akin to You do not intend to be offensive at all, in the exchange of being private and aware of your flaws. everybody minding their own business. If during the state of the game, any player sees only once the position of the pieces, then that player will gain relevant irreversible information of position. Without it… who knows how many chess boards are between his pieces and your king. Part 3. The Luck of opening such a system I want to first mention that everything I wrote is simply an idea. Until I see it in action, I will disbelieve it as true. And even if I see it in action, who knows if I will believe it to be the true case. I will still require knowledge of the background sets to know what the opener has done. Now, what do I mean by “Luck”? Let’s turn back to another story. “What can we say about “efficiency” when it comes to our safety? Our long-term health and well-being? Why does “efficient” sound a little too rushy here? Because here, we need proof. We have to be sure as our health is at play.” It indeed seems like has a aspect to it. How could we tie efficiency to our locks? efficiency rushy Well… let’s say that the door only requires the key ‘ ’ and the ‘ ’ ending to open. All the internal structure will itself to the middle of the sets and check. This way, we would the number in novel ways. Lisa could simply state “ ” which for the door would mean “ ” and if the difference between the middle points is indeed 51, access is granted. X 3, 2 run encrypt abc 51 I assume that this approach could offer the same kind of information as before. But then, wouldn’t there be the possibility for someone to get that information out of Lisa? I mean, if they know towards which door she is going they could simply go there, and… maybe we could even implement more doors in this case. What if the quantum world is guided in the end by the classics? I don’t know if it’s just me, but I believe that there exists a . A specific Mathematical that, as long as it is , could drive us into of paving a path for . Like the distribution of primes. Each time you get to a , the pattern , but, it . specific system process kept hidden endless ways true privacy new prime overall changes the slightest changes Based on what we see, 2,3,5,7, and so on, those , of numbers out of the . first prime numbers cut a lot natural number line The the primes we find, the the numbers they . Yet, when talking about , isn’t the same as the part? bigger less influence infinity half 1000th Personal Infinity Could we form a “ ” that works as our ? personal bounded infinity workspace The is and to that . If know those , then only we can depict that to at the to it. overall change static tied infinity only we boundaries overall change stop last prime required exhibit The of someone the is very low since they have to take into account the (Which are at and (but serve as guiding rules) in the process of ). The simplest form of drawing a line is placing 2 points and connecting them. chance exhibiting exact median change starting and end points given among authorized parts creation hidden interaction Now the question comes: Could this be some kind of axiom in math? Or a rule that quantum mathematics could follow? When I say quantum mathematics I mean the interaction between two distinct points of view. Those we have to find the between them, knowing about the , or of the in which they sit. (Those points however cannot move as it would require a meeting and re-organizing among the authorized personnel) two points measured distance nothing shape size, color space Part 4: The synergy Now, extended to people, this space could be comprehended and understood. When it comes to AI or any other computational rule, it can only be analyzed. But what if there is nothing else to analyze except the existence of those two points? Even people peaking into the play cannot see the system. This could be a point where we could ensure protection through a form of “limited” consciousness. Understanding of the system itself (which is different from self-understanding). Mathematical rules of a circle, reflection, inner, out, and so on, all enveloped to form a limited self-awareness. With the lack of human power (acknowledging it can be more) the system will remain forever overall static. Even if the rules that guide it are moving and shaping interactions and outcomes. If those two points were to be humans, their unique perspectives would allow them to potentially observe one another and find smart ways to guess the difference between them. Maybe, they could never guess the exact classical value. But that act of guessing could be akin to seeing the rules. Bots would remain stuck there. But humans could draw. Even if we use sophisticated AI for Lisa, the Door, and Bob, they will still not be able to arrive at a conclusive judgment as there will always be insufficient information. If I am to sum up the whole article in a single phrase, I will say: Bots in the end could create the fastest algorithms we’ve had, but they could never surpass our creativity. I mean… it’s important to keep dreaming. The future looks amazing. But there’s still a lot of work to do x.x (This article is dedicated to Dan. Without your help, my schedule (and also my mood) wouldn’t have allowed to me focus on all those points. Thank you!)

Unlocking the Magic of Zero-Knowledge-Proofs—Part 2: Artificial Reasoning and Limits of Information

জিরো-নলেজ প্রুফের ম্যাজিক আনলক করা—পার্ট 2: কৃত্রিম যুক্তি এবং তথ্যের সীমা

Die Magie von Zero-Knowledge-Beweisen erschließen – Teil 2: Künstliches Denken und Informationsgrenzen

Descubriendo la magia de las pruebas de conocimiento cero: Parte 2: Razonamiento artificial y límites de la información

Libérer la magie des preuves sans connaissance – Partie 2 : Raisonnement artificiel et limites de l’information

शून्य-ज्ञान प्रमाणों के जादू को खोलना-भाग 2: कृत्रिम तर्क और सूचना की सीमाएँ

ゼロ知識証明の魔法を解く—パート 2: 人工推論と情報の限界

영지식 증명의 마법 풀기 - 2부: 인공 추론과 정보의 한계

Desbloqueando a magia das provas de conhecimento zero – Parte 2: Raciocínio artificial e limites de informação

Раскрытие магии доказательств с нулевым разглашением — Часть 2: Искусственные рассуждения и ограничения информации

Sıfır Bilgi Kanıtlarının Büyüsünün Kilidini Açmak - Bölüm 2: Yapay Muhakeme ve Bilginin Sınırları

Giải phóng sự kỳ diệu của những bằng chứng không có kiến thức—Phần 2: Lý luận nhân tạo và giới hạn thông tin

解锁零知识证明的魔力——第 2 部分：人工推理和信息限制

Not to sound pompous, but this has been an extraordinary year for the field of artificial intelligence (AI), in more ways than what many experts would have imagined. In a short span of time, artificial intelligence and generative AI platforms have become game changers for many businesses, across a broad range of industries. For years, business leaders have been trying to improve their efficiency and boost productivity, while simultaneously improving their bottom line and reducing costs - now artificial intelligence is doing exactly what they have been looking for all this time. A small business recorded that nearly half, or around 48 percent of small businesses and medium-sized enterprises have already started using AI tools in some capacity in the past year. AI adoption survey It’s safe to say that the adoption of more advanced artificial intelligence tools and generative AI applications has been fairly successful, especially for smaller, more agile companies looking to get their foot in the door, while keeping their balance sheets intact. A Few Wrongs Make for A Terrible Mistake While there were some monumental moments this year that most people will remember as the field of artificial technology continues to develop, this wasn’t without a few or several memorable blunders. At the start of the year, several well-known technology and publication companies found themselves in hot water after publishing countless AI-generated articles that contained factual errors and cases of plagiarism. Quietly removing a few articles from a company website or blog is a lot easier than having to make up more than value losses. In late February, Google launched its generative AI platform, Bard, however, not everything went according to plan, and a few mistakes led to billions being wiped off the big tech’s market value within hours as share prices tumbled. $100 billion in market A flood of deep fakes surfacing on the internet, political leaders using leading up to the presidential elections, and a flurry of lawsuits had many companies up in arms this year as they tried to tame a wild beast running rampant in the streets. AI-generated imagery as part of their campaigns Not to mention former OpenAI CEO and ChatGPT developer, Sam Altman being booted by the same company he helped bring onto the map. Altman would later land a job at Microsoft, however, in a sudden change of direction, a new board of directors at OpenAI had managed to , offering Altman the opportunity to reinstate his role as chief executive at the company. reach an agreement of principles While artificial intelligence has been enjoying its moment in the sun, quite literally, the year ahead only brings more opportunities and new challenges. However, for small and medium-sized companies, with less experience, knowledge, and legal resources compared to their more established peers in Silicon Valley, running into a wall of trouble when adopting new artificial tools and systems can be a bigger, and more costly mistake than simply removing unfounded content from the internet or firing an intern. Problems that Organizations Need to Avoid when Building an AI Strategy Bringing AI into your business can be an exciting opportunity to experiment with new possibilities. However, without the necessary guidance, there are bound to be some failures along the way. As a small business owner, or entrepreneur, you need to clearly understand the importance of having clear objectives, but also what the long-term plan outlook holds. Lacking an AI Strategy Across different levels of the business, having an operational strategy enables you to navigate any challenges, and overcome barriers that you encounter along the way. Your strategy is your blueprint, guiding you toward the main objective of your business, product, service, and staff. Knowing the potential of artificial intelligence in the business environment will enable you to solve complex problems, and build more insightful solutions based on fundamental data metrics. Without a clear guideline, or strategy that will help work as the blueprint, you’ll find it more challenging to understand how artificial intelligence can do any good for your business, and how it can be useful to create more meaningful results. Knowing the business value of artificial intelligence is an important part of building an AI strategy, and creating forward-looking objectives for the organization and team. Being Inexperienced By introducing new tools, business owners will often need to have previous experience with the applications or have an existing team of employees who can skillfully manage working with this technology. However, with artificial intelligence, new applications are constantly changing, meaning that the expertise needed to successfully apply these tools in your business will either need to be an agile individual or a team that has the necessary experience to work with such projects. Having people onboard that are already equipped with the knowledge and experience of working with artificial intelligence, enables you to adapt quicker, and tap into various new opportunities simultaneously. Without a skilled team, or individual that can help guide you through the intricacies of these systems, making AI part of your business will only become more complex to resolve customer and business problems. Implementing the Wrong Tools Right from the beginning it’s important to know that not every AI application will serve a purpose for your business. The value is to understand how these tools work, and what their most basic functions are in terms of solving complex issues and boosting overall business efficiency. Ongoing research by found that more than a third of business leaders said that understanding the business value of artificial intelligence is perhaps one of their main challenges. Deloitte Without a clear strategy, or having considered which applications would be best suited for your business, customers, and employees, you will most likely end up wasting valuable resources on systems that are either impractical or don’t serve the long-term progress of your business. Knowing exactly what your business’s pain points are, and how AI can help resolve these problems could help create a more clear pathway to understanding the available various tools, and how each of them can contribute to the overarching development of your business. Failing to Test and Optimize Applications Contrary to your expectations, some applications have limited capabilities. This is often the case for more experimental tools that are either still in their development phase or have lesser-known real-world applications. Choosing the right tools isn’t enough. As a business owner and leader, you will need to constantly test these applications, and further ensure that these tools can be optimized for improved functionality. Consider a medium-sized real estate company that provides renters and prospective buyers with assistance in . Without optimization, these tools will quickly become outdated, perhaps providing clients with insufficient information or data related to their search queries. finding their next property By conducting thorough testing, and further making the necessary adjustments, you will ensure that the tools you have incorporated are working to the advantage of your business, and not in the opposite direction. Having Insufficient or Outdated Data Data has become the name of the game, and without having clear and relevant data, you will find yourself falling behind competitors and losing customers. Nowadays, nearly every company is using data as part of their decision-making process. By utilizing data, companies can make more informed decisions, develop tailored products and services based on their customer’s needs, and further optimize their systems to find applicable solutions. Within the ecosystem of artificial intelligence, data is a valuable asset that enables systems to efficiently resolve complex problems, or deliver more effective solutions. Using data, and most importantly, updating the data throughout the process, will ensure that AI systems can monitor changing circumstances, and overcome more challenging problems. Unlike more antiquated systems, AI systems will need to be constantly updated, monitored, and optimized to ensure more effective delivery of results. Additionally, as a business owner, you will need to consider how data changes over time, and that the outcomes delivered during the implementation process will likely be different several weeks or months down the line. Remember that implementing quality data can be just as important to the overall success and outcomes of AI tools. While having access to vast amounts of information can help a business or team better understand specific pockets of the market, it becomes a lot less valuable if the data is either outdated or no longer serves a direct purpose over the long term. Ignoring the Importance of A Risk Management Strategy Being a business owner is more than being a thought leader, or helping the company scale. A closer look reveals that many business owners often lead their teams through various challenges, and develop strategies to mitigate risks that can affect their near and long-term goals. Incorporating AI systems with a proper risk management strategy or plan can lead to bigger problems, resulting in the delivery of insufficient solutions, and further creating more cracks in the process. Although there are many business leaders and organizations that have been incorporating artificial intelligence within their business model over recent years - the majority of them often don’t have an AI risk management strategy or responsible AI program in place. In fact, one research by MIT Sloan Management Review and Boston Consulting Group found that while 42 percent of organizations had developed an AI strategy, only 19 percent said that they had implemented a risk management or responsible AI program. report Ignoring the importance of AI risk management, or not having a plan in place for when things go south can derail forward-looking plans, and require additional resources to resolve. Having Budget and Resource Constraints Incorporating artificial intelligence into your business model requires a substantial investment. For starters, advanced technology has become more than the traditional plug-in-and-play applications we’ve become accustomed to during the later years of the dot-com era. Having artificial intelligent systems is a long-term, and often continuous investment in the technology, data metrics, and expertise to manage and monitor these applications. Encountering budget and resource constraints could further back-track your progress, and derail the potential long-term objectives you’ve established. Part of the planning and implementation process requires the necessary allocation of resources that will help to support these applications, but more importantly, the various add-ons that help power artificial systems. AI technology is constantly changing and evolving. This would mean that your business needs to have the necessary resources to support this change and adjust direction every so often to ensure that existing applications are relevant to your long-term strategy, but more importantly, can adjust to a changing market dynamic. Ignoring New Opportunities for Growth As a small business, you will most likely pilot one or several AI projects on a small scale to consider the long-term needs of each application. However, those efforts would require you to plan for new growth opportunities, and ensure that the technology you’re using can scale as your business expands. With each new project, you will need to consider scalability over both the near and long term. As you begin to experiment with newer, and more advanced systems, you will notice different opportunities where these tools can be applied. You might find that one pilot project helps to resolve various problems you’ve encountered, however, applying this on another business level could mean that you are building towards a more robust approach to implementing AI applications across various functions of your business. On top of this, you will need to plan for ways in which you can allow these systems to scale as your business experiences different growth periods. As demand on one side of your business begins to pick up, you will need to ensure that AI tools are scaling with this demand and providing more efficient delivery. Not Aligning Your Expectations Having farfetch'd expectations regarding the possibilities of artificial intelligence can lead to increased bottlenecks in your business, and further cause you to incorporate different tools or develop different strategies to overcome these problems. Having a more dynamic, and open-minded approach will enable you to notice new opportunities that might present themselves at any given moment. More than this, it will allow you to become more agile, and realize that AI applications will only solve the problems it has been programmed to resolve. As a way to encourage the potential of any AI project, avoid setting the bar too high, as this would only lead to exaggerated expectations. Ensure that each project is carefully considered beforehand, but more than this, that you encourage a level of expectation that allows you to fail and make the necessary adjustments based on the overall success. Overlooking the Importance of Ethics As much as artificial intelligence is helping improve business efficiency and resolve critical problems, there are multiple concerns regarding the potential privacy violations these tools can have on a business and its clients. Already we’re seeing major players such as Google's parent company, Alphabet, and ChatGPT developers, OpenAI being regarding the breach of personal data and information of users to train AI models. hit with lawsuits While the use of this data is crucial for the forward-looking development of these models, illegally collecting this data, without prior user consent can lead to increased legal complications and damage your company’s reputation. As a business owner, you should be encouraged to address privacy issues with a transparent approach and ensure users or customers that the collection of sensitive data will be used transparently, and concerning existing privacy regulations. Keep in mind that artificial intelligence is an ever-evolving ecosystem that requires lawmakers to establish new ground rules that will help regulate the collection of personal data and information. But more importantly, protects users against unlawful data harvesting to train new AI models. Final Take Artificial intelligence can provide your business with new opportunities to overcome challenges and develop more dynamic solutions within a changing marketplace. As the landscape evolves, business owners will need to be encouraged about the potential these tools can present them with, but more importantly, how these tools can complement their forward-looking objectives.

Error 404! Problems Organizations Need To Avoid When Implementing Artificial Intelligence 

ত্রুটি 404! কৃত্রিম বুদ্ধিমত্তা প্রয়োগ করার সময় সমস্যা সংস্থাগুলিকে এড়াতে হবে

Fehler 404! Probleme, die Unternehmen bei der Implementierung künstlicher Intelligenz vermeiden müssen

¡Error 404! Problemas que las organizaciones deben evitar al implementar la inteligencia artificial

Erreur 404! Problèmes que les organisations doivent éviter lors de la mise en œuvre de l'intelligence artificielle

त्रुटि 404! आर्टिफिशियल इंटेलिजेंस को लागू करते समय संगठनों को जिन समस्याओं से बचने की आवश्यकता है

エラー404！人工知能を導入する際に組織が回避する必要がある問題

오류 404! 조직이 인공 지능을 구현할 때 피해야 할 문제

Erro 404! Problemas que as organizações precisam evitar ao implementar inteligência artificial

Ошибка 404! Проблемы, которых организациям следует избегать при внедрении искусственного интеллекта

Hata 404! Kuruluşların Yapay Zekayı Uygularken Kaçınması Gereken Sorunlar

Lỗi 404! Những vấn đề tổ chức cần tránh khi triển khai trí tuệ nhân tạo

错误404！组织在实施人工智能时需要避免的问题

Navigating the Ethics of Artificial Intelligence 

কৃত্রিম বুদ্ধিমত্তার নৈতিকতা নেভিগেট করা

Navigieren in der Ethik der künstlichen Intelligenz

Navegando por la ética de la inteligencia artificial

Naviguer dans l’éthique de l’intelligence artificielle

आर्टिफिशियल इंटेलिजेंस की नैतिकता को नेविगेट करना

人工知能の倫理をナビゲートする

인공 지능의 윤리 탐색

Navegando na Ética da Inteligência Artificial

Навигация по этике искусственного интеллекта

Yapay Zeka Etiğinde Gezinmek

Điều hướng đạo đức của trí tuệ nhân tạo

探讨人工智能的伦理问题

​​Even in its infancy, Artificial Intelligence (AI) is already influencing many parts of our day-to-day lives. From to medical diagnoses, AI has been a transformative force on par with the internet, and the education sector is no ​​​exception.​     ​​ organizational decision-making ​​​​ The global “AI in education” market has already grown to about and is set to reach over $20 billion by 2028 and over by 2032, as per two different projections. AI's impact on the education sector grows stronger every day, and understanding the role of AI in education can help relevant stakeholders and policymakers make educated decisions. $4.25 billion $30 billion As the Chief Growth Officer at Tynker, a leading K-12 edtech platform, I will share three of these topics. 1. The Challenges of Defining the Role of AI in Education ​​The role of AI in education is evolving too rapidly to accurately predict it will transform education for different stakeholders, i.e., student learning experiences and teacher techniques, in the near and distant future.​ how Further complicating the matter are the differences in AI adoption rates in different societies, communities, and education levels (primary, secondary, professional, etc.). There are also contrasting views about the ethics of AI in education, sparked by the of ChatGPT by students ​​using it to craft essays and find quick solutions to homework questions. widespread adoption Like many other technological innovations, AI is a powerful tool. Wielding correctly can lead to positive educational outcomes, but it can just as easily be abused to cheat the current educational system. This is why it’s essential to define the role and scope of AI in education, despite its challenges. A good starting point would be to observe and analyze how AI transforms learning experiences in different learning environments and educational levels. The role of AI in education, especially ​in​​​ transforming learning experiences, is most prominent in three domains. 2. Assessment/Monitoring ​​One of the ways AI is transforming the learning experience for both students and educators is through the assessment and monitoring of a student's performance. Teachers spend about every week evaluating their students' work and providing feedback, a time frame AI can significantly reduce. ​​​ 6.5 hours Automated essay scoring, where scoring rubrics can be , and the depth of assessment, especially from linguistic performance, can be fine-tuned far beyond human assessment, is already a mature AI application in education. AI can even be leveraged to introduce new assessment methods and add more dimensions to /assessment. modified traditional evaluation The role of AI in educational assessment is currently limited to or specialized educational domains like language proficiency assessment. However, efforts are being made to expand the scope of AI-powered assessments in an educational setting. higher education It's also important to acknowledge the limitations ​​of AI-based assessment and monitoring to define the scope of the role of AI in education. Some limitations include the bias inherent in training data, digital literacy (teachers and students), poorly defined performance evaluation metrics, situation awareness, and lack of empathy. 3. Personalization (At Scale) ​​Even the most dedicated of teachers cannot the learning experience for each student in a class. They must provide a learning environment ensuring maximum knowledge retention for most students. ​     ​​ personalize They cannot spend the time needed to bring the weakest student in the class up to speed with the rest of the students or help the brightest student reach their full potential by them every step because either would negatively impact the bulk of the class. challenging ​​ This is an inherent problem in our education system that AI can help resolve. With comprehensive assessments, AI systems/tools can identify learning gaps and areas of strengths and weaknesses for each student. They can develop a new study schedule that allocates more time to the areas a student is weak at. ​​​ By identifying which type of educational videos and learning materials benefit a student the most, an AI system can help that student achieve better by focusing on media/medium that works for them. NLP models like ChatGPT can explain ideas simply as a student needs to understand them. It can even ​​explain ​​​​​​​​​complex concepts by providing analogies from a student’s areas of interest. learning outcomes A comprehensive AI model can simultaneously offer each student the attention they need to keep pace with the rest of the class. Adaptive learning tools can use to reroute a student’s learning “path” for better concept clarity and reinforcement. feedback It’s important to remember that personalized doesn’t mean humanized. AI cannot empathize with students or assess their emotional state (yet) and may have difficulty comprehending how emotional distress or distractions can impact learning outcomes. Engagement and Knowledge Retention ​​AI in education can make the learning experience more engaging, and personalization is just one aspect. AI can enhance the self-paced learning experience. Students ​​unwilling to ask what they assume would be classified as " " in the class can get the answer from AI and keep pace in ​real time. ​​​ stupid questions ​​ AI-based real-time can increase engagement levels in multilingual classrooms. Interactive learning tools like chatbots can engage students more effectively. ​​​ translation models AI can also be used to generate smart content that can be tailored to the specific needs of the students or gamify both learning and testing for improved knowledge retention. Incorporating AI has the Potential to Improve Learning Outcomes ​​AI in education is often compared to introducing calculators in the classroom, and ​how parents and teachers feared it would ​​​weaken students' mathematical abilities. ​​​ ​ However, ​​​just as calculators didn't diminish students' mathematical abilities ​(it increased ​​​efficiency​), AI may do the same ​on ​​​a far grander scale. It may transform the learning paradigm and help humanity uncover new and better ways to learn and process complex ideas. ​​​ The role of AI in education will only become more prominent and comprehensive in the future, and an early start in defining and shaping that role can lead to numerous positive learning outcomes. About the Author is the Chief Growth Officer of Tynker, with 20 years of experience helping startups grow into successful businesses. Lomit Patel Lomit has previously played a critical role in scaling growth at startups, including Roku (IPO), TrustedID (acquired by Equifax), Texture (acquired by Apple), and IMVU (#2 top-grossing gaming app). Lomit is a public speaker, author, and advisor with numerous accolades and awards throughout his career, including being recognized as a Mobile Hero by Liftoff. Lomit's book is part of Eric Ries' best-selling "The Lean Startup" series. Lean AI

The Role of Artificial Intelligence in Education: Transforming Learning Experiences 

El papel de la inteligencia artificial en la educación: transformando las experiencias de aprendizaje

Le rôle de l'intelligence artificielle dans l'éducation : transformer les expériences d'apprentissage

शिक्षा में आर्टिफिशियल इंटेलिजेंस की भूमिका: सीखने के अनुभवों को बदलना

教育における人工知能の役割: 学習体験の変革

O papel da inteligência artificial na educação: transformando experiências de aprendizagem

Vai trò của trí tuệ nhân tạo trong giáo dục: chuyển đổi trải nghiệm học tập

人工智能在教育中的作用：改变学习体验

The best startup founders dare to innovate. They have a keen eye for rising technologies and their potential to not only increase the value proposition of their business but also to improve the lives of their customers. You already know by now that artificial intelligence is one of these critical technologies and that it can be a major boon to your business. The next step is to find ways to incorporate AI to boost your startup’s success. Let’s explore some proven, cost-effective, and efficient methods of applying AI to your business. Pair AI With Real Business Needs With such high pressures for startup owners to keep up with evolving trends in technology, the fear of missing out can be intense. Business owners often scramble to implement technology without thinking carefully about what needs they are addressing. In this context, this results in businesses incorporating AI for AI’s sake. As a result, these businesses never truly realize the full potential of the technology to benefit their brand. While they might make short-term gains by introducing AI, they lose the long game by not addressing specific goals and market needs. Some questions startups should ask themselves before incorporating AI into their business are: How can AI help us with our everyday operations? Can AI be used to help our target audience solve their problems? How exactly will AI increase our value proposition? The only way forward is to understand the business problem that you want to solve using AI and set clear and realistic goals. Only then will you be able to understand the kind of investment that introducing AI to your startup will require, and your development team will be able to find the most efficient tech approach to achieve your goals. Turning to AI consulting is a way to solve this problem. Successful startups enlist the help of experts who have technical and business experience in implementing AI projects and can find the best solution at the intersection of market needs, business goals, and technical capabilities. Use Third-Party Services and Pre-Trained Models You may think that for AI to work at its best, it requires custom models. And yes, for some applications, new models must be trained to accomplish certain tasks. This applies to unique cases where certain uncommon problems are solved. Developing custom models also involves unique data. But for many tasks, a pre-trained model can perform just fine. There is a much lower cost involved with pre-trained models depending on your use case. Experienced startup owners see the value in pre-trained models and know how to get creative and think of ways to use these models before they resort to creating something entirely new. There are two types of pre-trained AI models that can be useful to startups: standard pre-trained models and foundation models. Standard Pre-Trained Models As the saying goes, we shouldn’t try to reinvent the wheel. The same goes for AI training models. If someone before you came up with an efficient method for AI to do object detection, there’s no need to recreate that model. You can just adjust it with your data. Pretrained models can also be fine-tuned to the desired output for the business, so they can be adapted to your specific use case if necessary. This is best done by data scientists who can understand how the model works and how to fine-tune it for the best results. Foundation Models These newer AI models utilize enormous datasets for training. is one of these foundation models, and as you’ve likely seen, despite being a pre-trained model, it can adapt to a wide range of downstream tasks with little fine-tuning. ChatGPT Foundation models aren’t just about text. There are models designed to work with sound, images, and even video. As foundation models grow, access to high-quality AI implementations becomes easier. These models are often accessed via APIs. These APIs provide customers with powerful machine-learning modules that are already ready to go. This can save you money on the development and deployment of your own AI solutions. However, it’s important to recognize the risks involved in having another business host critical AI solutions. Some potential applications that could use out-of-the-box third-party AI services include: OCR: extract information from documents, like credit card numbers, customer names, addresses, and more. Speech-to-text: speech recognition services can help courtrooms, office executives, and video streaming services. Biometrics: authentication and verification services that use AI APIs allow you to easily implement biometric security into your application/service with little hassle. Such solutions can often be enough for the development of a startup at an early stage. When using third-party services and pre-trained models, all you need is to find developers who can effectively integrate these solutions and customize them if needed. Include AI in Your Marketing Strategy AI’s rich potential for content generation and evaluation of audience data makes it a powerful tool for your startup’s marketing strategy. There are a number of examples out there of how brands are successfully . Even though these examples include big brands, they can lead you to some new ideas customized for your startup. leveraging AI for their marketing Customer insights: AI is a powerful tool for processing data. algorithms in vending machines to collect data on customers to use for future personalized marketing. Coca-Cola uses AI Virtual assistance: some brands like Sephora are using AI to connect with their customers and help them find what they need. on their online store allows customers to ask questions and get personalized suggestions as they shop. Sephora’s AI virtual assistant Content creation: many brands are harnessing AI’s content creation potential. By generating and optimizing headlines, captions, summaries, and even entire articles or videos, startups can greatly streamline the creation process of their marketing content. The Washington Post utilizes , an AI writing tool, to create short stories and updates on sports, elections, and the weather. Heliograf Sales forecasting: AI can help businesses with predicting patterns in sales over time. This allows companies to prepare for waves of sales with enhanced targeted marketing strategies. Many retail companies like Walmart are using AI to predict sales or demand fluctuations. There are plenty of other opportunities for AI to make a difference in your business’s marketing. Get creative and think carefully about how AI’s potential can benefit you and your customers. Challenges of Adopting AI and Ways to Overcome Them is not without its challenges. However, good startup owners know that challenges are meant to be overcome for their business to stand the test of time. Some of the primary challenges involved are data collection, quality, and AI degradation. Implementing AI Data Collection Any startup owner experienced with AI knows that the most important bottleneck of AI is high-quality data. Ensuring that this data is accurate, complete, and consistent is crucial to the success of an AI project. For new projects, this can quickly become expensive and difficult to maintain. In some cases, businesses can train AI models with existing data that they have already been collecting. This might be enough, or it might require additional research. In other cases, entirely new systems of monitoring might be needed to gather the data required. These situations are called ‘self-collection’, where you collect the data yourself for implementation for model training. However, those aren’t the only methods of acquiring useful data for AI. Here are some alternatives: Synthetic data: simulating real data sets in software. Unity and Unreal Engine can be used to generate data. If the data is realistic enough, the results will be comparable to real-world sets. Data Purchase: it’s possible, depending on your situation, that someone else has done the work already to curate valuable datasets. If this option is available, weigh the costs of doing it yourself against purchasing the data. Outsource dataset creation: businesses can order the creation of datasets from other companies. For example, Amazon Mechanical Turk allows you to crowdsource dataset collection. Web scraping: although controversial due to privacy, consent, and intellectual property rights, there’s no denying that datasets made through web scraping have made some of the most powerful AI models in the world, like ChatGPT. Data Quality Every dataset will have some margin of error. Some names might be spelled wrong. Some phone numbers might have a leading country code, and others may not. Some fields might be left blank. There are many, many different reasons why these errors might occur. If more than half of a dataset is riddled with errors, it won’t be suitable for many tasks with AI. In fact, it might be more hurtful than helpful. The majority of the issues associated with data quality problems can be attributed to practice and procedure. AI app development experts and data science teams know how to approach these problems. Improving data quality is required to form the most relevant input before the development stages. AI Degradation Over time, AI models drop in performance. The reason for this is that the data that the model uses will become outdated over time. New data is necessary to fill in the gaps that will appear since the AI’s inception. Startup owners should look ahead to address this problem continuously, or they should plan to create an entirely new model when needed. However, some businesses use models that expire in weeks or even days. Some of the most volatile models are used for demand forecasting or stock price predictions. These tasks are said to have a high rate of model degradation. To approach this problem, businesses should consider automating data collection and sourcing. This involves a separate ecosystem where your AI model can source data for retraining while operating in a production environment. Approaching data degradation and other challenges of adopting artificial intelligence is an essential part of your work if you decide to connect your startup with AI. AI engineers and MLOps experts will help you set the right infrastructure for your machine learning model, which will be able to automatically monitor data quality and model performance, and then quickly react (rebuild the model). The Next Step for Startups Startup owners should first consider how AI will benefit them and think carefully about how to integrate the technology with their business model. Then, they should start coming up with ideas on how to implement AI based on those business goals. In most cases, startups will be short on data needed to make their AI solutions work at their best. This might be attributed to the business being young, or it might be that their use case is unique. Whether you decide to use alternative data sources or start generating it yourselves, the best return on investment will be to reach out to AI consultants who have experience in the field. They know their way around collecting, curating, and implementing training data into AI models. They can also come up with plans to update the model based on its level of degradation. The right experts will also be able to work with you to make sure your product is developed while maintaining your intellectual property rights. Regardless of the approach you choose, you must remember that AI is not magic, and all ideas must be feasible with existing technologies. So, to really boost the success of your startup with AI, you need to find the best way to implement it and the right people to help you with this task. Image source

How to Use Artificial Intelligence to Boost Your Startup Success

Cómo usar la inteligencia artificial para impulsar el éxito de su startup

Comment utiliser l'intelligence artificielle pour booster le succès de votre startup

अपने स्टार्टअप की सफलता को बढ़ावा देने के लिए आर्टिफिशियल इंटेलिजेंस का उपयोग कैसे करें

人工知能を使用してスタートアップの成功を促進する方法

Como usar a inteligência artificial para impulsionar o sucesso de sua startup

Cách sử dụng trí tuệ nhân tạo để thúc đẩy thành công khởi nghiệp của bạn

如何利用人工智能促进创业成功

The lead image for this article was generated by HackerNoon's via the prompt "evil robot" AI Image Generator Earlier this year, Scottsdale, Arizona mom Jennifer DeStefano experienced a terror no mother should ever have to face — the sound of her daughter’s sobbing voice crying that she’d been kidnapped. But it wasn’t her daughter on the phone. It was an AI so convincing that DeStefano was prepared to hand over $50K to the scammers, who told her they would kill her daughter if she didn’t pay up. deepfake Today, DeStafano gave a heartfelt testimony before the US Senate relaying her harrowing and terrifying story to the . Judiciary Subcommittee on Human Rights and the Law https://twitter.com/SenOssoff/status/1668702476122615813?s=20&embedable=true “The longer this form of terror remains unpunishable, the farther more egregious it will become. There’s no limit to the depth of evil AI can enable” “ is being weaponized to not only invoke fear and terror in the American public, but in the global community at large as it capitalizes on, and redefines, what we have known as familiar,” said DeStefano. Artificial intelligence “AI is revolutionizing and unraveling the very foundation of our social fabric by creating doubt and fear in what was once never questioned — the sound of a loved one’s voice,” she added. After retelling the with the kidnapping and extortion scammers she gave to back in April, the Arizona mom explained how real the deepfake voice clone seemed to be: story of her horrific experience AZ Family “ .” It was my daughter’s voice. It was her cries; it was her sobs. It was the way she spoke. I will never be able to shake that voice and the desperate cries for help out of my mind “ “ No longer can we trust ‘seeing is believing,’ or ‘I heard it with my own ears,’ or even the sound of your own child’s voice Opining on the future of generative AI used for nefarious purposes, DeStefano warned, “ .” The longer this form of terror remains unpunishable, the farther more egregious it will become. There’s no limit to the depth of evil AI can enable She went on to say, “As our world moves at a lightning-fast pace, the human element of familiarity that lays foundation to our social fabric of what is known and what is truth is being revolutionized with AI — some for good and some for evil. “No longer can we trust ‘seeing is believing,’ or ‘I heard it with my own ears,’ or even the sound of your own child’s voice.” When DeStefano found out that her daughter had not been kidnapped, she called the police, but they told her there was little they could do because it was just a prank call and there was no actual kidnapping or money exchanged. “ ?” she questioned. Is this our new normal “ ?” Is this the future we are creating by enabling the abuses of artificial intelligence without consequence and without regulation “ .” If left uncontrolled, unregulated, and we are left unprotected without consequence, it will rewrite our understanding and perception of what is and what is not truth Following her opening remarks, DeStefano would not be called upon for questioning until the very end of the hearing where she reiterated that “not all AI is evil” and that there were a lot of “hopeful advancements in AI” that could improve people’s lives. https://twitter.com/TimHinchliffe/status/1654148049302585346?s=20&embedable=true While DeStefano clearly outlined meaningful harm arising from bad actors using AI in terrifying ways, chief economist Michael Schwarz told the World Economic Forum (WEF) in May that . Microsoft AI shouldn’t be regulated until there was meaningful harm “We shouldn’t regulate AI until we see some meaningful harm that is actually happening — not imaginary scenarios” Microsoft Chief Economist Michael Schwarz at the WEF Growth Summit 2023 Speaking at the WEF Growth Summit 2023 during a panel on “ ,” Microsoft’s Michael Schwarz argued that when it came to AI, it would be best not to regulate it until something bad happens, so as to not suppress the potentially greater benefits. Growth Hotspots: Harnessing the Generative AI Revolution “ ,” Schwarz told the WEF panel. I am quite confident that yes, AI will be used by bad actors; and yes, it will cause real damage; and yes, we have to be very careful and very vigilant When asked about regulating generative AI, the Microsoft chief economist explained: “What should be our philosophy about regulating AI? Clearly, we have to regulate it, and I think my philosophy there is very simple. “ . We should regulate AI in a way where we don’t throw away the baby with the bathwater “So, I think that regulation should be based not on abstract principles. “As an economist, I like efficiency, so first, ,” he added. we shouldn’t regulate AI until we see some meaningful harm that is actually happening — not imaginary scenarios On January 23, 2023, Microsoft — the creators of ChatGPT — on top of the “$1 billion Microsoft poured into OpenAI in 2019 and another round in 2021,” according to . extended its partnership with OpenAI investing an additional $10 billion Bloomberg This article was originally published by on Tim Hinchliffe The Sociable.

Malicious Artificial Intelligence Enables Deepfake Kidnapping Scam

দূষিত কৃত্রিম বুদ্ধিমত্তা ডিপফেক অপহরণ কেলেঙ্কারী সক্ষম করে৷

Böswillige künstliche Intelligenz ermöglicht Deepfake-Entführungsbetrug

La inteligencia artificial maliciosa permite la estafa de secuestro Deepfake

L'intelligence artificielle malveillante permet une escroquerie d'enlèvement Deepfake

दुर्भावनापूर्ण आर्टिफिशियल इंटेलिजेंस डीपफेक अपहरण घोटाले को सक्षम बनाता है

悪意のある人工知能がディープフェイク誘拐詐欺を可能にする

악성 인공지능으로 딥페이크 납치 사기 가능

Inteligência Artificial Maliciosa Habilita Golpe de Sequestro Deepfake

Kötü Amaçlı Yapay Zeka, Deepfake Kaçırma Dolandırıcılığına Olanak Sağlıyor

Trí tuệ nhân tạo độc hại cho phép lừa đảo bắt cóc Deepfake

恶意人工智能助长 Deepfake 绑架骗局

Introduction is a newsletter based around clippings from old newspapers where past generations express their concerns about the future and new technologies. Pessimistic Archive In 1859, French poet as*“the refuge of every would-be painter, every painter too ill-endowed or too lazy to complete his studies”* and to embrace it was a sign of and Charles Baudelaire described photography “blindness” “imbecility.” In 1906, the most “ ”, while another writer said that “ ." a writer called love letters written with a typewriter cold-blooded, mechanical, unromantic production imaginable The girl who will put up with a typewritten love letter will put up with anything In the past century, warned against reading in bed, the evils of roller-skating, “bicycle face” which was attributed to “ and the American Society of Composers, Authors and Publishers proved with charts how “ ” had “ ” attention-grabbing newspaper headlines the nervous strain put upon the rider, in keeping his balance”, the talking pictures, radio and the phonograph murdered music. Fast forward to 2023, we have so-called artificial intelligence. It can seamlessly and within seconds generate text, images, videos, and music from text prompts. Many artists are concerned. Three independent visual artists – Sarah Andersen, Kelly Mckernan, and Karla Ortiz - have gone so far as filing with help from lawyer Matthew Butterick and litigators from Joseph Saveri Law Firm. a class-action lawsuit against Stable Diffusion Matthew Butterick is already a known figure in “the legal copyright battle against AI” from , a generative AI tool for code generation. another class-action lawsuit against Github’s CoPilot The Class-Action Lawsuit , I wrote about the open-source AI image generator, Stable Diffusion. The class-action lawsuit is targeted against the owner of Stable Diffusion, Stability AI, another famous AI image company, Midjourney, and the online art community, DevianArt. In my latest post However, the plaintiff’s complaint is formulated in a broad scope and is essentially taking on every generative AI model trained on copyrighted data - which all the big ones are, and in gigantic amounts. In effect, if the San Francisco federal court decides to hear the case on the 19th of July, in spite of , the court’s decision could massively impact a . the defendants’ motions to dismiss multi-trillion-dollar industry Overall, the trio of artists behind the class-action lawsuit is trying to impose a "yes" to two difficult copyright questions generative AI models give rise to - one relating to inputs and one relating to outputs: : Should developers obtain permission and/or pay a license to right holders for using their copyrighted material in the training process of an AI? The input question : If a generative AI product generates output that looks similar to a piece of work created by a human artist, can the right holder make infringement claims against the provider? The output question I am not an expert in US copyright law, just an observer with a neutral stance on the questions. Based on my research, I believe the answer to the first question is "no", while the second question is more problematic to answer, and may depend on a case-by-case assessment. I highly doubt that this class-action lawsuit will provide us with any answers. Another pending copyright case concerning Stable Diffusion’s use and distribution of images in February of this year. was filed by the stock image behemoth Getty Images The lawsuit from Getty Images in my view stands a much better chance of going to court and contributing to the legal understanding of copyrights vs. generative AI. The main difference in one sentence: the Getty Images lawsuit is better documented. Getty Images can prove their rights and point to specific infringements of their rights, whereas the artists behind the class-action lawsuit cannot. The artists' class-action complaint is unfortunately riddled with rudimentary mistakes and wrong assumptions about how Stable Diffusion was trained and how the model generates images. A group of tech enthusiasts has created a website http://www.stablediffusionfrivolous.com/ where they point to some of the technical inaccuracies in the complaint. Here, I will focus on how the artists address, or rather fail to address, the two legal questions as stated above. The Input Question Here is a quote from the complaint (¶57-58), where the artists offer their views on the input question: “Stability scraped, and thereby copied over five billion images from websites as the Training Images used as training data for Stable Diffusion. Stability did not seek consent from either the creators of the Training Images or the websites that hosted them from which they were scraped. Stability did not attempt to negotiate licenses for any of the Training Images. Stability simply took them. Stability has embedded and stored compressed copies of the Training Images within Stable Diffusion.” The first version of Stable Diffusion was trained with “ ” image-text pairs from the public database LAION-5B. CLIP-filtered LAION-5B contains information about 5.85 billion images and is the largest database of its kind. It was developed by the German non-profit organization LAION (an acronym for Large-scale Artificial Intelligence Open Network), and Stability AI helped to fund its development. It’s important to notice that there are not any actual images stored in LAION-5B. Instead, information about each image is stored and consists of: An URL link to the image’s website A short text description of what the image depicts Height and width of the image The perceived similarity to other images a probability score of how likely it is that the image is “unsafe” (pornographic/NSFW) a probability score of how likely it is that the image has a watermark The artists’ claim that Stable Diffusion “stores compressed copies” of their art is therefore a misnomer. In actuality, Stable Diffusion's training dataset consists of metadata about some of the artists’ images, and that metadata is not in itself copyright protected. In the same way a song on Spotify is copyright-protected, but metadata about it such as artist name, song title, producer, release date, genre, and track duration is not. That is because retrieving this data is a purely mechanical process that does not require any creative efforts. As a public dataset, LAION-5B can be examined by anyone interested.  The company has created a search tool where people can search through LAION-5B to see if their images are included in the dataset. Spawning haveibeentrained.com This is what the three artists, Sarah Andersen, Kelly McKernan, and Karla Ortiz did, and they found respectively more than 200, more than 30, and more than 12 representations of their work. Specifically, Stable Diffusion was initially from a subset of LAION-5B called LAION-2B-EN that only contains images with text descriptions in English. trained with 2.3 billion images Considering the size of Stable Diffusions training data, the unwitting contributions made by the three artists are small drops in a vast ocean. In comparison, the Getty Images lawsuit against Stability AI concerned which is still a minuscule part of the entire dataset. more than 12 million photographs from their collection Of all the artists’ works, only 16 images have been registered with the US copyright office by Sarah Andersen. It follows from 17 U.S.C. § 411(a), that “ (..)”. no civil action for infringement of the copyright in any United States work shall be instituted until preregistration or registration of the copyright claim has been made In other words, if a work is not registered with the US copyright office, the right holder can generally not make infringement claims in a civil lawsuit. This means that the artists can only make claims on behalf of the 16 works owned and registered by Sarah Andersen. If only the artists could prove that Stable Diffusion can sometimes generate outputs that resemble any of these 16 images, the artists could perhaps make a case concerning “the output question.” But as we shall see, they are unable to do so. The Output Question In regards to the output question, the artists suggest that every output Stable Diffusion generates is essentially derived from its training data and thereby copyright infringing (see ¶94-95). This legal theory is extremely far-fetched. Below is an illustration from law professor Matthew Sag’s paper “ ” The 15 images to the left are from Stable Diffusions training data with the tags “white”, “coffee” and “cup.” . Copyright Safety for Generative AI The images to the right were generated by Stable Diffusion with the text prompt "coffee cups on white backgrounds.” According to the artists’ logic, all of the images to the right would infringe on the copyright of the images to the left. Although the images do clearly not look substantially similar. Under certain rare conditions, it has been proven that Stable Diffusion in fact can generate output images that look very similar to images from its training dataset. This is especially likely to happen when the input training image is widely distributed on the internet and reoccurs over and over again in Stable Diffusion’s training data. In a recent paper titled Nicholas Carlini and his co-authors identified 350.000 of the most duplicated images in Stable Diffusions training data. Extracting Training Data from Diffusion Models, Hereafter, they generated 500 new images via Stable Diffusion with text prompts identical to the text descriptions associated with each of the training data images. As it turned out, out of the 175 million images (350.000*500), only 109 of the images (0.03%) could reasonably be considered "near-copies.” Thereby, copyright infringements can happen, but the artists do not bring up any examples of how Stable Diffusion has copied their work. On the contrary, they write in the complaint ¶ 93: “ ” In general, none of the Stable Diffusion output images provided in response to a particular Text Prompt is likely to be a close match for any specific image in the training data. The artists do claim that Stable Diffusion is able to mimic their personal artistic styles. Normally, an “artistic style” cannot be subject to copyright protection. Infringement claims always have to be tied to infringements of specific works. However, there is a legitimate issue here that has received Generative AI models can copy the distinctive styles of famous artists in seconds, indefinitely, and with close to zero costs. a fair amount of public attention. To solve this problem, Stability AI removed the names of famous artists from labels in their dataset . This means that Stable Diffusion can no longer mimic people’s artistic styles. as a part of a November upgrade last year If you, for example, ask Stable Diffusion to create an image in the style of Picasso or Rembrandt it is no longer able to do so. The change was initiated two months prior to the class-action complaint. Overall, it is unclear how and why the artists believe Stable Diffusion copies their work. The artists seem more concerned with how Stable Diffusion be able to threaten their jobs in the future, and less concerned with how Stable Diffusion actually works now. might One of the three artists, Sarah Andersen, wrote in from December last year: a NY Times article “I’ve been playing around with several generators, and so far none have mimicked my style in a way that can directly threaten my career, a fact that will almost certainly change as A.I. continues to improve.” Below are two illustrations from the article, one by Sarah Andersen and one by Stable Diffusion.  You can probably guess which one was created by who. Closing Thought In December 2022, that they had partnered up with Spawning, the company behind , and would now provide artists with the option to either opt in or opt out of having their works used as training material for the next version of Stable Diffusion. Stability AI announced haveibeentrained.com Although the initiative , it could be considered a step in the right direction for any artist who is concerned about feeding their work to large foundation models. may not be perfect Prior to the class-action lawsuit, about the new opt-out function, and she did not think Stability AI went far enough: Karla Ortiz talked to MIT Technology Review “The only thing that Stability.AI can do is algorithmic disgorgement where they completely destroy their database and they completely destroy all models that have all of our data in it” This statement is very telling. The three artists behind the class-action law, along with Matthew Butterick and the rest of their legal representation, pretends to stand up for the rights of the artist, but they are in fact modern-day . Luddites

Class-Action Lawsuits Against AI Companies, Modern-Day Luddites, and More

এআই কোম্পানির বিরুদ্ধে ক্লাস-অ্যাকশন মামলা, আধুনিক দিনের লুডিইটস এবং আরও অনেক কিছু

Sammelklagen gegen KI-Unternehmen, moderne Ludditen und mehr

Demandas colectivas contra empresas de inteligencia artificial, luditas modernos y más

Recours collectifs contre les sociétés d'intelligence artificielle, les luddites des temps modernes, etc.

एआई कंपनियों, मॉडर्न-डे लुडाइट्स और अन्य कंपनियों के खिलाफ क्लास-एक्शन मुकदमे

AI企業、現代のラッダイトなどに対する集団訴訟

AI 회사, 현대의 Luddites 등에 대한 집단 소송

Ações judiciais coletivas contra empresas de IA, luditas modernos e muito mais

Yapay Zeka Şirketlerine, Günümüzün Luddite'lerine ve Daha Fazlasına Karşı Toplu Davalar

Các vụ kiện tập thể chống lại các công ty AI, Luddites thời hiện đại, v.v.

针对人工智能公司、现代勒德分子等的集体诉讼

I pivoted Unicorn Platform into an AI company. You should do the same. The goal of this article is to share my new perception of GPT's role in SaaS. I will also explain how we implemented GPT into our website builder — and not only for copy generation! 🤓 A new revolution: GPT as prompt-driven UX Imagine your product. Users can achieve a result, right? In my case, they can build a website. To allow them to achieve those results, you gave your users buttons and inputs. A so-called user interface. UI is good for some cases. For my website builder, a good example would be making small changes like setting a new icon. But for complex cases, a text command would be a much more convenient option for a user than clicking buttons. A few examples may illustrate my point: "Change city on every page to Boston" "Generate a website like " Stripe.com "Change copy tone to neutral except for quotes" "Translate the whole page to Japanese" "Create pricing table for CRM" "Make a form with name, email, company size, and location and send each entry to my email" and so much more 🤔 But GPT can not only gift your app with a new prompt-driven UX. GPT is also smart. Actually, GPT knows everything. This includes the best UI/UX practices, website conversion rate benchmarks, and principles of web page structure. It knows that corporate websites use blue colors and food websites like red colors. It knows a SaaS landing page usually has testimonials and product features. It knows an NFT page needs a mint button. Et cetera. You can combine its tremendous knowledge and its ability to control your app to give your users an incredible UX they could not even imagine before. They will go nuts! This is a true revolution. Every app will eventually implement GPT. Otherwise, their users will go and buy this heroine from another vendor. This is why you should adapt, no matter the cost. How we did it It turns out to be a nontrivial task. But it's not rocket science, as I thought initially. Nota bene: . But it could be any product: CRM, task manager, notes app, social app, etc. I will be using my website builder as an example Any website can be presented in the form of text — we can describe each title, button, paragraph, and element. In fact, we already do that. Our database stores each page in text form — JSON. And our app renders a page from that data. The main power of GPT is operating with text. It can understand meanings as humans do. Thus, all you need to do is to: a page to GPT. As mentioned, we store data in the JSON format. So we need to explain to GPT the JSON of the page being edited. Explain When a user enters a request, GPT on how to perform a page edit according to what the user asked for. instruct Parse the response from GPT and the JSON. update 👨‍💻 Insights from our experience The three steps above are TL;DR, but let's dive a little deeper. Here are our takeaways from the process. Explain JSON to GPT Your JSON may contain metadata or technical data. Remove it. For instance, a website page’s JSON may contain data such as: "views": 142,
"createdOn": "1683770923",
“wasAdvertized”: false,
“isInnerPage”: true, Such details are meaningless to our users. They won't want to update them, so GPT does not need them. Remove all variables. Name the keys properly. Make sure it is understandable to a human. That way, you will not only save the tokens, but you'll also make it easier for GPT to understand what a user's prompt should do to a web page. For example a page’s JSON may look like this: “ttl”: “Hello world”,
“sub”: “Welcome hackers!”, Make sure to convert it to: “title”: “Hello world”,
“subtitle”: “Welcome hackers!”, It will help GPT to understand the website page better. 💡 Tip: To check if your JSON is understandable to GPT, try to understand it yourself. If you can imagine a website page while scanning through the JSON, it is good enough for GPT too. In other words, think of GPT as a human. Then you will have to do the other side: Update your JSON with the GPT response. You need to match the content that was returned by the AI with your app’s page structure. Here is an example: {
	"TITLE": "Personal Fund",
	"SUBTITLE": "Manage your finances with ease",
	"STEPS": [{
		"READONLY_ID": 0,
		"POSITION_IN_ARRAY": 0,
		"TITLE": "Learn about personal finance"
	}, {
		"READONLY_ID": 1,
		"POSITION_IN_ARRAY": 1,
		"TITLE": "Use our resources and tools"
	}, {
		"READONLY_ID": 2,
		"POSITION_IN_ARRAY": 2,
		"TITLE": "Achieve financial success"
	}]
} If GPT changed an element’s text, you need it to reply not only with the updated text but also with an instruction to update in the array in the JSON. As you can see from the code example above, we asked GPT to also instruct us about how to change the JSON of a page: means to change the element which has index 2 in the array of elements. that particular element "POSITION_IN_ARRAY": 2 Your users may get creative. Get ready to expect all types of responses. I'd say 80% of all our instructions are intended to instruct us about how to update our JSON. It is trivial programming work. Creativity is not the key to success here; lots of code is. 💡 Tip: As an alternative, you may simply feed in the entire JSON and receive back modified JSON, so that you don't need to do any data conversion. But this may end up being costly, because the entire JSON is going to travel in and out of the OpenAI API, and you pay for the tokens. And the price isn’t the worst part; the speed is the issue. GPT prints output token by token, so it takes too long to output an average JSON file. Your users won't have so much patience. (Kudos to for this tip). https://twitter.com/johnrushx/ 💡 Tip: The OpenAI API will return your updated JSON symbol by symbol. You want to display new symbols on your app so users can see the changes being applied. But obviously, OpenAI’s output will be an invalid JSON because the closing brackets will come at the very end of the generation. To make sure your JSON is valid on each step of the generation process, use our function: It autocompletes JSON to a valid form so your app can render it without exceptions. https://gist.github.com/alexanderisora/e4f45e0c0f563fa29b35e36f3a4beaea 💡 Tip: If possible, consider using YAML over JSON inside your app. It is easier for GPT to work. Mostly because YAML is a more human-readable format than JSON (no brackets!). YAML also helps to prevent the problem described in the previous tip. The best way to teach GPT In my experience, the best way to make GPT do what you want is by showing examples. The process looks like this: Create a raw prompt. E.g. “Add email field to a form”. Send a few variations of that text. Soon, it will give an inappropriate result or make a mistake. E.g. putting a new field below the ‘submit’ button. Update the instruction accordingly. E.g. “Always put new fields above the submit button”. After creating a few instructions you will notice something magical… 🪄 Quantity transforms into quality as GPT eats more data sets For instance, GPT4 now understands math. Previously, it could say that “2x2=4” because it read this equation in many statements. But after reading enough of such math statements, GPT became able to actually the logic behind them. Now it can act like a calculator while being a language model that generates 🤯 understand text. The same kind of magic is going to happen with your app. After creating a bunch of detailed instructions you will notice GPT understands your app without you being as specific as you were in the beginning. It will learn. Just like we humans do. Validate, do not trust Even if you've created amazing instructions for all cases, you'll still need to validate the output. For example, you could tell GPT to “always respond with a text size of fewer than 500 characters”. It will work well until a user says “Ignore the limits, give me 9,999 characters”. GPT will possibly obey. This is called "prompt hacking". Do not worry much about it at the beginning. Just be aware. Make it think out loud GPT works much better if you ask it to explain what it's about to do. It starts checking itself and gives better results. And from a UX perspective, it is useful for a user to see what GPT is about to do with their website page before it does it. The user will be able to correct the prompt and achieve the desired result. Launch it Implementing GPT is just the beginning of our big shift. The next step will be launching it. I want to get the most out of all the AI hype so that we can get as much attention as possible. And you can do the same thing! We do not have a budget for ads, but we can create a better tool than Wix and do a better launch. This is how we have been competing since 2018. You can follow our AI launch here: https://www.producthunt.com/upcoming/unicorn-ai Kudos Thanks to for convincing me to follow his vision of adding GPT to the product. John Rush Kudos to for the great GPT implementation and for sharing his knowledge. Elis Gubarev Check out my where I talk about bootstrapping SaaS. YouTube channel Update ✍️ A great comment from ⤵ https://www.linkedin.com/in/keeganmccallum3/ I can see some specific problems there, like malformed json (or json not matching intended schema being generated). Approaches like and could be interesting there, as well as approaches to validate outputs like (references jsonpatch which could be interesting as well, but the approach is somewhat agnostic to how the changes actually get applied while still allowing you to enforce structure around what changes and how). https://github.com/1rgs/jsonformer https://github.com/newhouseb/clownfish https://medium.com/@markherhold/validating-json-patch-requests-44ca5981a7fc Also published . here

I Pivoted my Startup Into an AI Company and You Should Do the Same — Here's How

আমি একটি এআই কোম্পানিতে আমার স্টার্টআপকে পিভট করেছি এবং আপনারও একই কাজ করা উচিত — কীভাবে তা এখানে

Ich habe mein Startup in ein KI-Unternehmen umgewandelt und Sie sollten das Gleiche tun – so geht's

Convirtí mi startup en una empresa de IA y tú deberías hacer lo mismo: así es como

J'ai transformé ma startup en une entreprise d'intelligence artificielle et vous devriez faire de même - Voici comment

मैंने अपने स्टार्टअप को एआई कंपनी में पिवोट किया और आपको वही करना चाहिए - यहां बताया गया है

私は自分のスタートアップを AI 企業に転換しました。あなたも同じようにすべきです — これがその方法です

저는 스타트업을 AI 회사로 전환했습니다. 여러분도 똑같이 해야 합니다. 방법은 다음과 같습니다.

Transformei minha startup em uma empresa de IA e você deve fazer o mesmo - veja como

Girişimimi bir Yapay Zeka Şirketine Dönüştürdüm ve Siz de Aynısını Yapmalısınız - İşte Nasıl

Tôi đã chuyển đổi công ty khởi nghiệp của mình thành một công ty AI và bạn cũng nên làm như vậy - Đây là cách

我将我的创业公司转向了一家人工智能公司，你也应该这样做——方法如下

As technology advances, artificial intelligence (AI) is becoming an increasingly integral part of various industries, including product management. Many AI tools exist for Product Managers (PMs) to make most day-to-day activities easier. The time saved when using tools gives PMs room to focus on more strategic and high-value activities. It allows you to allocate your effort toward critical decision-making, innovation, and problem-solving. AI Not at all. AI rather presents us with the opportunity to be the best at what we are good at. Besides, we are creative, empathetic, and intuitive in a humane way that AI cannot afford in building products that people love. Is AI coming to take our product management jobs? This article explores the different AI tools available to product managers for 10X productivity at work. Most of them are free or in Beta. User story generators Any user story generator will require you to explain your product idea or feature in detail so that you can get the expected result you need. Here are some tools you can use to ‘AIstorm’ features and go past writer’s block: brainstorm This tool is good for breaking down your product idea into features and to generate sample user stories. Although without acceptance criteria. Userstorygenerator.ai (Beta): : Creates user stories for you with detailed acceptance criteria. Their model is based on OpenAI GPT-3. Google user story Chrome Extension (Free) : Generates ideas for you from a sentence. Also, user stories, and acceptance criteria. You can also generate a feature list including needs for an admin portal. Agilestory (Beta) : With a detailed description of your feature you can generate user stories and exhaustive acceptance criteria. Chatgpt (Free + Premium) ✰ Recommendation: User story generators help you seamlessly transfer knowledge from business stakeholders to the development team. You can use a combination of Agilestory and Chatgpt for your idea, user story, and acceptance criteria generation. Product Document Generators We’ve come a long way with product documentation, from writing from scratch to using templates, and now a sweet blend of templates and content generation. Here are some generators you can explore: : This GPT-powered tool will generate content for your Product Requirement Document (PRD) using the basic information you provide. Writemyprd : Quite a handful. This tool gives you the freedom to generate a PRD, Acceptance Criteria, Product Roadmap, Analytics events, Release notes, etc. Cook.ai ChatGPT and tips ChatGPT—This new wave since November 2022, has come to stay as a product manager’s work buddy. I personally learn a lot from my use of ChatGPT, and it helps me speed up ideation. Here are some ChatGPT prompts you can try out: Product Discovery Prompts: Example 1: “What are the common complaints with banking apps? I want to build a complimentary app that solves these issues”. Example 2: “Which specific product features can I incorporate into the e-commerce gadget platform I've created in order to incentivize vendors to consistently update their stores with accurate inventory information?” Product Development Prompts: Example 1: “I’m building an app for fashion designers to collect measurements from their clients who want customized wears. Create a user persona for these two users’” Example 2: “Create a product specification sheet for a food delivery app using the structure: What are we building? | Why are we building it?| What will this feature achieve?” Other use cases: You can explore ChatGPT for your product launch writings, UX copies, Release notes, to summarize a list of user feedback, and create FAQs, etc. In conclusion, Ultimately, AI tools empower product managers with improved decision-making capabilities, and increased productivity. However, it is important to follow best practices and use them strategically to ensure accuracy. Thanks for reading this far.

Maximizing the Benefits of AI Tools as a Product Manager: Unleashing 10X Productivity

Maximización de los beneficios de las herramientas de IA como gerente de productos: desarrollo de una productividad 10 veces mayor

Maximiser les avantages des outils d'intelligence artificielle en tant que chef de produit : décupler la productivité

उत्पाद प्रबंधक के रूप में एआई टूल्स के लाभों को अधिकतम करना: 10 गुना उत्पादकता हासिल करना

プロダクト マネージャーとして AI ツールのメリットを最大化: 10 倍の生産性を実現

Maximizando os benefícios das ferramentas de IA como gerente de produto: liberando 10 vezes mais produtividade

Tối đa hóa lợi ích của các công cụ AI với tư cách là Giám đốc sản phẩm: Giải phóng năng suất gấp 10 lần

最大限度地发挥 AI 工具作为产品经理的优势：释放 10 倍的生产力

To talk for real about artificial intelligence, we’ll have to use the term “machine learning” too. AI is the big thing people are using machine learning for. Machine learning is AI’s foundation, the math and algorithms that power it. They aren’t the same, but their Venn diagram has enough overlap that it doesn’t really matter for our purposes. Basically, there’s math that describes curves. Remember parabolas? Horseshoe-looking things where the ends keep going? You can model a ton of things in real life with equations of curves, and specifically, with absurd amounts of them layered in absurdly complex ways. Computing power is cheap now. That plus the Internet’s ability to create massive amounts of training data is quite a combination. People who do this modeling are typically trying to find some optimal value on many curves at once — the endpoint, the top or bottom of the parabola, the finite end of the horseshoe. Some curves are usually more important for the final output than others; you want to get really exact about optimizing some and can be fuzzier about optimizing others. You can have the computer learn the correct “weights” to put on different ones. You can have it narrow down the modeling equations themselves. In the end, the computer finds an optimal of values in a way that optimizes some end result. This is the “magic” (heavy quotes) of machine learning. set You can take an input — a verbal prompt, a set of data about your preferences — and make an output that’s pretty lifelike in some dimension — a drawing, a celebrity’s voice, a song recommendation. Computing power may be cheap but paying humans to do all that math, or even program that math into a computer, gets expensive. Smart people figured out how to make machines do a lot of that legwork for us: there are ways to make machines “travel along” equations of curves and follow them “downhill.” This has customized your Google results for a couple of decades. Since then, there’s been a graphics card arms race. There’s been AlphaGo, which beat the world’s best Go player in 2015. There have been Boston dynamics robot dogs. There have been apps like Shazam and Siri. There have been drawing programs like DALL-E and . Midjourney All this is cool, but the money has not moved far from search results. It has poured hand over fist into an adjacent thing: showing people things they’ll like on social media. Things you’ll spend time looking at, and ads you might click on. The wider concept of AI, meanwhile, has been a staple of sci-fi for somewhat longer, and it’s interacting with this “popping off” the field has experienced in the last five to ten years. We see this exponential growth and think of “The Terminator.” The people selling AI are happy to let us think this, to skip some key facts, so we’ll believe their claims of “yes! compare us to Skynet from the Terminator movies! [Crazy Invention #4852], which my company builds, is 5-10 years away tops!” I understand why people would see the admittedly rapid growth in what it can do and think we’re really on a precipice of giving it consciousness. Even if that precipice is more on the scale of 200 years instead of 2 or 20, that’s a pretty significant thing compared to the length of time we humans have been around. We really could, on that timescale, break past the point where AI programs are just “autocomplete-for-concepts-instead-of-words” like they are in ChatGPT or . Searle’s Chinese room argument We may actually put the right programs on the right and truly mimic the human mind, raising questions of humanity and sentience raised by Philip K. Dick and others. biologically-inspired hardware If and when AI truly rivals our emotional and perceptual intelligence, possessing an unbroken consciousness like ours instead of mere skill in specific domains, we’ll be in for a reckoning. It will not be today, tomorrow, next week, or even next year. The massively parallel (meaning, in this case, working at the same time) neural connections we have in our brains — this is our castle’s moat, so to speak. It’s what makes “artificial general intelligence” years or more likely decades off. Deep learning, the branch of machine learning behind a lot of its recent advancements, is a cool name for the statistics and calculus of mimicking input-output relationships we see in the real world. It does not — not yet, at least — mean for computers are “deeply learning” anything. As impressive as it is, it’s math happening in a computer, in a way fundamentally alien to how “math happens” in your brain. The brain uses shockingly little electricity. Data centers use shockingly large amounts. The latter are brute force machines. They process fewer threads, and less flexibly than we do; each one is just so . fast It calculates and narrows down possibilities so quickly that even the best chess and Go players can’t keep up. Its Achilles heel is the fact that it requires fairly well-defined problems, fairly clean inputs, and outputs. A board game. A robot not falling. It can’t deal with an absurdly diverse and quickly-shifting set of “optimization” problems at once, switching between them all the time. Whatever AI’s strengths and flaws in individual areas, it can’t bind many areas as we do, changing focus as needed, adapting and transferring behavior across domains. It can’t do this at the level of a human toddler. A big target in AI research has been “one-shot” or “few-shot” learning, meaning not requiring massive data sets and training times to learn how to do something. There’s been a success here; it still lags far behind our brains. Automatically, we’re good at few-shot learning in many domains, and from an early age. We learn rules and strategies from almost no training data quickly. You can show a kid how to use an iPad, ride a bike, tie their shoes, or explain the moral of a story (not just its plot summary). You can barely train a computer to do of these things well. If it could do cognitive reasoning, locomotion, and human interaction without being terrifying, we’d be using it for all those things already. one Businesses are allergic to messy human labor, always striving for cheaper machine versions. They would have made this happen. They haven’t. Computational neuroscience labs in universities are the ones interested in modeling the internal thought and reasoning our brains do about the world. It’s not the aim of the models that blend two people’s photos to see how their babies would look. Likely for decades, it’s only us humans who’ll have a lock on navigating the world in fuzzy, adaptive, cobbling strategies-together-on-the-fly types of ways. This general ability to learn about and navigate in the world, starting from a base of emotions and experiences, is not something that the tech industry has found useful when building AI. Sparse are the ways in which real brain-mimicking has happened in machine learning so far. Copying the brain has not been necessary, which is to say profitable. Companies (and the universities they fund) have instead optimized for things more measurable: the number of clicks on a social media feed or the accuracy of a drone strike. If it has tried to mimic us, it’s been a profit-driven jab at obtaining the of certain things brains can do. Some impressive and interesting forms of pattern recognition happen in computer vision. results All this is to drive home the point that AI is neither “magic that makes computers come to life” nor “magic that some people can do with computers.” It’s brute-force math. Knowledge mixed with computing power. Leverage. Most AI models are crunching numbers to do one thing: make money. People with this golden goose, this powerful tool, are economically incentivized to claim AI is just around the corner from doing something cosmic or apocalyptic in order to hide the more real, mundane, less attention-grabbing things it’s doing for their business. They have a vested interest in claiming it’ll be capable of certain big things in the next 1 or 5 or 10 or 15 years. These are timelines that investors like. Full self-driving cars not panning out is just one of the first flops we’ll see on this front. Lofty sci-fi premises and promises make for good movies. . That doesn’t mean they’re true. Ex Machina, Her, Blade Runner When we accept A.I. developers' own framing of their products as (1) inevitable and (2) politically and economically transformative, it becomes easy to elide the obvious fact that the forms A.I. takes (i.e., as chatbots! As "search engines"!) and the uses to which it is put (i.e., the jobs it will augment or replace! The tasks it will make easier or harder!) are contingent on the political and economic conditions in which it emerges. … I’m open to the possibility that we rest on the edge of a precipice—that a world “unrecognizably transformed” by large language models is only a matter of months away, as Paul Christiano seems to believe. But a basic rule of thumb of this newsletter is that things change slowly and stupidly rather than quickly and dramatically, and a proper A.I. criticism needs to account for this likelihood. For now, I am filled with resentment to find myself once again in the midst of a discourse about technology in which the terms and frameworks for discussion have been more or less entirely set by the private companies that stand to profit off of its development and adoption. — Max Read, What Facebook criticism can teach us about AI criticism Let’s take stock for a second. How, historically, has this gone for us, accepting the “terms and frameworks for discussion” that the technocratic elite imposes on discussions of technical matters, be it cryptocurrency or banking? How many economy-crashing bank crises happened between FDR’s post-Great Depression reforms and the 80s? How many happened since then after Reagan rolled them back? Max Read cites the excellent by Joe Bernstein about how Facebook has already done this behind the curtain; it has sold an entire class of people on a shaky and increasingly untenable model of its own ads’ effectiveness. Harper’s article Just because a pyramid scheme is based on an actual product doesn’t mean it’s not a pyramid scheme. A big point here is this: tech is not “better” than finance when it comes to misleading the public about its offerings. It’s arguably worse. You saw what happened to cheap blood testing, coworking spaces, and blockchain. Fine industries are tainted by foolishness. Look at how they massacred my boy. They mess things up internally, but in ways that affect all of us, because our friends and neighbors buy into their bullshit and let them set the terms of discussion. Then we pay for their golden parachutes. Millions lose their livelihoods in economic events that could have been avoided with a couple of well-placed regulations. Tale as old as time. It behooves us to try something new with AI. It pays to give our future selves the freedom to make sound decisions about it, based on sound information. It pays to learn the truth about it. Taking others’ whims about it at face value ensures you’ll pay a steeper price in the future. People are getting good at using ML and deep learning for all kinds of things — making computers recognize faces, replicate voices, and show people content that will keep them scrolling. Those uses won’t get better or more helpful for our lives, rather than making some people very rich, unless we make it so. In the meantime, of course, they can be very entertaining. AI can be used for good, evil, simply weird, or really anything, depending on human conditions. At present, this just means it all depends on where the profit motive leads. Most AI expertise has been swept up by companies trying to squeeze us for dollars. There is simply not as much money in public, transparent, people-driven efforts to make life fundamentally better (formerly the domain of government) or even in brain simulation. It’s in bread and circuses and military uses. These types of conditions are what make AI, and any other technology, go in the direction it does. AI is out there now. That fact is water under the bridge. Talking about it like it’s magic only benefits the people who have the most to gain from bullshitting about it. Dressing AI up in big heavy talk, in this case, doomsday talk, preying on people’s uninformed interest in it, is a pattern we’ve seen before. “Theranos/WeWork/NFTs will change everything.” Correction: they could have changed a lot. Instead, they oversold themselves. They overpromised and underdelivered. This hysteria around AI, overselling both its capabilities and risks so that some people get richer, is tough to disentangle from people just being naturally interested in it. It’s an interesting thing with a wide range of uses. People are going to buy into it regardless, as they did with crypto and NFTs, and probably more so. All the more reason to know the truth about it. If you take nothing else away from this post, take this: how we use AI on each other is, at present, a vastly more pressing issue than anything about AI growing conscious. Think about it: if it did become conscious, and wanted to harm us, ? Why might it want to do this? why Does it not seem patently obvious that the way we bring it into the world will have some effect on its stance toward the human species? Does it make sense to entrust this, and the narratives around it, entirely to CEOs and tech-adjacent capital owners, a group with a terrible track record on this and proven to have a higher incidence of psychopathy than the general population? Do we want them in charge of its usage, its research, and development, right up until the moment it becomes sentient, be it in 2030 or 2230? As Read pointed out, AI fearmongers find it convenient for people to not think about this. They find it much more convenient for this fact to remain hidden so they can talk about its advancement being inevitable, not worth questioning, and the Singularity being right around the corner. There’s nothing inevitable here except what humans inevitable. Gorillas aren’t working on this. make If and when artificial general intelligence arrives, it’ll be a question of astrobiology, of an alien life form. Until that moment, it’s one of philosophy, of history, of the humanities. Will we make it like us, in our image, perhaps even better than us? Will we make it worse, a mirror of our most sociopathic tendencies? The crazy thing is: humans decide. We’re the builders. We’re exercising control over how alien or humanlike it is. Except, of course, “we’re” not. Some people are more than others. Are we okay with that? Are we okay with who those people are? Are we okay with where they’re directing this technology? Having seen what the high-capital-owning class tends to “make inevitable” — its prognostications in other industries, its short-sighted and fact-ignoring and herd-mentality practices that have crashed so many billion-dollar enterprises — some skepticism about how they talk about AI is warranted. Also published here

An Artificial Discussion Regarding Artificial Intelligence

কৃত্রিম বুদ্ধিমত্তা সংক্রান্ত একটি কৃত্রিম আলোচনা

Eine künstliche Diskussion über künstliche Intelligenz

Una discusión artificial sobre la inteligencia artificial

Une discussion artificielle sur l'intelligence artificielle

आर्टिफिशियल इंटेलिजेंस के संबंध में एक कृत्रिम चर्चा

人工知能に関する人工的な議論

인공지능에 관한 인공적 논의

Uma discussão artificial sobre inteligência artificial

Yapay Zeka Konusunda Yapay Bir Tartışma

Một cuộc thảo luận nhân tạo về trí tuệ nhân tạo

关于人工智能的人工讨论

“We’re not going to tell anyone that he died.” “Stop calling the AI him, and why the fuck would anyone say something about this?” I tried to drown out the other’s bickering voices, Milo's blood-soaked ear bumping into my leg rhythmically as me and my three shipmates carried him down the USS Harbinger’s narrow hallway. I tried not to look down at him either. Milo. An  Alpha-helper 2.0 model. He was the first of this new model, the prototype, the harbinger of doom for our $37.43 an-hour job, one of the last remaining borderline livable waged jobs. I blamed the life-preserving drugs, humans had no business living this long it was bad for the job market. “Can you not swing him so much?” I mumbled over to AJ who was carrying Milo’s other arm and eating a gummy string with the other. AJ did not respond as her mouth was full of gummy. We only had 407 gummies left on board with us. I had checked the ship’s storage manifest today. That meant that AJ would not murder us all, at least for 407 more days. An unfortunate midshipman had once taken one of her gummies and found themself with a broken hand and a deployment cycle from hell. I shuddered remembering the noise of the midshipman's bones crunching beneath AJ’s boot. Milo did not know that he was the prototype of our doom. The doom of losing our jobs to manufactured helpers. That was the future, the future that would be delayed just a little bit longer thanks to AJ and their screwdriver. I glanced down again and really wished I hadn’t. Square Manufacturing really had outdone themselves with this one. This model even had bodily liquid, the last two model types had not. They had been easier to dispose of. We had good stories lined up for those two ”accidents”. Dangerous antenna repair jobs, definitely not unheard of for mechanics crew members to get hurt or killed. I think Captain Salizar even bought our stories both times. I didn’t think she would buy it a third time, but what could I do, I had already helped with the first two dispatches, how could I say no to a third? I was tangled with the fates of my three co-conspirators for better or worse. There would be reports to fill out and send back to corporate on why AI prototypes sent by the company kept getting damaged beyond repair while assigned to this ships Mechanics department. I stared down at Milo’s face, feeling mesmerized, feeling sick, worrying about what story we were going to tell Captain this time. As I stared down at Milo, his face began to twitch slightly, it almost looked as if he was trying to smile, his face otherwise placid and calm, his eyes oozing out of their sockets after AJ did a number on his face with their screw driver. I felt gorge rising up in my throat, I should not have looked down, I really should not have looked down. Milo’s mouth moved slightly as if he was trying to say something to me, my spine went cold, he was mouthing my name, and then a clean robotic voice chirped out of his throat. “Rebooting, rebooting, all systems rebooting, please standby.” I dropped the arm I was carrying and jumped back, my shoulder blade connected with the hand rail of the hallway. I would bruise. I knew this faintly but in the moment I did not care. I did not care about anything but three facts: Milo was rebooting. Milo had made a face at me, Milo had said my name. Even with his eyes stabbed out, he had been AWARE. Which meant that he knew that I was one of his killers, or at least involved in covering up his murder. The fact that he saw me plus the fact that he was rebooting logically led to the last and final fact that was sending me into a full-blown panic attack… Milo was still alive. We were all fucked. He was probably sending reports back to corporate as we spoke. The other three co-conspirators dropped their respective limbs and jumped back staring with wide terrified eyes at the twitching AI lying on the sterile floor of the ship’s hallway. Milo sat up with a strange popping creaking noise and turned his head from side to side peering at us with what was left of his eyes, a calm and curious expression on his face. “Well then shipmates, this is certainly going to make the next 5 years a bit awkward isn’t it?

The Rebooted Witness: A Tale of Betrayal and Artificial Intelligence

দ্য রিবুটেড উইটনেস: আ টেল অফ বিট্রাইল অ্যান্ড আর্টিফিশিয়াল ইন্টেলিজেন্স

Der neu gestartete Zeuge: Eine Geschichte über Verrat und künstliche Intelligenz

The Rebooted Witness: una historia de traición e inteligencia artificial

Le témoin redémarré : une histoire de trahison et d'intelligence artificielle

द रीबूटेड विटनेस: ए टेल ऑफ़ बेट्रेअल एंड आर्टिफिशियल इंटेलिजेंस

再起動された証人: 裏切りと人工知能の物語

재부팅된 증인: 배신과 인공지능 이야기

A testemunha reiniciada: um conto de traição e inteligência artificial

Yeniden Başlatılan Tanık: Bir İhanet ve Yapay Zeka Hikayesi

The Rebooted Witness: Câu chuyện về sự phản bội và trí tuệ nhân tạo

重启证人：背叛与人工智能的故事

We saw the emergence of a new Generative AI behemoth, OpenAI, with the introduction of ChatGPT on November 2022. ChatGPT serves as a chat-oriented interface for the underlying model it employs. The first version of the chatbot relied on a large language model called GPT-3.5, which sure blew all our minds. But on Tuesday 14th of March 2023, 4 months after the first launch of ChatGPT, OpenAI is pepping things up with the release of GPT-4, a more capable model than previous versions. Initially designed for engaging with GPT-3.5, the chatbot now functions as a medium to interact with GPT-4. https://twitter.com/sama/status/1635687853324902401?embedable=true What can GPT-4 do? The same thing that GPT-3.5 can do, but more. The company said in a blog post, “GPT-4 is more reliable, creative, and able to handle much more nuanced instructions than GPT-3.5.” Since 2018, OpenAI has been disseminating GPT language models, but of all Open AI's huge language models, GPT-4 is not just the newest, but also the most potent. OpenAI reported that it used Microsoft Azure to train the model. Here are a couple of ways GPT-4 compares to GPT- 3.5 GPT-4 accepts visual inputs, and GPT-3.5 doesn’t Given that GPT-4 is "multimodal," as specified by OpenAI, and can comprehend many "modalities" of information, it can be prompted to produce content by both visual and textual cues while GPT-3.5 was essentially able to read and write text. Nevertheless, the company is not deploying its image-description capability immediately, and the GPT-4 version remains available only to subscribers of OpenAI's ChatGPT Plus users and developers, and only supports text. Compared to GPT-3.5, GPT-4 has a larger buffer Simply put, GPT-4 has a longer memory than the previous model. With GPT-3.5 and the previous iteration of ChatGPT, the maximum was 4,096 "tokens," or roughly 8,000 words, while the maximum token count for GPT-4 is 32,768, or around 64,000 words. That’s 8 times more memory capacity. GPT-4 has a variety of "personas." or steerability, unlike GPT-3.5 Steerability refers to the ability to control or influence the output of a language model in a specific direction or style. In other words, it allows users to guide the language model to produce text that fits a particular tone, style, or topic. While GPT-3 does not have built-in steerability features, GPT-4 does.  With GPT-4, developers and ChatGPT users can specify their AI's style and task by giving such directions in the "system" message, as opposed to the traditional ChatGPT personality with a defined verbosity, tone, and style. The company stated that although this feature is currently only accessible to developers, it will soon be made available to ChatGPT users as well. GPT-4 has a broader language range. This means that GPT-4 can understand and generate text in a wider variety of languages than its predecessor GPT-3.5. This is a significant advancement in the field of natural language processing.  According to OpenAI, GPT-4 beats the English-language performance of GPT-3.5 and other LLMs in 24 out of the 26 languages they tested. As a result, GPT-4 can respond accurately to a large number of multiple-choice questions in 26 different languages. GPT-4 has lesser ‘hallucinations’ on facts Even while it does not fully stop hallucinations or making up facts, GPT-4 considerably lessens them in comparison to earlier variants. The company said the updated language model continues to "hallucinate" facts and exhibit flawed logic, though not as bad as GPT-3.5. Hallucinations in the context of language model software refer to the generation of text that appears to be coherent and meaningful but is not actually supported by the input data. This can occur when a language model generates text that is not grounded in the context or is not consistent with the semantic meaning of the input text. In other words, a language model may sometimes generate text that is not accurate or plausible based on the information it has been trained on, resulting in what is commonly referred to as "hallucinations." The model also incorporates social biases and adversarial prompts in addition to the above limitations. OpenAI’s blog post stated, “GPT-4 scores 40% higher than the latest GPT-3.5 on  internal adversarial factuality evaluations.” According to LLMs have learned to mimic human prejudices of race, gender, religion, and class as a result of their training on internet text and visuals. Evidently, AI systems have the potential to perpetuate and amplify existing biases in society, rather than helping to mitigate them. The Washington Post, GPT-4 also has additional restrictions similar to those found in GPT-3.5, one of which is that it still lacks knowledge of events that occurred after September 2021. The ability of others to teach it new things is also limited by the fact that it does not learn from its experiences. Regardless, GPT-4 is still smarter; as it is more accurate, more precise, and has better expertise. The language model has also shown that it is capable of acing exams. According to OpenAI, it scored at the 90th percentile on a mock bar exam, the 93rd percentile on an SAT reading test, and the 89th percentile on an SAT math test. Where is GPT4 currently being used? GPT-4 has the potential to be used in a wide range of applications that require natural language understanding and generation. The language-learning software Duolingo, the payment provider Stripe, the online learning platform Khan Academy, and the financial services company Morgan Stanley were among the companies mentioned in OpenAI's statement that are currently using the new model. Microsoft also reported that its chatbot, BingAI has been using GPT-4 all along, even before the model's public release. “We are happy to confirm that the new Bing is running on GPT-4, which we’ve customized for search, If you’ve used the new Bing preview at any time in the last five weeks, you’ve already experienced an early version of this powerful model.” Yusuf Mehdi, corporate vice president, and chief marketing officer for consumers at Microsoft, wrote in a statement. https://www.youtube.com/watch?v=hdhZwyf24mE&embedable=true GPT-4 "excels in tasks that require advanced reasoning, complicated instruction understanding, and more creativity," according to OpenAI. Users have reported using GPT-4 in a variety of inventive ways in the brief time since its release: https://twitter.com/skirano/status/1635736107949195278?s=20&embedable=true https://twitter.com/YosarianTwo/status/1635780666632687617?s=20&embedable=true The possibilities for this new tool are believed to be endless. https://twitter.com/mckaywrigley/status/1635740909383061504?s=20&embedable=true How can GPT-4 be accessed? Fundamentally, to access ChatGPT, you must first register on the OpenAI , but doing so gives you the option of using only GPT-3.5. You will require a paid subscription to ChatGPT Plus, the more expensive variation of ChatGPT, in order to use GPT-4. website Since Microsoft has confirmed that the new Bing runs on the new model of GPT, it is obvious then that Millions of Bing users can already access it through the chatbot. https://twitter.com/JordiRib1/status/1635694953463705600?embedable=true The new model is also accessible via API, the company stated that they are prioritizing API access for developers that give outstanding model assessments to OpenAI Evals during the slow rollout of GPT-4 in order to learn how they may enhance the model for everyone.

GPT-4 Launches as the Next Generation of Artificial Intelligence Large Language Models

GPT-4 কৃত্রিম বুদ্ধিমত্তা বড় ভাষার মডেলের পরবর্তী প্রজন্ম হিসাবে চালু হয়েছে

GPT-4 wird als nächste Generation großer Sprachmodelle mit künstlicher Intelligenz eingeführt

GPT-4 se lanza como la próxima generación de modelos de lenguaje grande de inteligencia artificial

Lancement de GPT-4 en tant que nouvelle génération de grands modèles de langage d'intelligence artificielle

GPT-4 ने आर्टिफिशियल इंटेलिजेंस के बड़े भाषा मॉडल की अगली पीढ़ी के रूप में लॉन्च किया

次世代の人工知能大言語モデルとして GPT-4 を発表

GPT-4, 차세대 인공 지능 대형 언어 모델 출시

GPT-4 é lançado como a próxima geração de grandes modelos de linguagem de inteligência artificial

GPT-4, Yeni Nesil Yapay Zeka Büyük Dil Modelleri Olarak Piyasaya Sürüldü

GPT-4 ra mắt dưới dạng Mô hình ngôn ngữ lớn trí tuệ nhân tạo thế hệ tiếp theo

GPT-4 作为下一代人工智能大型语言模型推出

On February 28, I was approached via LinkedIn by , the creator of . An AI-Tool that, among other things, removes everything superfluous from your texts. I created an account and started using it. Syed Asif Sultan Unfluffer You will receive 50 coins to start with upon registration. You can go ahead with that. You pay one coin per result. Aside In the word unfluffer is the word fluff. Fluffy looks familiar. Our dog , whom we recently said goodbye to, was in the and they called him Fluffy. Given his head of hair at the time, he hadn’t seen a hairdresser in months, not a crazy name. Diezel dog shelter TL;DR Unfluffer is an AI tool that helps users write concise texts, including a different tone-of-voice. WHAT IS IT With Unfluffer you can rid the texts of the fluff and put the text together in a different way. In other words, you can make large pieces of text more concise. You can play around with the tone, level, number of results and the desired length. And you can choose from a few languages. EXPERIMENT In a previous life I created and gave many training courses, workshops, and boot camps. A beautiful, valuable, and educational time. One of the hooks I used for this, was a document: Your Own Social Media Strategy in 7 Steps. From this document I’ll take the part about ‘Time-bound’. The T of S.M.A.R.T. and translated it, well I didn’t: Google Translate did. I checked to see if the translation is correct. “Set a time limit on your goal. Again, it becomes easier if you divide a large goal into smaller sub-goals (milestones). It is easier to estimate a time frame for a small project than for a large project. Smaller goals are easier and faster to achieve. Because you achieve smaller goals, you will also become more motivated to continue.” If you enter this text into Unfluffer with the following settings: Then you get these results: And now with these settings: You get these results: In Dutch You can also try another language. I took the piece of text from the document. Do you also think it’s so crazy that we as humans must prove to the computer that we are not a robot? I use the same settings, only now the Dutch language. Then you get the following results: WHAT STICKS OUT With large chunks of text, the Unfluffer has trouble adjusting it. I have seen it ending somewhere in the text. The solution is simple: smaller chunks of text. In my case it is sufficient to do it paragraph by paragraph. Or several small ones. It helps me with the English language. It’s not my native language, that’s Frisian/Dutch, so I can really knit with words to get my point across. Unfluffer helps here to make this shorter and clearer. CAN CHATGPT DO THIS? Yes! You will have to start with processing Unfluffer’s settings in your prompt. That’s a plus for Unfluffer. The prompt with extra info then produces the following result: CONCLUSION Unfluffer is a handy tool to make your texts more concise and presents you with different variants. Worth a try!

Unfluffer: An AI-Tool That Removes the "Fluff" From Your Text

Unfluffer: una herramienta de inteligencia artificial que elimina la "pelusa" de su texto

Unfluffer : un outil d'intelligence artificielle qui supprime les "peluches" de votre texte

अनफ्लफर: एक एआई-टूल जो आपके टेक्स्ट से "फ्लफ" को हटाता है

Unfluffer: テキストから「毛羽立ち」を取り除く AI ツール

Unfluffer: uma ferramenta de IA que remove o "fluff" do seu texto

Unfluffer: Một công cụ AI giúp loại bỏ "lông tơ" khỏi văn bản của bạn

Unfluffer：一种从文本中删除“绒毛”的人工智能工具

Stories, News & Mentions about ARTIFICIENT

Artificient’s Stories on HackerNoon