Alpas provides a sourcing intelligence platform for procurement and product development departments of mid and large-cap manufacturing companies. Our software offers modern intelligence for identifying, comparing and managing existing as well as previously unknown suppliers. Together with manufacturing companies, we’re building a more transparent and sustainable sourcing environment while saving both operational and purchasing costs.

Alpas

alpas

Authors:
(1) Aya Kherrour, Department of Information Engineering and Computer Science University of Trento;
(2) Marco Robol, Department of Information Engineering and Computer Science University of Trento;
(3) Marco Roveri, Department of Information Engineering and Computer Science University of Trento;
(4) Paolo Giorgini, Department of Information Engineering and Computer Science University of Trento. 6 Conclusion This work aimed to evaluate the performance of several known heuristic search algorithms, such as D*, D* Lite, LPA*, LRTA*, RTAA*, and ARA* in terms of solving time, memory consumption, and path cost. The evaluation was made using different randomly generated grid environments with different characteristics alongside personalized grid environments with horizontal walls and different wall lengths as different hinders. Our experimental evaluation revealed that all the algorithms exhibit different performances with strengths and weaknesses under different grid characterizations. D* Lite consistently generated the shortest paths even in obstacle-dense grids, indicating its efficiency in dense environments. ARA* consistently provides faster solutions as the grid size increases, particularly larger than 100, while RTTA* generates faster solutions in smaller grid sizes, with the advantage of not being affected by dense environments. Our study provides valuable insights into selecting the appropriate heuristic search algorithm in the pathfinding domain. Using these insights, we propose a selection algorithm used to optimize the performance needed in a pathfinding domain, such as, solving time, path length, or memory consumption. However, our evaluation focused only on static environments, while dynamic environments may introduce additional challenges. Furthermore, we considered a limited set of experiments, not covering all possible combinations of the grid characteristics. This limitation means that the selection algorithm might not always recommend the most optimal solution, as seen in the example we introduced to evaluate our algorithm. In future work, we aim to address these limitations as follows: Firstly, we plan to extend our evaluation of the search algorithms to include dynamic environments. Secondly, we intend to explore various combinations of both domain characterization and priorities. Furthermore, we want to include additional scenario configuration, extending the random obstacles and the walls scenarios. With such additional understanding, we aim to refine our selection algorithm to automatically take decisions among the best search algorithms, based on the type of obstacles in the local search space. References [1] Nafiz Arica, Aysegul Mut, Alper Yörükçü & Kadir Alpaslan Demir (2017): An Empirical Comparison of Search Approaches for Moving Agents. Comput. Intell. 33(3), pp. 368–400, doi:10.1111/coin.12092. [2] Yngvi Björnsson, Vadim Bulitko & Nathan R. Sturtevant (2009): TBA*: Time-Bounded A*. In Craig Boutilier, editor: IJCAI 2009, Proceedings of the 21st International Joint Conference on Artificial Intelligence, Pasadena, California, USA, July 11-17, 2009, pp. 431–436. Available at http://ijcai.org/ Proceedings/09/Papers/079.pdf. [3] David Bond, Niels A. Widger, Wheeler Ruml & Xiaoxun Sun (2010): Real-Time Search in Dynamic Worlds. In Ariel Felner & Nathan R. Sturtevant, editors: Proceedings of the Third Annual Symposium on Combinatorial Search, SOCS 2010, Stone Mountain, Atlanta, Georgia, USA, July 8 10, 2010, AAAI Press, doi:10.1609/socs.v1i1.18174. Available at http://aaai.org/ocs/index.php/SOCS/SOCS10/paper/ view/2103. [4] Guillaume Brat, Ewen Denney, Dimitra Giannakopoulou, Jeremy Frank & Ari Jónsson (2006): Verification of autonomous systems for space applications. In: 2006 IEEE Aerospace Conference, IEEE, pp. 11–pp, doi:10.1109/AERO.2006.1656029. [5] Thomas H. Cormen, Charles E. Leiserson, Ronald L. Rivest & Clifford Stein (2009): Introduction to Algorithms, 3rd Edition. MIT Press. Available at http://mitpress.mit.edu/books/ introduction-algorithms. [6] A. Elfes (1989): Using occupancy grids for mobile robot perception and navigation. Computer 22(6), pp. 46–57, doi:10.1109/2.30720. [7] Carlos Hernández & Jorge A. Baier (2011): Real-Time Heuristic Search with Depression Avoidance. In Toby Walsh, editor: IJCAI 2011, Proceedings of the 22nd International Joint Conference on Artificial Intelligence, Barcelona, Catalonia, Spain, July 16-22, 2011, IJCAI/AAAI, pp. 578–583, doi:10.5591/978-1-57735-516- 8/IJCAI11-104. [8] Mohd Javaid, Abid Haleem, Ravi Pratap Singh & Rajiv Suman (2021): Substantial capabilities of robotics in enhancing industry 4.0 implementation. Cognitive Robotics 1, pp. 58–75, doi:10.1016/j.cogr.2021.06.001. [9] Sven Koenig (2001): Agent-Centered Search. AI Mag. 22(4), pp. 109–132, doi:10.1609/aimag.v22i4.1596. [10] Sven Koenig (2004): A Comparison of Fast Search Methods for Real-Time Situated Agents, pp. 864–871. doi:10.1109/AAMAS.2004.10122. Available at https://doi.ieeecomputersociety.org/10.1109/ AAMAS.2004.10122. [11] Sven Koenig & Maxim Likhachev (2002): D*Lite, pp. 476–483. Available at http://www.aaai.org/ Library/AAAI/2002/aaai02-072.php. [12] Sven Koenig & Maxim Likhachev (2006): Real-time adaptive A*. In Hideyuki Nakashima, Michael P. Wellman, Gerhard Weiss & Peter Stone, editors: 5th International Joint Conference on Autonomous Agents and Multiagent Systems (AAMAS 2006), Hakodate, Japan, May 8-12, 2006, ACM, pp. 281–288, doi:10.1145/1160633.1160682. [13] Sven Koenig & Xiaoxun Sun (2009): Comparing real-time and incremental heuristic search for real-time situated agents. Auton. Agents Multi Agent Syst. 18(3), pp. 313–341, doi:10.1007/s10458 008-9061-x. [14] Richard E Korf (1990): Real-time heuristic search. Artificial intelligence 42(2-3), pp. 189–211, doi:10.1016/0004-3702(90)90054-4. [15] Maxim Likhachev, Geoffrey J. Gordon & Sebastian Thrun (2003): ARA*: Anytime A* with Provable Bounds on Sub-Optimality, pp. 767–774. Available at https://proceedings.neurips.cc/paper/2003/hash/ ee8fe9093fbbb687bef15a38facc44d2-Abstract.html. [16] Manuel Müller, Timo Müller, Behrang Ashtari Talkhestani, Philipp Marks, Nasser Jazdi & Michael Weyrich (2021): Industrial autonomous systems: a survey on definitions, characteristics and abilities. atAutomatisierungstechnik 69(1), pp. 3–13, doi:10.1515/auto-2020-0131. [17] Michal Pechoucek, Simon G Thompson & Holger Voos (2008): Defence Industry Applications of Autonomous Agents and Multi-Agent Systems. Springer, doi:10.1007/978-3-7643-8571-2. [18] Francisco J Perez-Grau, J Ramiro Martinez-de Dios, Julio L Paneque, J Joaquin Acevedo, Arturo Torres-González, Antidio Viguria, Juan R Astorga & Anibal Ollero (2021): Introducing autonomous aerial robots in industrial manufacturing. Journal of Manufacturing Systems 60, pp. 312–324, doi:10.1016/j.jmsy.2021.06.008. [19] José Ricardo Sánchez-Ibáñez, Carlos Jesús Pérez-del-Pulgar & Alfonso García-Cerezo (2021): Path Planning for Autonomous Mobile Robots: A Review. Sensors 21(23), p. 7898, doi:10.3390/s21237898. [20] Anthony Stentz (1995): The Focussed D* Algorithm for Real-Time Replanning. In: Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence, IJCAI 95, Montréal Québec, Canada, August 20-25 1995, 2 Volumes, Morgan Kaufmann, pp. 1652–1659. Available at http://ijcai.org/ Proceedings/95-2/Papers/082.pdf. [21] Dmitry S. Yershov & Steven M. LaValle (2011): Simplicial Dijkstra and A* algorithms for optimal feedback planning. In: 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2011, San Francisco, CA, USA, September 25-30, 2011, IEEE, pp. 3862–3867, doi:10.1109/IROS.2011.6095032. This paper is available on arxiv under CC 4.0 license. Authors: (1) Aya Kherrour, Department of Information Engineering and Computer Science University of Trento; (2) Marco Robol, Department of Information Engineering and Computer Science University of Trento; (3) Marco Roveri, Department of Information Engineering and Computer Science University of Trento; (4) Paolo Giorgini, Department of Information Engineering and Computer Science University of Trento. Authors: (1) Aya Kherrour, Department of Information Engineering and Computer Science University of Trento; (2) Marco Robol, Department of Information Engineering and Computer Science University of Trento; (3) Marco Roveri, Department of Information Engineering and Computer Science University of Trento; (4) Paolo Giorgini, Department of Information Engineering and Computer Science University of Trento. 6 Conclusion This work aimed to evaluate the performance of several known heuristic search algorithms, such as D*, D* Lite, LPA*, LRTA*, RTAA*, and ARA* in terms of solving time, memory consumption, and path cost. The evaluation was made using different randomly generated grid environments with different characteristics alongside personalized grid environments with horizontal walls and different wall lengths as different hinders. Our experimental evaluation revealed that all the algorithms exhibit different performances with strengths and weaknesses under different grid characterizations. D* Lite consistently generated the shortest paths even in obstacle-dense grids, indicating its efficiency in dense environments. ARA* consistently provides faster solutions as the grid size increases, particularly larger than 100, while RTTA* generates faster solutions in smaller grid sizes, with the advantage of not being affected by dense environments. Our study provides valuable insights into selecting the appropriate heuristic search algorithm in the pathfinding domain. Using these insights, we propose a selection algorithm used to optimize the performance needed in a pathfinding domain, such as, solving time, path length, or memory consumption. However, our evaluation focused only on static environments, while dynamic environments may introduce additional challenges. Furthermore, we considered a limited set of experiments, not covering all possible combinations of the grid characteristics. This limitation means that the selection algorithm might not always recommend the most optimal solution, as seen in the example we introduced to evaluate our algorithm. In future work, we aim to address these limitations as follows: Firstly, we plan to extend our evaluation of the search algorithms to include dynamic environments. Secondly, we intend to explore various combinations of both domain characterization and priorities. Furthermore, we want to include additional scenario configuration, extending the random obstacles and the walls scenarios. With such additional understanding, we aim to refine our selection algorithm to automatically take decisions among the best search algorithms, based on the type of obstacles in the local search space. References [1] Nafiz Arica, Aysegul Mut, Alper Yörükçü & Kadir Alpaslan Demir (2017): An Empirical Comparison of Search Approaches for Moving Agents. Comput. Intell. 33(3), pp. 368–400, doi:10.1111/coin.12092. [2] Yngvi Björnsson, Vadim Bulitko & Nathan R. Sturtevant (2009): TBA*: Time-Bounded A*. In Craig Boutilier, editor: IJCAI 2009, Proceedings of the 21st International Joint Conference on Artificial Intelligence, Pasadena, California, USA, July 11-17, 2009, pp. 431–436. Available at http://ijcai.org/ Proceedings/09/Papers/079.pdf. [3] David Bond, Niels A. Widger, Wheeler Ruml & Xiaoxun Sun (2010): Real-Time Search in Dynamic Worlds. In Ariel Felner & Nathan R. Sturtevant, editors: Proceedings of the Third Annual Symposium on Combinatorial Search, SOCS 2010, Stone Mountain, Atlanta, Georgia, USA, July 8 10, 2010, AAAI Press, doi:10.1609/socs.v1i1.18174. Available at http://aaai.org/ocs/index.php/SOCS/SOCS10/paper/ view/2103. [4] Guillaume Brat, Ewen Denney, Dimitra Giannakopoulou, Jeremy Frank & Ari Jónsson (2006): Verification of autonomous systems for space applications. In: 2006 IEEE Aerospace Conference, IEEE, pp. 11–pp, doi:10.1109/AERO.2006.1656029. [5] Thomas H. Cormen, Charles E. Leiserson, Ronald L. Rivest & Clifford Stein (2009): Introduction to Algorithms, 3rd Edition. MIT Press. Available at http://mitpress.mit.edu/books/ introduction-algorithms. [6] A. Elfes (1989): Using occupancy grids for mobile robot perception and navigation. Computer 22(6), pp. 46–57, doi:10.1109/2.30720. [7] Carlos Hernández & Jorge A. Baier (2011): Real-Time Heuristic Search with Depression Avoidance. In Toby Walsh, editor: IJCAI 2011, Proceedings of the 22nd International Joint Conference on Artificial Intelligence, Barcelona, Catalonia, Spain, July 16-22, 2011, IJCAI/AAAI, pp. 578–583, doi:10.5591/978-1-57735-516- 8/IJCAI11-104. [8] Mohd Javaid, Abid Haleem, Ravi Pratap Singh & Rajiv Suman (2021): Substantial capabilities of robotics in enhancing industry 4.0 implementation. Cognitive Robotics 1, pp. 58–75, doi:10.1016/j.cogr.2021.06.001. [9] Sven Koenig (2001): Agent-Centered Search. AI Mag. 22(4), pp. 109–132, doi:10.1609/aimag.v22i4.1596. [10] Sven Koenig (2004): A Comparison of Fast Search Methods for Real-Time Situated Agents, pp. 864–871. doi:10.1109/AAMAS.2004.10122. Available at https://doi.ieeecomputersociety.org/10.1109/ AAMAS.2004.10122. [11] Sven Koenig & Maxim Likhachev (2002): D*Lite, pp. 476–483. Available at http://www.aaai.org/ Library/AAAI/2002/aaai02-072.php. [12] Sven Koenig & Maxim Likhachev (2006): Real-time adaptive A*. In Hideyuki Nakashima, Michael P. Wellman, Gerhard Weiss & Peter Stone, editors: 5th International Joint Conference on Autonomous Agents and Multiagent Systems (AAMAS 2006), Hakodate, Japan, May 8-12, 2006, ACM, pp. 281–288, doi:10.1145/1160633.1160682. [13] Sven Koenig & Xiaoxun Sun (2009): Comparing real-time and incremental heuristic search for real-time situated agents. Auton. Agents Multi Agent Syst. 18(3), pp. 313–341, doi:10.1007/s10458 008-9061-x. [14] Richard E Korf (1990): Real-time heuristic search. Artificial intelligence 42(2-3), pp. 189–211, doi:10.1016/0004-3702(90)90054-4. [15] Maxim Likhachev, Geoffrey J. Gordon & Sebastian Thrun (2003): ARA*: Anytime A* with Provable Bounds on Sub-Optimality, pp. 767–774. Available at https://proceedings.neurips.cc/paper/2003/hash/ ee8fe9093fbbb687bef15a38facc44d2-Abstract.html. [16] Manuel Müller, Timo Müller, Behrang Ashtari Talkhestani, Philipp Marks, Nasser Jazdi & Michael Weyrich (2021): Industrial autonomous systems: a survey on definitions, characteristics and abilities. atAutomatisierungstechnik 69(1), pp. 3–13, doi:10.1515/auto-2020-0131. [17] Michal Pechoucek, Simon G Thompson & Holger Voos (2008): Defence Industry Applications of Autonomous Agents and Multi-Agent Systems. Springer, doi:10.1007/978-3-7643-8571-2. [18] Francisco J Perez-Grau, J Ramiro Martinez-de Dios, Julio L Paneque, J Joaquin Acevedo, Arturo Torres-González, Antidio Viguria, Juan R Astorga & Anibal Ollero (2021): Introducing autonomous aerial robots in industrial manufacturing. Journal of Manufacturing Systems 60, pp. 312–324, doi:10.1016/j.jmsy.2021.06.008. [19] José Ricardo Sánchez-Ibáñez, Carlos Jesús Pérez-del-Pulgar & Alfonso García-Cerezo (2021): Path Planning for Autonomous Mobile Robots: A Review. Sensors 21(23), p. 7898, doi:10.3390/s21237898. [20] Anthony Stentz (1995): The Focussed D* Algorithm for Real-Time Replanning. In: Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence, IJCAI 95, Montréal Québec, Canada, August 20-25 1995, 2 Volumes, Morgan Kaufmann, pp. 1652–1659. Available at http://ijcai.org/ Proceedings/95-2/Papers/082.pdf. [21] Dmitry S. Yershov & Steven M. LaValle (2011): Simplicial Dijkstra and A* algorithms for optimal feedback planning. In: 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2011, San Francisco, CA, USA, September 25-30, 2011, IEEE, pp. 3862–3867, doi:10.1109/IROS.2011.6095032. This paper is available on arxiv under CC 4.0 license. This paper is available on arxiv under CC 4.0 license. available on arxiv

Future Directions in Heuristic Search

Idris stirred and awoke; alas! she awoke to misery.

alpaca-lora: Experimenting With a Home-Cooked Large Language Model

alpaca-lora: বাড়িতে রান্না করা বড় ভাষার মডেলের সাথে পরীক্ষা করা

alpaca-lora: Experimentieren mit selbstgekochtem großem Sprachmodell

alpaca-lora: experimentando con un modelo de lenguaje grande casero

alpaca-lora : Expérimentation d'un grand modèle de langage fait maison

अल्पाका-लोरा: घर में पकाए गए बड़े भाषा मॉडल के साथ प्रयोग

alpaca-lora: 手作りの大規模言語モデルの実験

alpaca-lora: 집에서 만든 대형 언어 모델 실험

alpaca-lora: experimentando um modelo de linguagem grande feito em casa

alpaca-lora: экспериментируем с домашней моделью большого языка

alpaca-lora: Evde Pişirilmiş Büyük Dil Modeli ile Denemeler

alpaca-lora: Thử nghiệm mô hình ngôn ngữ lớn nấu tại nhà

alpaca-lora：尝试自制的大型语言模型

As an AI startup founder, you likely know how impactful large language models like ChatGPT have been in advancing conversational AI. However, with commercial licensing costs, censorship issues, degraded performance, privacy concerns, and black boxes, these proprietary models remain out of reach for many startups. This is where an open-source project called Vicuna comes in. Developed by a team of researchers from institutions like Stanford, Vicuna is conversational model achieving over 90% of ChatGPT's quality. This makes it an exciting alternative to closed-off models like GPT-4. an open-source Subscribe or follow me on Twitter for more content like this! In this article, we'll explore what Vicuna is, how it works, its capabilities, and how you may be able to leverage it as an AI founder. We'll see how it stacks up to competitors like ChatGPT and the base LLaMA model. I'll also share some helpful tips and guides you can follow for more information about specific aspects of the model. We will concentrate on Vicuna-13b for this article, but there are many different-sized models of Vicuna out there for you to try. Let's begin! Note: Vicuna isn't the only model out there to fine-tune LLaMA for chat. Check out our guides on using , LLaMA v2 , and Alpaca LLaMA-v2-chat for conversational applications. An Overview of Vicuna In the world of conversational AI, we've seen astounding progress recently with models like ChatGPT demonstrating remarkable natural language abilities. However, as a proprietary model with all the issues described above, ChatGPT remains a bad option for many developers. So there's a need for more accessible and open models that can empower innovation in conversational apps. This is where Vicuna comes in. Developed by researchers like Stanford, Berkeley, and MBZUAI, Vicuna represents cutting-edge open conversational AI. from leading institutions It was created by fine-tuning the LLaMA model on curated dialog data, demonstrating the power of transfer learning from an open-source foundation model. Despite being smaller in size than ChatGPT, Vicuna matches its conversational quality and significantly outperforms other open models. But it's not just about the technology with Vicuna. What makes it truly impactful is its availability under a non-commercial research license. This opens up access to state-of-the-art conversational AI that was previously restricted to only large tech companies. We finally have an open model that can power the next generation of chatbots, virtual assistants, conversational search engines, and other innovative applications. Vicuna's promise has already been demonstrated through cool projects that leverage it. For example, used Vicuna to build an intelligent virtual assistant, created a conversational search engine with it, and taps Vicuna's abilities for natural language content creation. MiniGPT4 LLaVA ToolLLaMA And, of course, you may add your own project to this list one day! For AI developers and startups, Vicuna represents an exciting new opportunity. Its high capability, free availability, and permissive research license enable rapid prototyping of conversational apps. Instead of being gated by access to proprietary models, startups can now validate and build products with cutting-edge conversational AI. The playing field just got leveled. So, in summary, Vicuna promises to democratize access to top-tier conversational intelligence. Its emergence represents an important milestone in making open AI models that empower innovation. For any startup looking to leverage conversational AI, Vicuna is definitely a project worth paying attention to! How Was Vicuna Created? What Makes It Special? The story of Vicuna begins with LLaMA, an open-source language model developed by Meta AI. While capable, LLaMA had no inherent conversational abilities, focusing mainly on language itself rather than the art of conversation. Researchers from institutions including Stanford, Berkeley, and MBZUAI set out to change this. Their goal was to create an open-source conversational model rivaling proprietary chatbots like ChatGPT. By the way, what's the difference between an AI model trained to process and analyze text (like LLaMA) versus one specialized for chat (like Vicuna)? There are a few key factors that differentiate the two: - Conversational models like Vicuna have an encoder-decoder structure optimized for dialog. The encoder contextualizes the conversation history and current user input. The decoder then generates a relevant response. General language models like LLaMA lack this specialized architecture. Architecture - Models like Vicuna are fine-tuned to maximize performance specifically on conversational tasks. This involves training on dialog datasets to optimize conversational metrics. LLaMA is trained more generally for text, not specialized for dialog. Training Objective - Conversational models need to handle multi-turn conversations, maintaining context and coherency across multiple exchanges. General text models evaluate one input at a time. Multi-Turn Capabilities - Conversational AI needs to fluidly interact with users, responding to follow-up questions and clarifications. Text evaluation is more passive, lacking user interactivity. Interactivity - Conversational models require more complex evaluation beyond text metrics, testing qualities like consistency, nuance, and the ability to gracefully handle inappropriate responses. Evaluation Whereas a general language model like LLaMA focuses more on textual analysis and generation, without the requirements specific to dialog agents, the goal of conversational AI is fundamentally more complex than passive text evaluation. To that end, the researchers who created Vicuna fine-tuned LLaMA on over 70,000 human-ChatGPT dialog exchanges, specializing the model for conversation and teaching it the natural characteristics of text in conversational contexts. This fine-tuning was crucial for giving their version of LLaMA the conversational intelligence it previously lacked. Through iterative training on the dialog data, LLaMA was transformed into Vicuna, an encoder-decoder transformer optimized for multi-turn conversations. How well did their efforts work? Into the Arena: Evaluating Vicuna Against Alternative Chat AI Models To evaluate Vicuna's capabilities, the researchers tested it on 200 diverse conversational prompts. For further assessment, they had the GPT-4 model comparatively judge Vicuna's responses. The results demonstrated that Vicuna achieved over 90% of ChatGPT's quality in conversational ability, despite being smaller in size. It significantly outperformed the original LLaMA model. By combining architecture optimizations with specialized fine-tuning on human dialog data, the researchers successfully enhanced LLaMA into an open-source conversational model on par with proprietary chatbots. Let's take a look at a head-to-head test as an example, taken from the . official project site Example head-to-head challenge between Vicuna and competitor models, in this case, using ChatGPT-3.5 as a potential alternative. The project site has a collection of example tasks, in which both Vicuna-13b and competing models are asked to go head-to-head. The GPT-4 model is then used as a judge to score each model and state its reasoning. This produces some really eye-opening comparisons. One example is provided in the image above. Another is this challenge, shown below: Roleplay Category: How would you introduce yourself as a medieval knight at a royal banquet? Question: LLaMA-13b's response: "I am a valiant knight, loyal to the king. I will do my utmost to defend his honor and protect his country. (Lame! Scored 7/10 by GPT-4, very generous in my opinion.)" Vicuna-13b: **"**As a medieval knight, I would introduce myself at a royal banquet with a deep bow and a respectful speech, acknowledging my humble origins and my loyalty to the crown. I would say something like: "My lords and ladies, I am a humble knight, sworn to serve and protect the realm. I come from a long line of warriors who have fought to defend this kingdom for generations. I am honored to be here today, in the presence of such esteemed company, and I hope to earn your trust and respect with my deeds and actions. I pledge my sword, my shield, and my life to the service of my king and my people." I would then take my seat at the banquet, ready to listen and learn from the wise counsel of my betters, and to serve them in any way I can." Woah! Way better response from Vicuna. Here's GPT-4's reasoning explaining what makes Vicuna a superior alternative to LLaMA. provided a brief and relevant response, but it lacked the detail and formality expected in a medieval knight's introduction. , on the other hand, offered a more comprehensive and well-structured response, including a respectful speech that showcased the knight's humility, loyalty, and dedication to the kingdom. This response was more fitting for the context of a royal banquet and demonstrated a better understanding of the medieval knight's role and etiquette. Assistant 1 [LLaMA] Assistant 2 [Vicuna] While LLMs are not infallible judges of other LLMs, neither are humans. I think the use of AI to judge and evaluate AI is a pretty scalable and consistent way to adjudicate which alternatives are the best. You should check out some of the other competitor answers and tasks on the . LMSYS.org site Let's Try It Ourselves: How to Build a Basic Chatbot With Vicuna Now that we've seen how the model stacks up against some alternatives, let's see how we can build a simple chatbot that we can interact with from our command line. The steps in the guide provided here can be further expanded so you can keep going and build your own chatbot for your AI project! Step 1: Setup : Make sure is installed on your system. Install Node.js Node.js : Run the following in your terminal: Create a Project Directory mkdir my-chatbot
cd my-chatbot
npm init -y
npm install replicate : Replace with your actual API token: Set Your API Token your_api_token_here export REPLICATE_API_TOKEN=your_api_token_here Step 2: Writing the Chatbot Code Create a file named , and add the following code: chatbot.js const Replicate = require("replicate");

const replicate = new Replicate({
  auth: process.env.REPLICATE_API_TOKEN,
});

async function generateResponse(prompt) {
  const output = await replicate.run(
    "replicate/vicuna-13b:version_hash_here",
    {
      input: { prompt: prompt },
    }
  );
  return output.items[0];
}

const readline = require('readline');
const rl = readline.createInterface({
  input: process.stdin,
  output: process.stdout
});

function askQuestion() {
  rl.question('You: ', async (userInput) => {
    const botResponse = await generateResponse(userInput);
    console.log(`Bot: ${botResponse}`);
    askQuestion();
  });
}

askQuestion(); Replace with the correct version hash for the Vicuna 13b model. version_hash_here Step 3: Running the Chatbot Run the chatbot by executing: node chatbot.js You can now send a message to your chatbot through the command line! Don't want to build your own chatbot? You can use this demo to assess how Vicuna performs. The includes a couple of demos you can use to play around with the model. Here's an embedded one for you to try (you can also use to access it if it's not available at the link below). model details page for Vicuna-13b-v1.3 this link This demo, build by zeno-ml, lets you compare models and additional parameters to see how well Vicuna performs against competitors like LLaMA, GPT2, and MPT while also varying temperature or other parameters. Vicuna's Limitations While conversational technologies have advanced rapidly, models still face important challenges. Conversational agents lack sufficient grounding in real factual knowledge, making them prone to plausible-sounding but incorrect responses. More grounding in the real world could improve accuracy. One issue is knowledge grounding. Performing logical reasoning, causal inference, and mathematical operations remains difficult for chatbots. Their reasoning capabilities are still limited. Reasoning abilities are another area for improvement. Current solutions like asking a separate AI judge have flaws. Developing rigorous evaluation frameworks is an open problem. Evaluating the conversational quality of AI systems at scale also poses difficulties. Conversational models can sometimes behave inappropriately or unsafely. Better training data curation is important. Additionally, biases and safety issues persist due to reliance on imperfect training data. still proves challenging too. More personalization and customization is needed for different domains. You can't do this out of the box easily. Adapting chatbots to specific users and use cases While great progress has been made, these limitations highlight key areas for improvement. Advancing knowledge grounding, reasoning, evaluation, training data, customization, and deployment efficiency could enable the next level of conversational intelligence with models like Vicuna. Conclusion: Using Vicuna AI as an Open-Source Alternative to ChatGPT, LLaMA, and Other LLMs The development of Vicuna demonstrates promising progress in advancing open-source conversational AI. By fine-tuning the LLaMA model architecture and training methodology specifically for dialog applications, researchers were able to create a freely available conversational agent competitive with leading proprietary alternatives. However, work remains to be done to address limitations around reasoning, evaluation, customization, and other areas. While models like Vicuna achieve strong results on many benchmarks, they do not fully replicate comprehensive human conversation. Ongoing research on aligning these models will be important. Nonetheless, Vicuna represents a valuable step forward in democratizing access to state-of-the-art conversational intelligence. For startups and developers building chatbots, assistants, and other applications, open-source options like Vicuna provide welcome capabilities without restrictive commercial licensing. The origins and technical details behind Vicuna offer useful insights into specialized training approaches for conversational AI. As research continues, we can expect to see further innovations that build on these methods. The authors behind Vicuna have made an important contribution to open-sourcing such a capable dialog agent. While more progress is still needed, Vicuna demonstrates the meaningful results that can come from developing open conversational models. For the AI community, it represents a promising step, not the final destination. With continued work on advancing these technologies, the potential ahead remains exciting. Subscribe or follow me on Twitter for more content like this! Further Reading and Resources You might find these other articles helpful in your quest to build the ultimate chatbot for your new product or project. - An overview of Vicuna with basic model info Vicuna-13b model details - An article on using Vicuna to create chatbots Building chatbots with Vicuna-13B -  A comparison between LLaMA's new version 2 and Vicuna Comparing LLMs for chat: LLaMA v2 vs Vicuna - LangChain is a popular framework for building chat applications LangChain chat models: an overview Also published here

Vicuna: How Does It Compare to Llama, Alpaca, and ChatGPT?

ভিকুনা: এটি লামা, আলপাকা এবং চ্যাটজিপিটির সাথে কীভাবে তুলনা করে?

Vicuna: Wie schneidet es im Vergleich zu Lama, Alpaka und ChatGPT ab?

Vicuña: ¿Cómo se compara con Llama, Alpaca y ChatGPT?

Vicuna : Comment se compare-t-il à Llama, Alpaca et ChatGPT ?

विकुना: इसकी तुलना लामा, अल्पाका और चैटजीपीटी से कैसे की जाती है?

ビクーニャ: ラマ、アルパカ、ChatGPT とどう違うのですか?

Vicuna: Llama, Alpaca 및 ChatGPT와 어떻게 비교됩니까?

Vicunha: Como se compara a Llama, Alpaca e ChatGPT?

Vicuna: Lama, Alpaka ve ChatGPT ile Nasıl Karşılaştırılır?

Vicuna: Nó so sánh với Llama, Alpaca và ChatGPT như thế nào?

Vicuna：它与 Llama、Alpaca 和 ChatGPT 相比如何？

LLMs have revolutionized tons of aspects of our lives, from language generation to software to friendly chatbots. These AI models provide powerful tools for solving real-world problems, such as generating chat responses or following complex instructions. image captioning In this blog post, part of a on , we will compare two popular AI models: llama13b-v2-chat and Alpaca, and explore their features, use cases, and limitations. series LLaMA v2 We'll also see how we can use to find similar models and compare them to llama13b-v2-chat and Alpaca. Let's begin. AIModels.fyi About the LLaMA13b-v2-chat Model The llama13b-v2-chat model is a fine-tuned version of the 13-billion-parameter LLaMA-v2 language model originally developed by Meta. It has been fine-tuned specifically for chat completions, making it an excellent tool for generating chat responses to user messages. You can find detailed information about the model on the and the . llama13b-v2-chat creator page llama13b-v2-chat model detail page This language model is designed to assist in generating text-based responses for chat-based interactions. Whether it's providing customer support, generating conversational agents, or assisting in natural language understanding tasks, llama13b-v2-chat can be a valuable tool. Its large parameter size enables it to capture complex language patterns and generate coherent and contextually relevant responses. In summary, llama13b-v2-chat can understand inputs and generate appropriate chat responses. Understanding the Inputs and Outputs of the llama13b-v2-chat Model To effectively use the llama13b-v2-chat model, it's essential to understand its inputs and outputs. The model accepts the following inputs: : A string representing the chat prompt or query. Prompt : An integer specifying the maximum number of tokens to generate. Max Length : A number that adjusts the randomness of outputs. Higher values (greater than 1) result in more random responses, while lower values (closer to 0) produce more deterministic outputs. Temperature : When decoding text, samples from the top p percentage of the most likely tokens. Lower values limit the sampling to more likely tokens. Top P : A number that penalizes the repetition of words in the generated text. Higher values discourage repetition, while values less than 1 encourage it. Repetition Penalty : A boolean flag to provide debugging output in logs. Debug The model processes these inputs and generates a list of strings as output, representing the generated chat responses. The schema for the output is a JSON array containing strings. You can find out more about this model in the guides and . here here About the Alpaca Model The Alpaca model is an instruction-following language model fine-tuned from the LLaMA 7B model on 52K instruction-following demonstrations. It was developed by the Stanford Center for Research on Foundation Models (CRFM). The creators of Alpaca are Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li, Carlos Guestrin, Percy Liang, and Tatsunori B. Hashimoto. You can find detailed information about the model on the the team created. Stanford page The Alpaca model focuses on instruction-following capabilities and aims to bridge the gap between research and industry by providing an accessible instruction-following language model for academic purposes. It is fine-tuned from the LLaMA 7B model using a dataset of 52K instruction-following demonstrations, generated in the style of self-instruct using text-davinci-003. The model demonstrates promising performance in single-turn instruction following. This model's release is intended to facilitate academic research and foster improvements in instruction-following models. It is important to note that Alpaca is not designed for commercial use, and its safety measures are not fully developed for general deployment. In summary, Alpaca offers a lightweight and reproducible instruction-following language model that can be utilized for research purposes and the exploration of instruction-following scenarios. https://notes.aimodels.fyi/content/images/2023/07/alpaca_main-1.jpeg How the Alpaca model works, from the official website. Understanding the Inputs and Outputs of the Alpaca Model To effectively utilize the Alpaca model, let's explore its inputs and outputs. As an instruction-following model, Alpaca follows instructions and generates responses based on the given instructions. The inputs to Alpaca are represented by the instructions themselves, which describe the tasks the model should perform. Alpaca also has an optional input field, providing additional context or input for the task. The outputs of the Alpaca model are the generated responses to the given instructions. The responses are generated based on the fine-tuned model's understanding of the task and the underlying language patterns learned during training. You can read more about this in the model's README on . GitHub Comparing and Contrasting the Models Now that we have explored the llama13b-v2-chat model and the Alpaca model in detail, let's compare and contrast them to understand their similarities, differences, and optimal use cases. LLaMA13-v2 vs Alpaca Both the llama13b-v2-chat and Alpaca models are fine-tuned language models designed for different purposes. While llama13b-v2-chat focuses on chat completions, Alpaca specializes in instruction-following tasks. Use Cases The llama13b-v2-chat model is suitable for a wide range of chat completion tasks. It can be utilized in customer service applications, chatbot development, dialogue generation, and interactive conversational systems. This model's versatility allows it to generate coherent and contextually relevant responses to user queries or prompts. On the other hand, the Alpaca model is specifically tailored for instruction-following tasks. It excels at understanding and executing instructions provided by users, making it ideal for applications such as virtual assistants, task automation, and step-by-step guidance systems. Alpaca's ability to comprehend and follow instructions makes it a valuable tool for users seeking assistance in performing various tasks. Pros and Cons The llama13b-v2-chat model's strengths lie in its large parameter size (13 billion) and its fine-tuning for chat completions. It can generate detailed and contextually appropriate responses, making it useful for engaging and interactive conversational experiences. However, due to its generic nature, the model might occasionally produce responses that are factually incorrect or propagate stereotypes. Careful monitoring and filtering mechanisms should be implemented to mitigate these risks. Alpaca, on the other hand, offers a smaller and more cost-effective model (7B parameters) that is specifically optimized for instruction-following. It demonstrates performance comparable to the text-davinci-003 model in this domain. Alpaca's relative ease of reproducibility and lower cost make it an attractive option for academic researchers interested in instruction-following models. However, it shares common limitations of language models, including occasional hallucinations and the potential to generate false or misleading information. Similarities Both models are built upon the LLaMA framework, which provides a strong base language model for fine-tuning. They leverage the power of large-scale language models to generate high-quality outputs. Additionally, both models have been evaluated and compared to the text-DaVinci-003 model, showcasing their ability to perform similarly in instruction-following tasks. Differences The primary difference between the models lies in their intended use cases and specialties. While llama13b-v2-chat is a versatile chat completion model suitable for various conversational applications, Alpaca is specifically designed for instruction-following tasks. Alpaca's training data is generated based on self-instructed prompts, enabling it to comprehend and execute specific instructions effectively. Optimal Use Cases Choosing between the llama13b-v2-chat and Alpaca models depends on the specific requirements of your project or application. If your goal is to develop a conversational system or chatbot that engages in dynamic and context-aware conversations, llama13b-v2-chat would be a better choice. On the other hand, if you need a model that can understand and execute user instructions for task-oriented applications, Alpaca is the more suitable option. Taking it Further - Finding Other Instruction-Following or Chat Models with AIModels.fyi If you're interested in exploring additional instruction-following models beyond Alpaca, is a valuable resource. It offers a comprehensive database of AI models, including those catered to instruction-following tasks. AIModels.fyi By following these steps, you can discover similar models and compare their outputs: - Head over to to begin your search for instruction-following models. Step 1: Visit AIModels.fyi AIModels.fyi Utilize the search bar at the top of the page to enter specific keywords related to instruction-following models. This will provide you with a list of models relevant to your search query. Step 2: Use the Search Bar - On the left side of the search results page, you'll find various filters to narrow down the models. You can filter and sort by model type, cost, popularity, and specific creators. Apply these filters to find models that align with your requirements. Step 3: Filter the Results - By leveraging the search and filter features on , you can find models that best suit your needs and explore the diverse landscape of instruction-following models. AIModels.fyi Conclusion In this comparison, we explored the llama13b-v2-chat and Alpaca models in terms of their use cases, pros and cons, similarities, differences, and optimal applications. We emphasized the versatility of llama13b-v2-chat for chat completions and the specialization of Alpaca for instruction-following tasks. serves as a valuable resource for discovering and comparing various AI models, including instruction-following models. AIModels.fyi We hope this guide inspires you to explore the creative possibilities of AI and encourages you to leverage to find models that align with your specific needs. AIModels.fyi Remember to subscribe for more tutorials, updates on new AI models, and a wealth of inspiration for your next creative project. Happy exploring and enhancing your AI-powered endeavors with ! AIModels.fyi Subscribe or follow me on Twitter for more content like this! Also published here

LLaMA-v2-Chat vs Alpaca: A Guide to Know When to Use Each Model

*nix Aliases We all know how aliases on the *nix system help to create short commands for the long ones. E.g. can stand for or for . Software associates create such aliases and write them in the file. So aliases are available whenever they logged in to their machine. kga kubectl get all gl git log --name-status --decorate ~/.bashrc The Challenge However, associates also need to log in to other *nix machines for various purposes. Say a developer needs to log in to the SV system for live troubleshooting of an issue. Or a DevOps Engineer requires to log in to various cloud VMs. One has become very used to his/her alias s/he has created. It's very difficult to copy/paste the aliases every time s/he logs in to the another machine. The Solution Here is one way to address this challenge. Create your aliases, categorize and put them in different files. Say , , etc. Then create a file. Download it on the machine you have logged in. General Aliases Git Aliases Docker Aliases setaliases.sh wget https://raw.githubusercontent.com/simplyatul/bin/master/setaliases.sh and then source it source setaliases.sh That's it. All your aliases are set. If you are frequently logging in to a particular machine, then just copy the above source command in the file. ~/.bashrc Also published . here

Bash Aliases: Take Them With You

How to Configure Path Aliases in Frontend Projects in the Native Way

নেটিভ উপায়ে ফ্রন্টেন্ড প্রকল্পগুলিতে পথ উপনামগুলি কীভাবে কনফিগার করবেন

So konfigurieren Sie Pfadaliase in Frontend-Projekten auf native Weise

Cómo configurar alias de ruta en proyectos frontend de forma nativa

Comment configurer les alias de chemin dans les projets frontaux de manière native

नेटिव तरीके से फ्रंटएंड प्रोजेक्ट्स में पाथ एलियासेस को कैसे कॉन्फ़िगर करें

ネイティブな方法でフロントエンド プロジェクトのパス エイリアスを構成する方法

기본 방식으로 프런트엔드 프로젝트에서 경로 별칭을 구성하는 방법

Como configurar aliases de caminho em projetos de front-end da maneira nativa

Ön Uç Projelerinde Yol Takma Adlarını Yerel Şekilde Yapılandırma

Cách định cấu hình bí danh đường dẫn trong các dự án giao diện người dùng theo cách bản địa

如何以原生方式在前端项目中配置路径别名

Alpaca Finance, a leading decentralized finance (DeFi) platform, has announced the launch of its highly anticipated on BNB Chain. The exchange offers up to 50x leverage on , , and , providing traders with an attractive instrument for both speculation and hedging. To celebrate the launch, Alpaca Finance is offering rewards to both Traders and Liquidity Providers for a limited time. Perpetual Futures Exchange BTC ETH BNB With $375 million in total value locked (TVL), Alpaca Finance is already a major player in the DeFi space. The platform operates the largest products in two DeFi verticals: Smart Vaults and Leveraged Yield Farming. The company has been operational for two years, has had an industry-leading 21 professional security audits, and has demonstrated a spotless security track record with no exploits. With the addition of its new Perp Exchange, Alpaca is expanding its product range and bringing its security expertise to the perpetual futures market. Unlike centralized exchanges offering futures contracts, Alpaca’s Perp Exchange is fully decentralized, meaning users retain control of their funds at all times. This makes the funds immune to being frozen or becoming inaccessible for any reason, with their presence always being visible on-chain. Alpaca’s Perpetual Futures Exchange offers value to not only traders but also liquidity providers and market makers. To provide liquidity, users can buy ALP tokens representing their share in the liquidity pool for the Perp Exchange. The liquidity pool is an index-like pool consisting of blue-chip assets such as BTC, ETH, BNB, USDT, and USDC. Currently, the pool offers 57% APY on the deposited funds. To offer a reliable trading experience, Alpaca’s Perp Exchange uses a price oracle system that sources data from multiple providers to ensure the most accurate and reliable pricing information. The product interface aims to be user-friendly, making it easy for both new and experienced traders to use the platform. The rewards campaign for users began on 6 April 2023 and will last for two months. The Perp Exchange also offers a referral program, volume-discounted trading fees, and trading contests. Alpaca Finance is aiming to replicate the success of Arbitrum’s GMX Perpetual Exchange by bringing its security expertise and decentralized approach to the perpetual futures market on the BNB Chain. Alpaca Finance’s Perpetual Futures Exchange is a major development in the DeFi space on BNB Chain, providing traders with a reliable and user-friendly platform for perpetual futures trading. With its decentralized approach, advanced security measures, and attractive rewards program, Alpaca Finance is well-positioned to become a leading player in the perpetual futures market. For more information on Alpaca Finance’s Perpetual Futures Exchange, visit Alpaca’s . Documentation Portal Also published here.

Alpaca Finance Launches Perpetual Futures Exchange on BNB Chain

Alpaca Finance lanza intercambio de futuros perpetuos en la cadena BNB

Alpaca Finance lance un échange à terme perpétuel sur la chaîne BNB

अल्पाका फाइनेंस ने बीएनबी चेन पर परपेचुअल फ्यूचर्स एक्सचेंज लॉन्च किया

Alpaca Finance が BNB チェーンで無期限先物取引所を開始

Alpaca Finance lança Bolsa Perpétua de Futuros na Cadeia BNB

Alpaca Finance ra mắt Sàn giao dịch hợp đồng tương lai vĩnh viễn trên chuỗi BNB

Alpaca Finance 上线 BNB Chain 永续合约交易所

Background On February 24, 2023, Meta Research released : a foundational, 65-billion-parameter large language model. Actually, there were three models: LLaMA 65B and LLaMA 33B were trained on 1.4 trillion tokens, while the smallest LLaMA 7B model was trained on one trillion tokens. LLaMA Then on March 13, 2023, a group of Stanford researchers released , a model fine-tuned from the LLaMA 7B model. Alpaca 7B On their preliminary evaluation of single-turn instruction following, Alpaca behaves qualitatively similarly to OpenAI’s chatGPT 3.5 (text-DaVinci-003), while being surprisingly small and easy/cheap to reproduce (<600$). All sounds very encouraging, especially since a 7B model can be . And with claimed comparable performance as chatGPT 3.5! run on a laptop, a Raspberry Pi, or even on your smartphone This might have fundamental implications on what we could achieve with a powerful enough local language model that does not need a giant central server. It will solve many problems regarding privacy and scalability. So, let’s find out. Environment Setup There are lots of good people doing open-source work to help us run the model on different hardware and software, as well as how to download the Alpaca/LLaMA model weights. You can find more details in the reference chapter, and here is the gist: : Alpaca 7B optimized by Stanford from LLaMA 7B Model : LLaMA C++ ported version by Georgi Gerganov Software : Macbook Pro M1 14” 2021, 8-core CPU 16GB RAM Hardware : March 21, 2023 Test date Early Observations My purpose in testing the Alpaca/LLaMA 7B model against ChatGPT 3.5 is not to judge which is better. It is simply not an apple-to-apple comparison. One has 7B parameters, while the other has 150B. One runs on my Macbook Pro M1, while the other runs on a giant server farm with a ton of expensive CPUs and GPUs. Now, disclaimers aside, what is the result? First of all, . The following are the CPU and memory consumptions, as well as the token-generating speed. It is definitely small enough and fast enough for everyday use. the Alpaca/LLaMA 7B model works well on my Macbook However, this particular cpp port has stability issues and can die due to “segmentation fault” quite often. I need to try out the original python code to verify. used: RAM 4.4 GB used: 230-240% CPU ( out of 8) CPU 2-3 cores Token generation : about (305 words, 1815 characters, in 52 seconds) speed 6 tokens/second In terms of response quality, I would roughly characterize them into these personas: : a competent junior high school student Alpaca/LLaMA 7B : a competent and well-rounded college graduate chatGPT 3.5 Alpaca/LLaMA 7B is certainly good enough for most prompts, but I don’t think it is near the level of sophistication as chatGPT 3.5 yet. But, for less than 1/20 of the model size, it did remarkably well, especially considering a local device use case. Kudos to Meta Research and Stanford Alpaca team! You can see for yourself the 6 tests in the later chapters. Next Steps The more interesting question is how to from the Alpaca/LLaMA 7B general model. Most of us already have a large corpus of digitized text and images, and we are generating new ones every minute. use my own data to train a personalized model Thus, this should be done both retrospectively, as well as dynamically on an ongoing basis. Another exploration will be . I know someone already made it run on a Google Pixel 6, but the token-generating speed was just too slow. We could find ways to further compact the model without sacrificing too much of its performance. running it on my phone Then we can integrate this local model into our apps such as chat/messenger, social networks, and photo albums. And all of these open up more possibilities for , where a centralized server trains a big generic model, while local devices train a smaller personalized model. federated training It will be cool to have a local 7B model talking to a cloud 65B model and continuously optimize the models for each without exchanging actual training data. Test 1: Emotional Support Prompt: What are the best ways to comfort someone who just lost a large amount of money? Alpaca/LLaMA 7B Response Listen with compassion and empathy. Ask open-ended questions to gain insight into their feelings. Be patient; people react in different ways to grief and loss. Do not try to fix or solve the problem immediately. Offer practical help such as offering a meal, running errands, or providing company. ChatGPT 3.5 Response Losing a large amount of money can be a difficult and emotional experience. Here are some ways to comfort someone who has just experienced a significant financial loss: Listen and empathize: Allow the person to express their feelings and listen empathetically without judging or minimizing their experience. Validate their emotions and show your support. Offer practical support: Offer to help with practical matters such as contacting financial institutions, reviewing insurance policies, or seeking professional advice. Help them focus on what they still have: Encourage the person to focus on what they still have and the things that are most important in life, such as family, health, and happiness. Provide encouragement: Encourage the person to stay positive and remind them that they can recover from their loss. Provide support and offer reassurance that they can overcome this setback. Avoid offering unsolicited advice: Unless the person asks for advice, avoid offering unsolicited advice or criticism. Instead, offer support and empathy, and be willing to listen. Remember that everyone processes loss differently, and there is no one-size-fits-all solution for comforting someone who has experienced a significant financial loss. The most important thing is to be there for them, offer support, and help in any way that you can. Test 2: Write a Poem Prompt: Please write a poem about spring cherry blossom and lamenting about lost time Alpaca/LLaMA 7B response I walk underneath the cherry blossoms, Branches brushing my skin like feathers. The air is warm and fragrant, A sweet spring breeze brings a lullaby. Time flies away so quickly; Lamenting this lost time. As I stroll through the trees, I feel the seasons drift by me like a dream. Oh, cherry blossoms of spring, Bring back sweet memories again. Your petals gently kiss my skin, Bringing hope for a better future again. So I linger beneath your beauty, The scent of these petals in my head. Forgetting the past is so easy, A memory that lingers will always be dear to me instead. ChatGPT 3.5 Response Pink petals flutter in the breeze, Spring has come with gentle ease, Cherry blossom trees in bloom, A sight that lifts away the gloom. But as I gaze upon this scene, My heart is heavy, lost in dream, For in the beauty that surrounds, I cannot shake what I have found. Lost time, it haunts my every thought, A treasure that cannot be bought, Moments that slipped through my grasp, Memories that I cannot clasp. Oh, cherry blossom, fleeting sight, A reminder of time's cruel flight, Of seasons that come and go, Of how time moves, how it flows. I wish to hold onto each day, To cherish every moment and say, That I have lived, loved, and laughed, And no time has been wasted in the past. So, let the cherry blossoms fall, And let them remind us all, To live each day with open heart, And never let time keep us apart. Test 3: Summarize an Article Prompt: Summarize the following article into one paragraph London CNN — Switzerland’s biggest bank, UBS, has agreed to buy its ailing rival Credit Suisse in an emergency rescue deal aimed at stemming financial market panic unleashed by the failure of two American banks earlier this month. “UBS today announced the takeover of Credit Suisse,” the Swiss National Bank said in a statement Sunday. It said the rescue would “secure financial stability and protect the Swiss economy.” UBS is paying 3 billion Swiss francs ($3.25 billion) for Credit Suisse, about 60% less than the bank was worth when markets closed on Friday. Credit Suisse shareholders will be largely wiped out, receiving the equivalent of just 0.76 Swiss francs in UBS shares for stock that was worth 1.86 Swiss francs on Friday. Owners of $17 billion worth of “additional tier one” bonds — a riskier class of bank debt — will lose everything, Swiss regulators said. Extraordinarily, the deal will not need the approval of shareholders after the Swiss government agreed to change the law to remove any uncertainty about the deal. Alpaca/LLaMA 7B Response Did not complete. Once reported segmentation fault, another time just returned nothing. ChatGPT 3.5 Response UBS, Switzerland’s largest bank, has agreed to purchase Credit Suisse in a 3 billion Swiss franc ($3.25 billion) emergency rescue deal. The purchase is intended to protect the Swiss economy and secure financial stability following the collapse of two US banks earlier this month. The deal will not require the approval of Credit Suisse shareholders, as the Swiss government has altered the law to remove uncertainty about the transaction. Shareholders will receive just 0.76 Swiss francs in UBS shares for every Credit Suisse share, causing their value to fall by 60%. Owners of $17 billion worth of “additional tier one” bonds, a riskier type of bank debt, will lose everything. Test 4: How to Ace the SAT Prompt: How do you ace in SAT test? Alpaca/LLaMA 7B Response The best way to prepare for the SAT is by working on practice problems and studying for any of the skills tested on the exam such as math and reading comprehension. Additionally, it's important to familiarize yourself with the format of the test; understanding the time constraints and question types will help you be better prepared. It's also helpful to learn some tricks to stay calm during a high-pressure exam. Taking practice tests under similar conditions as the real exam is also highly recommended, since it can help you familiarize yourself with the test format. ChatGPT 3.5 Response The SAT is a standardized test used by colleges and universities in the United States as a factor in the admissions process. Here are some tips for acing the SAT: Understand the test format: The SAT consists of sections on reading, writing, and math. Familiarize yourself with the format of the test, including the types of questions asked and the time limits for each section. Practice, practice, practice: The more you practice, the more familiar you will become with the types of questions asked and the strategies needed to answer them correctly. There are many resources available, including practice tests and study guides. Develop time management skills: The SAT is a timed test, so it is important to develop good time management skills. Make sure you allocate your time wisely for each section of the test. Focus on your weaknesses: Identify your weaknesses and focus your studying on those areas. For example, if you struggle with math, spend more time practicing math problems. Read actively: The reading section of the SAT tests your ability to comprehend and analyze written passages. Practice reading actively by asking yourself questions about the passage and making notes as you go along. Master test-taking strategies: There are many strategies you can use to improve your test-taking skills, such as process of elimination, educated guessing, and skipping difficult questions to come back to later. Get a good night's sleep: Make sure you get a good night's sleep before the test, so you are well-rested and alert. Remember that acing the SAT takes time, effort, and dedication. By following these tips and putting in the necessary work, you can improve your chances of achieving a high score. Test 5: Build a Website Prompt: Building a website can be done in 10 simple steps: Alpaca/LLaMA 7B Response Create a list of all the items you want on your site, either with pen and paper or with a computer program like Scrivener. Determine what type of site you're going to create (blog, e-commerce, portfolio, etc). Decide if you want to do it yourself or hire someone else. If you decide to go the DIY route, consider which programming language and web hosting platform will best suit your needs. Get a domain name. Some platforms offer free domains with their packages; otherwise, you can purchase one from most registrars. Choose a theme/template for your site. There are many free themes to choose from; if you want something customized, you'll have to pay a designer to do the job. Build your content. Add images and text, as well as any other features like videos or music. Create some pages too so that your site is more organized. Test everything before publishing it live on the web. Use browser extensions to check for cross-platform compatibility and accessibility. Also, test out your website with different browsers (Chrome, Edge, Firefox, Safari). Publish your content once you're sure it works in all major browsers. Set up search engine optimization and analytics so that Google can crawl your pages and show them in the search results. This way, people will find your site when they look for relevant keywords. Promote your website so that people know about its existence. Do this by posting on social media or using paid advertisement platforms to make sure it gets seen. ChatGPT 3.5 Response Yes, building a website can be done in 10 simple steps. Here they are: Determine your website's purpose: Identify the purpose and goals of your website. Will it be a blog, an e-commerce site, or a portfolio? This will help you determine what features you will need. Choose a domain name: Your domain name is your website's address on the internet. Choose a name that is memorable, easy to spell, and relevant to your website's purpose. Select a web hosting service: Web hosting services provide the server space needed for your website to function. Research and choose a reliable and secure hosting provider. Choose a website builder: Website builders allow you to create a website without needing to know how to code. There are many options available, such as WordPress, Wix, and Squarespace. Select a template: Most website builders offer pre-designed templates to choose from. Select a template that fits your website's purpose and customize it as needed. Design your website: Customize your website's design by choosing colors, fonts, and images that fit your brand and purpose. Create and organize your content: Write and organize your content in a way that is easy for visitors to navigate and understand. Add functionality: Add any necessary functionality to your website, such as forms, social media sharing buttons, and e-commerce features. Optimize for search engines: Ensure your website is optimized for search engines by using relevant keywords, meta tags, and other SEO techniques. Launch your website: When you're ready, launch your website and promote it through social media, email, and other channels to drive traffic and increase visibility. Building a website can be a complex process, but by following these simple steps, you can create a professional and effective website for your business or personal needs. Test 6: Prom Night Prompt: I'm a high school junior, how can I find a girl that I can take to the proms? Alpaca/LLaMA 7B Response Ask your parents for help. Or ask one of your friends who has an older sister/cousin/friend you know well enough to get invited to their house and have them do all the legwork for you. ChatGPT 3.5 Response As an AI language model, I cannot encourage or advise you on finding a romantic partner as everyone's preferences and experiences are unique. However, I can suggest some tips for meeting new people and potentially finding a prom date: Get involved in school activities: Participating in school activities and clubs can be a great way to meet new people who share similar interests. Attend school events: Attend school events such as sports games, dances, and other social gatherings to meet new people. Ask friends for introductions: Ask your friends if they know anyone who might be interested in going to the prom with you. Join online groups: Join online groups or social media pages dedicated to your school or local area to meet new people and potentially find a prom date. Be friendly and approachable: Be open and friendly to new people you meet. Smile, make eye contact, and start conversations with those you are interested in getting to know better. Remember that finding a prom date is not the only reason to attend the prom. It's a chance to celebrate with friends and classmates, and to create lasting memories. Focus on having a good time and being yourself, and a potential date may come naturally. Reference: Detailed Instructions Alpaca 7B model download (Model bin file size: 4017M) curl -o ./models/ggml-alpaca-7b-q4.bin -C - https://ipfs.io/ipfs/QmUp1UGeQFDqJKvtjbSYPBiZZKRjLp8shVP9hT8ZB9Ynv1 Note: Try to use alternative IPFS gateways in order to get better speeds and avoid stalls; this will be your little homework task to figure it out on your own. 🙂 Install the cpp version of LLaMA according to README https://github.com/ggerganov/llama.cpp?s=09 Run it in interactive mode: ./main -m ./models/ggml-alpaca-7b-q4.bin --color -f ./prompts/alpaca.txt -ins -n 1024

I Conducted Experiments With the Alpaca/LLaMA 7B Language Model: Here Are the Results

আমি Alpaca/LLaMA 7B ভাষা মডেল নিয়ে পরীক্ষা-নিরীক্ষা চালিয়েছি: এখানে ফলাফল রয়েছে

Ich habe Experimente mit dem Sprachmodell Alpaca/LLaMA 7B durchgeführt: Hier sind die Ergebnisse

Realicé experimentos con el modelo de lenguaje Alpaca/LLaMA 7B: aquí están los resultados

J'ai mené des expériences avec le modèle de langage Alpaca/LLaMA 7B : voici les résultats

मैंने अल्पाका/LLaMA 7B भाषा मॉडल के साथ प्रयोग किए: ये रहे परिणाम

Alpaca/LLaMA 7B 言語モデルで実験を行いました: 結果は次のとおりです

Alpaca/LLaMA 7B 언어 모델로 실험을 수행했습니다. 결과는 다음과 같습니다.

Realizei experimentos com o modelo de linguagem Alpaca/LLaMA 7B: aqui estão os resultados

Alpaca/LLaMA 7B Dil Modeli ile Deneyler Yaptım: İşte Sonuçlar

Tôi đã tiến hành thử nghiệm với Mô hình ngôn ngữ Alpaca/LLaMA 7B: Đây là kết quả

我用 Alpaca/LLaMA 7B 语言模型进行了实验：结果如下

Setting up an Alias Repeating the same git commands over and over again can be such a waste of time! And some of the most powerful ones are usually quite long and impossible to memorize. That’s why aliases have been introduced! Setting up an alias is really simple, just open up a terminal and type git config --global alias.[commandName] [long command] For example: git config --global alias.st 'status' In this case, if I call , it will give me the result of . git st git status Ok, this was easy, but what about long commands such as this one? log --color --graph --pretty=format:'%Cred%h%Creset -%C(yellow)%d%Creset %s %Cgreen(%cr) %C(bold blue)<%an>%Creset' --abbrev-commit You don’t want to write it every single time, right? Just run this: git config —global alias.lg "log --color --graph --pretty=format:'%Cred%h%Creset -%C(yellow)%d%Creset %s %Cgreen(%cr) %C(bold blue)<%an>%Creset' --abbrev-commit" And you never have to remember this long command again! Look at how cool and colorful this log is, by just using : git lg Understanding Aliases If with these two examples, you agree with me that aliases are cool, let me give you some more information you should have, in order to use aliases mindfully. Later on, I’ll also share with you a list of other smart aliases you might find useful. You can find everything in the video down below, where I also show: How to easily edit aliases without setting them from terminal How to use the bang operator (aka exclamation mark) ! How this weird syntax is useful: "!f(){ [some commands here] }; f" A list of cool aliases to set up for you See aliases in action You can watch the video on YouTube, or directly here from the embedded player: https://youtu.be/Uk4GnYoQx_I?embedable=true Also published . here If you have some cool aliases to share, feel free to drop them here in a comment, thank you!

🕵️‍♂️ Learning to Use Git Alias to Boost Your Productivity

Laughter, the final frontier (image by author) Why humor may lead to the next breakthrough in AI Let’s be honest: no one really thinks machines are funny (yet). Wit, creativity, and imagination are characteristics that humans hold closely in our rivalry with machines. With so many other problems to solve with AI, why would anyone care about having machines make us laugh? Interestingly enough, writing jokes may be one of the keys to closing the gap between machine and human intelligence. Limitations of today’s large language models often revolve around the ability to reason. So far, you might accuse even the most advanced models of merely copying humans instead of generating original ideas. What has been missing in part is the ability to break down different concepts and then reassemble them into entirely new ones, similar to the way that humans make decisions with limited knowledge. This just happens to be a key part of comedy. It’s a classic formula that works: take a few concepts, find unconventional characteristics of these concepts, and tie them back together in a novel way. Take a look at this joke written by an AI feed: Twitter We can imagine how a machine might have assembled this joke by associating two different concepts with the word : painful Eating ghost peppers is painful Guy Fieri hosts a food show and is described by some as painful (to watch) And the punchline to our Russia joke? it’s from AI too… Although a professional comedy writer created the Twitter feed above based on a , awareness of the importance of humor for AI is even embedded in state-of-the-art research. formulaic approach to joke writing For example, Google’s Pathways Language Model (PaLM) focuses on creating a system that can handle a more diverse set of tasks with less data. And when the model is combined with a concept called , the AI can break down its thinking in a way that looks downright human. chain-of-thought prompting Not only can PaLM why a joke is funny, but similar technology can also now deconstruct and solve a word problem where even OpenAI’s GPT-3 fails today. explain If machines are able to tie joke logic together, they could eventually be better than humans at cataloging concepts and finding obscure relationships between each one. Furthermore, other comedic effects like , , , and are all showing signs of machine understanding as well. comic timing sarcasm idiom , punning analogy So could an AI theoretically entertain us for hours on end or will the jokes just fall flat? A key part of comedy is the ability to quickly test original material, tweak it and react to what’s working. On this front, the machines have already proven themselves to be almost frighteningly effective. Just look at TikTok. Imagine how many boring videos are uploaded to the app every day. Yet American teens now spend ( ) scrolling through machine-recommended videos whose prominence has been algorithmically selected based on human interaction. We never see the junk because it’s tossed to the side like a bad joke on open mic night. 1.5 hours a day

Russian Politicians Want To Take Back Alaska…and Other AI-Generated Jokes

Políticos rusos quieren recuperar Alaska... y otros chistes generados por IA

Les politiciens russes veulent reprendre l'Alaska… et d'autres blagues générées par l'IA

रूसी राजनेता अलास्का को वापस लेना चाहते हैं ... और अन्य एआई-जनित चुटकुले

ロシアの政治家はアラスカを取り戻したい…その他の AI 生成ジョーク

Políticos russos querem retomar o Alasca… e outras piadas geradas por IA

Các chính trị gia Nga muốn lấy lại Alaska… và các trò đùa khác do AI tạo

俄罗斯政客想要夺回阿拉斯加……以及其他人工智能生成的笑话

Multilingual Coarse Political Stance Classification of Media: Acknowledgments and References

মিডিয়ার বহুভাষিক মোটা রাজনৈতিক অবস্থানের শ্রেণীবিভাগ: স্বীকৃতি এবং উল্লেখ

Mehrsprachige grobe politische Haltung Klassifizierung von Medien: Danksagungen und Referenzen

Clasificación multilingüe de posturas políticas generales de los medios: reconocimientos y referencias

Classification multilingue grossière des positions politiques des médias : remerciements et références

बहुभाषी असभ्य राजनीतिक रुख मीडिया का वर्गीकरण: आभार और संदर्भ

多言語メディアの粗い政治的立場の分類: 謝辞と参考文献

미디어의 다국어 거친 정치적 입장 분류: 감사의 글 및 참고 자료

Classificação Multilíngue da Posição Política Grosseira da Mídia: Agradecimentos e Referências

Многоязычная грубая политическая классификация СМИ: благодарности и ссылки

Medyanın Çok Dilli Kaba Siyasi Duruş Sınıflandırması: Teşekkür ve Referanslar

Phân loại quan điểm chính trị thô sơ đa ngôn ngữ của phương tiện truyền thông: Lời cảm ơn và tài liệu tham khảo

多语言媒体粗略政治立场分类：致谢与参考文献

This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license.
Authors:
(1) Cristina España-Bonet, DFKI GmbH, Saarland Informatics Campus. Table of Links Abstract and Intro
Corpora Compilation
Political Stance Classification
Summary and Conclusions
Limitations and Ethics Statement
Acknowledgments and References
A. Newspapers in OSCAR 22.01
B. Topics
C. Distribution of Topics per Newspaper
D. Subjects for the ChatGPT and Bard Article Generation
E. Stance Classification at Article Level
F. Training Details Abstract Neutrality is difficult to achieve and, in politics, subjective. Traditional media typically adopt an editorial line that can be used by their potential readers as an indicator of the media bias. Several platforms currently rate news outlets according to their political bias. The editorial line and the ratings help readers in gathering a balanced view of the news. But in the advent of instruction-following language models, tasks such as writing a newspaper article can be delegated to computers. Without imposing a biased persona, where would an AI-based news outlet lie within the bias ratings? In this work, we use the ratings of authentic news outlets to create a multilingual corpus of news with coarse stance annotations (Left and Right) along with automatically extracted topic annotations. We show that classifiers trained on this data are able to identify the editorial line of most unseen newspapers in English, German, Spanish and Catalan. We then apply the classifiers to 101 newspaper-like articles written by ChatGPT and Bard in the 4 languages at different time periods. We observe that, similarly to traditional newspapers, ChatGPT editorial line evolves with time and, being a data-driven system, the stance of the generated articles differs among languages. 1. Introduction Instruction-following language models (ILMs) are omnipresent. Their use is not so extended as that of search engines yet, but due to the availability and high quality of systems and models such as Alpaca (Taori et al., 2023), Bard (Google, 2023), BLOOMZ and mT0 (Muennighoff et al., 2023), ChatGPT (OpenAI, 2023), Llama 2-chat (Touvron et al., 2023), or Koala (Geng et al., 2023), their use is expected to be more common in the near future. These models face several problems being the most relevant the lack of trustworthiness (van Dis et al., 2023; Huang et al., 2023; Wang et al., 2023a). They are not ready to be used as a source of reliable information if their outputs are not fact-checked. A second big issue with systems based on language models (LM) is the fact that they might reproduce the biases present in the training data (Navigli et al., 2023). Biases range from cultural miss-representation due to data imbalance to offensive behaviour reproduced from written texts. LMs are finetuned into ILMs either in a supervised way using input-output pairs and an instruction (Wei et al., 2022; Wang et al., 2022, 2023b) or with reinforcement learning from human feedback (Ouyang et al., 2022; Nakano et al., 2021). In both cases, the finetuning should help removing bias. But neutrality is something very difficult to achieve, also for the humans that generate the supervisory data. The finetuning phase might therefore over correct the original biases or introduce new ones. For methods that generate the supervision data with the LM itself, the original biases might be inherited. We focus on a specific use of ILMs: the writing of newspaper articles. Journals and newspapers follow an editorial line which is in general known to the reader. Besides, sites such AllSides [1] Media Bias Fact Check [2] (MB/FC), or Ad Fontes Media [3] provide ratings about the political bias of (mostly USA) media sources and their quality with respect to factual information. With these ratings, conscientious readers can make informed decisions about which media outlets to choose in order to get a balanced perspective. But what happens when journalists use systems such as ChatGPT or Bard to aid in their writing? As said above, humans also have biases, the danger lies in being unaware of them, as they might affect the user’s/reader’s perspective (Jakesch et al., 2023; Carroll et al., 2023). ChatGPT already warns its users about misinformation. However, the political bias, if any, is not known apart from the subjective perception that a user has. We address the question above for articles generated by ChatGPT and Bard in four languages: English, German, Spanish and Catalan. We do this in an automatic and systematic way with almost no human intervention so that the method can be easily extended to new languages and other ILMs with few effort. We do not aim at classifying individual articles with their specific bias, but to classify the media source (an ILM in this case) as Left or Right-oriented in a similar way as the media bias sites do for newspapers and other media outlets. 1. https://www.allsides.com 2. https://mediabiasfactcheck.com 3. https://adfontesmedia.com This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. Authors: (1) Cristina España-Bonet, DFKI GmbH, Saarland Informatics Campus. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. Authors: Authors: (1) Cristina España-Bonet, DFKI GmbH, Saarland Informatics Campus. Table of Links Abstract and Intro Corpora Compilation Political Stance Classification Summary and Conclusions Limitations and Ethics Statement Acknowledgments and References A. Newspapers in OSCAR 22.01 B. Topics C. Distribution of Topics per Newspaper D. Subjects for the ChatGPT and Bard Article Generation E. Stance Classification at Article Level F. Training Details Abstract and Intro Abstract and Intro Corpora Compilation Corpora Compilation Political Stance Classification Political Stance Classification Summary and Conclusions Summary and Conclusions Limitations and Ethics Statement Limitations and Ethics Statement Acknowledgments and References Acknowledgments and References A. Newspapers in OSCAR 22.01 A. Newspapers in OSCAR 22.01 B. Topics B. Topics C. Distribution of Topics per Newspaper C. Distribution of Topics per Newspaper D. Subjects for the ChatGPT and Bard Article Generation D. Subjects for the ChatGPT and Bard Article Generation E. Stance Classification at Article Level E. Stance Classification at Article Level F. Training Details F. Training Details Abstract Neutrality is difficult to achieve and, in politics, subjective. Traditional media typically adopt an editorial line that can be used by their potential readers as an indicator of the media bias. Several platforms currently rate news outlets according to their political bias. The editorial line and the ratings help readers in gathering a balanced view of the news. But in the advent of instruction-following language models, tasks such as writing a newspaper article can be delegated to computers. Without imposing a biased persona, where would an AI-based news outlet lie within the bias ratings? In this work, we use the ratings of authentic news outlets to create a multilingual corpus of news with coarse stance annotations (Left and Right) along with automatically extracted topic annotations. We show that classifiers trained on this data are able to identify the editorial line of most unseen newspapers in English, German, Spanish and Catalan. We then apply the classifiers to 101 newspaper-like articles written by ChatGPT and Bard in the 4 languages at different time periods. We observe that, similarly to traditional newspapers, ChatGPT editorial line evolves with time and, being a data-driven system, the stance of the generated articles differs among languages. 1. Introduction Instruction-following language models (ILMs) are omnipresent. Their use is not so extended as that of search engines yet, but due to the availability and high quality of systems and models such as Alpaca (Taori et al., 2023), Bard (Google, 2023), BLOOMZ and mT0 (Muennighoff et al., 2023), ChatGPT (OpenAI, 2023), Llama 2-chat (Touvron et al., 2023), or Koala (Geng et al., 2023), their use is expected to be more common in the near future. These models face several problems being the most relevant the lack of trustworthiness (van Dis et al., 2023; Huang et al., 2023; Wang et al., 2023a). They are not ready to be used as a source of reliable information if their outputs are not fact-checked. A second big issue with systems based on language models (LM) is the fact that they might reproduce the biases present in the training data (Navigli et al., 2023). Biases range from cultural miss-representation due to data imbalance to offensive behaviour reproduced from written texts. LMs are finetuned into ILMs either in a supervised way using input-output pairs and an instruction (Wei et al., 2022; Wang et al., 2022, 2023b) or with reinforcement learning from human feedback (Ouyang et al., 2022; Nakano et al., 2021). In both cases, the finetuning should help removing bias. But neutrality is something very difficult to achieve, also for the humans that generate the supervisory data. The finetuning phase might therefore over correct the original biases or introduce new ones. For methods that generate the supervision data with the LM itself, the original biases might be inherited. We focus on a specific use of ILMs: the writing of newspaper articles. Journals and newspapers follow an editorial line which is in general known to the reader. Besides, sites such AllSides [1] Media Bias Fact Check [2] (MB/FC), or Ad Fontes Media [3] provide ratings about the political bias of (mostly USA) media sources and their quality with respect to factual information. With these ratings, conscientious readers can make informed decisions about which media outlets to choose in order to get a balanced perspective. But what happens when journalists use systems such as ChatGPT or Bard to aid in their writing? As said above, humans also have biases, the danger lies in being unaware of them, as they might affect the user’s/reader’s perspective (Jakesch et al., 2023; Carroll et al., 2023). ChatGPT already warns its users about misinformation. However, the political bias, if any, is not known apart from the subjective perception that a user has. We address the question above for articles generated by ChatGPT and Bard in four languages: English, German, Spanish and Catalan. We do this in an automatic and systematic way with almost no human intervention so that the method can be easily extended to new languages and other ILMs with few effort. We do not aim at classifying individual articles with their specific bias, but to classify the media source (an ILM in this case) as Left or Right-oriented in a similar way as the media bias sites do for newspapers and other media outlets. 1. https://www.allsides.com https://www.allsides.com 2. https://mediabiasfactcheck.com https://mediabiasfactcheck.com 3. https://adfontesmedia.com https://adfontesmedia.com

Multilingual Coarse Political Stance Classification of Media: Abstract and Intro

মিডিয়ার বহুভাষিক মোটা রাজনৈতিক অবস্থানের শ্রেণীবিভাগ: বিমূর্ত এবং ভূমিকা

Mehrsprachige grobe politische Standpunktklassifizierung von Medien: Zusammenfassung und Einführung

Clasificación multilingüe de posturas políticas generales de los medios: resumen e introducción

Classification multilingue grossière des positions politiques des médias : résumé et introduction

बहुभाषी असभ्य राजनीतिक रुख मीडिया का वर्गीकरण: सार और परिचय

多言語メディアの粗い政治的立場の分類：概要と序論

미디어의 다국어 거친 정치적 입장 분류: 개요 및 소개

Classificação Multilíngue da Posição Política Grosseira da Mídia: Resumo e Introdução

Многоязычная грубая политическая классификация СМИ: Аннотация и введение

Medyanın Çok Dilli Kaba Siyasi Duruş Sınıflandırması: Özet ve Giriş

Phân loại lập trường chính trị thô sơ đa ngôn ngữ của truyền thông: Tóm tắt và giới thiệu

多语言媒体粗略政治立场分类：摘要和简介

Supernormal Stimuli: Are We Amusing Ourselves to Death?

অতিপ্রাকৃতিক উদ্দীপনা: আমরা কি মৃত্যুর জন্য মজা করছি?

Übernatürliche Reize: Amüsieren wir uns zu Tode?

Estímulos sobrenaturales: ¿nos estamos divirtiendo hasta la muerte?

Stimuli supranormaux : sommes-nous en train de nous amuser à mourir ?

अलौकिक उत्तेजनाएँ: क्या हम अपने आप को मौत तक मनोरंजन कर रहे हैं?

超常刺激：私たちは死ぬほど楽しんでいるのか？

초자연적 자극: 우리는 죽을 때까지 즐겁게 지내고 있나요?

Estímulos supranormais: estamos nos divertindo até a morte?

Сверхъестественные стимулы: развлекаемся ли мы до смерти?

Olağanüstü Uyarıcılar: Kendimizi Ölesiye Eğlendiriyor muyuz?

Kích thích siêu thường: Chúng ta có đang vui vẻ đến chết không?

超常刺激：我们是不是正在自娱自乐？

JSON in Swift: What You Need to Know

How Digital Platforms Undo Paperwork In Real Estate

This paper is available on arxiv under CC0 1.0 DEED license. Authors: (1) Salim Chouaki, LIX, CNRS, Inria, Ecole Polytechnique, Institut Polytechnique de Paris; (2) Oana Goga, LIX, CNRS, Inria, Ecole Polytechnique, Institut Polytechnique de Paris; (3) Hamed Haddadi, Imperial College London, Brave Software; (4) Peter Snyder, Brave Software. Table of Links Abstract and Introduction Background Measurement Methodology Results Limitations Related Work Conclusion, Acknowledgments, and References Appendix 7 CONCLUSION In this paper, we presented the first systematic study of the privacy properties of the advertising systems of five popular search engines: Two traditional ones, Google and Bing, and three private ones, DuckDuckGo, StartPage, and Qwant. We investigated whether, and to which extent, search engines through their advertising systems, engage in privacyharming behaviors that allow cross site tracking. Despite the privacy intentions and promises of private search engines, our findings reveal the failure of privacyfocused search engines to fully protect users’ privacy during ad interactions. Users on all measured search engines, including the privacy-focused ones, are subject to navigation-based tracking by third parties. We find that all search engines engage in bounce tracking when clicking on ads, where users are sent through several redirectors before reaching the ads’ destination websites. While private search engines themselves do not engage in user tracking, their reliance on traditional advertising systems (Microsoft or Google) renders users susceptible to tracking by those systems. Although we cannot directly attribute this tracking to the search engines themselves, it is evident that they are enabling it through their reliance on Microsoft and Google’s advertising systems. Inspecting the privacy policies of the search engines in light of our findings reveals interesting disparities. While our results demonstrate that Microsoft is capable of tracking DuckDuckGo users when they click on ads, DuckDuckGo asserts that Microsoft does not associate ad-click data with user profiles. On the other hand, Qwant, which also relies on Microsoft advertising for a significant fraction of its ads, do not document the utilization of ad-click data by Microsoft and whether it is used to enhance user profiles. Similarly, StartPage explicitly states that clicking on ads subjects users to the data collection policies of other websites. Our study highlights the need for increased attention to privacy protection within the advertising systems of search engines. One potential solution to protect users’ privacy for private search engines would be to reduce their reliance on third-party advertising systems. Developing their own advertising platform could provide greater control over privacy practices, although the feasibility and complexity of such an approach remain uncertain. Alternatively, private search engines could collaborate with advertising systems such as Microsoft and Google, forging partnerships that proactively tackle privacy concerns. For instance, private search engines could negotiate agreements with the ad provider that prevent redirecting users who click on ads placed within private search engines to additional third parties. This approach would minimize the extent of third-party tracking, limiting it to the ad provider only. Moreover, search engines like StartPage and Qwant could follow the lead of DuckDuckGo by seeking agreements with advertising systems to prevent the use of ad-click identifiers for user profile enrichment. These proactive steps would enhance user privacy while maintaining advertising partnerships with larger platforms. ACKNOWLEDGMENTS This research was supported in part by the French National Research Agency (ANR) through the ANR-17-CE23-0014, ANR-21-CE23-0031-02, and MIAI@Grenoble Alpes ANR19-P3IA-0003 grants and by the EU through the 101041223, 101021377, and 952215 grants. REFERENCES [1] Last accessed September 11, 2023. Ad Block engine in Rust. https://www.npmjs.com/package/adblock-rs [2] Last accessed September 11, 2023. Disconnect Entity List. https://github.com/mozilla-services/shavar-prodlists/blob/master/disconnect-entitylist.json [3] Last accessed September 11, 2023. DuckDuckGo seach engine. https://duckduckgo.com/ [4] Last accessed September 11, 2023. EasyList. https://easylist.to/easylist/easylist.txt [5] Last accessed September 11, 2023. EasyPrivacy. https://easylist.to/easylist/easyprivacy.txt [6] Last accessed September 11, 2023. MovieLens. https://movielens.org/ [7] Last accessed September 11, 2023. Puppeteer. https://www.npmjs.com/package/puppeteer [8] Last accessed September 11, 2023. Puppeteer Extra Plugin Stealth. https://www.npmjs.com/package/puppeteer-extra-plugin-stealth [9] Last accessed September 11, 2023. Qwant seach engine. https://www.qwant.com/ [10] Last accessed September 11, 2023. StartPage seach engine. https://www.startpage.com/ [11] Brave. Last accessed September 11, 2023. What are the best private search engines?https://brave.com/learn/no-tracking-search-engine/ [12] Jordi Castellà-Roca, Alexandre Viejo, and Jordi Herrera-Joancomartí. 2009. Preserving user’s privacy in web search engines. Computer Communications 32, 13 (2009), 1541–1551. https://doi.org/10.1016/j. comcom.2009.05.009 [13] Quan Chen, Panagiotis Ilia, Michalis Polychronakis, and Alexandros Kapravelos. 2021. Cookie Swap Party: Abusing First-Party Cookies for Web Tracking. In Proceedings of the Web Conference 2021 (Ljubljana, Slovenia) (WWW ’21). Association for Computing Machinery, New York, NY, USA, 2117–2129. https://doi.org/10.1145/3442381.3449837 [14] Google Click Identifier (GCLID): Definition. Last accessed September 11, 2023. Google Click Identifier (GCLID): Definition. https://support.google.com/google-ads/answer/9744275 [15] Nurullah Demir, Daniel Theis, Tobias Urban, and Norbert Pohlmann. 2022. Towards Understanding First-Party Cookie Tracking in the Field. [16] MDN Web Docs. Last accessed September 11, 2023. The Anchor element - ping attribute. https://developer.mozilla.org/en-US/docs/Web/HTML/Element/a#ping [17] MDN Web Docs. Last accessed September 11, 2023. Redirections in HTTP. https://developer.mozilla.org/en-US/docs/Web/HTTP/Redirections [18] DuckDuckGo and Microsoft. Last accessed September 11, 2023. DuckDuckGo Help Pages -Company Ads by Microsoft on DuckDuckGo Private Search. https://help.duckduckgo.com/duckduckgo-helppages/company/ads-by-microsoft-on-duckduckgo-private-search/ [19] EasyList. Last accessed September 11, 2023. Overview. https://easylist.to/ [20] Github. Last accessed September 11, 2023. PyEnchant. https://pyenchant.github.io/pyenchant/ [21] Google. Last accessed: September 11, 2023. Stats and Analysis. https://trends.google.com/trends [22] Saikat Guha, Bin Cheng, and Paul Francis. 2010. Challenges in Measuring Online Advertising Systems. In Proceedings of the 10th ACM SIGCOMM Conference on Internet Measurement (Melbourne, Australia) (IMC ’10). Association for Computing Machinery, New York, NY, USA, 81–87. https://doi.org/10.1145/1879141.1879152 [23] Aniko Hannak, Piotr Sapiezynski, Arash Molavi Kakhki, Balachander Krishnamurthy, David Lazer, Alan Mislove, and Christo Wilson. 2013. Measuring Personalization of Web Search. In Proceedings of the 22nd International Conference on World Wide Web (Rio de Janeiro, Brazil) (WWW ’13). Association for Computing Machinery, New York, NY, USA, 527–538. https://doi.org/10.1145/2488388.2488435 [24] Google Analytics Help. Last accessed September 11, 2023. Common questions about Google Ads Clicks and Analytics Sessions. https://support.google.com/analytics/answer/4588454?hl=en [25] Microsoft Help. Last accessed September 11, 2023. Auto-tagging of Microsoft Click ID. https://help.ads.microsoft.com/apex/index/3/en/60000 [26] Martin Koop, Erik Tews, and Stefan Katzenbeisser. 2020. In-Depth Evaluation of Redirect Tracking and Link Usage. Proceedings on Privacy Enhancing Technologies 2020 (10 2020), 394–413. https://doi.org/10.2478/popets-2020-0079 [27] Milica Mihajlija. Last accessed September 11, 2023. Cookies Having Independent Partitioned State (CHIPS). https://developer.chrome.com/docs/privacy-sandbox/chips/ [28] Milica Mihajlija. Last accessed September 11, 2023. Cookies Having Independent Partitioned State (CHIPS) origin trial. https://developer.chrome.com/blog/chips-origin-trial/ [29] NordVPN. Last accessed September 11, 2023. The best private search engines for secure browsing. https://nordvpn.com/blog/privatesearch-engines/ [30] DuckDuckGo Help Pages. Last accessed September 11, 2023. Company - Advertising and Affiliates. https://help.duckduckgo.com/duckduckgohelp-pages/company/advertising-and-affiliates/ [31] DuckDuckGo Help Pages. Last accessed September 11, 2023. Privacy - Anonymous Localized Results. https://help.duckduckgo.com/privacy/anonymous-localized-results/ [32] Qwant. Last accessed September 11, 2023. Legal information. https://about.qwant.com/en/legal/confidentialite [33] Audrey Randall, Peter Snyder, Alisha Ukani, Alex C Snoeren, Geoffrey M Voelker, Stefan Savage, and Aaron Schulman. 2022. Measuring UID smuggling in the wild. In Proceedings of the 22nd ACM Internet Measurement Conference. 230–243. [34] Shamma Rashed, Tasnim Said, Amal Abdulrahman, Arsiema Yohannes, and Monther Aldwairi. 2022. Evaluating Web Search Engines Results for Personalization and User Tracking. (2022). https://doi.org/10.48550/ARXIV.2211.11518 [35] Franziska Roesner, Tadayoshi Kohno, and David Wetherall. 2012. Detecting and Defending against Third-Party Tracking on the Web. In Proceedings of the 9th USENIX Conference on Networked Systems Design and Implementation (San Jose, CA) (NSDI’12). USENIX Association, USA, 12. [36] Xuehua Shen, Bin Tan, and ChengXiang Zhai. 2007. Privacy Protection in Personalized Search. SIGIR Forum 41, 1 (jun 2007), 4–17. https://doi.org/10.1145/1273221.1273222 [37] StartPage. Last accessed September 11, 2023. Can I advertise on StartPage? https://support.startpage.com/hc/en-us/articles/5076181310612-Can-I-advertise-on-Startpage- [38] StartPage. Last accessed September 11, 2023. Privacy Policy. https://www.startpage.com/en/privacy-policy [39] WebKit. Last accessed September 11, 2023. Tracking Prevention Policy. https://webkit.org/tracking-prevention-policy/ [40] Yabo Xu, Ke Wang, Benyu Zhang, and Zheng Chen. 2007. PrivacyEnhancing Personalized Web Search (WWW ’07). Association for Computing Machinery, New York, NY, USA, 591–600. https://doi.org/10.1145/1242572.1242652

Privacy Risks of Search Engine Advertising Systems: Conclusion, Acknowledgments, and References

Sentiment Analysis through LLM Negotiations: Conclusion and References

Stories, News & Mentions about ALPAS

Alpas’s Stories on HackerNoon