PERSUIT

persuit

Operation Weather.

113 Stories To Learn About Python3

279 Stories To Learn About Programming

373 Stories To Learn About Learning To Code

This article is really a precursor to cool things you can do with calculus such as the persuit curve which is used in air-to-air missiles, and rocket launch equations. Let's revisit some calculus topics you most likely haven't touched on in a while and use Python to take a refresher, and go over common derivatives and rules used. This is intended to be a calculus tutorial, but you won't use an online calculator such as Wolfram-Alpha, which requires an API. Also useful if you ever want to write your own code, or just do painless calculations. This is also a great way to check that your calculations done by hand are accurate. I will go over through differentiation rules in the easiest way possible, providing examples which you can execute using python. The areas covered in this how to install SymPy, and go through the following: Power Rule, Product Rule, Quotient Rule, Chain Rule, Exponential, Partial Derivatives I will use Lagrange's derivative notation (such as 𝑓(𝑥), 𝑓 (𝑥), and so on) to express formulae as it is the easiest notation to understand whilst you code along with python. Whilst it is more common to use the Leibniz  notation, (d/dx), it didn't feel natural when running differential equations in Python. ′ Using SymPy This post will use , which is a Python library. It is also possible to use the SciPy library, but SymPy prints the output in an easy to read way, and is more useful in getting a grasp of differentiation and integration. SymPy To install it, open up the terminal and run the following command: pip install sympy I would strongly recommend writing the python code in a new Jupyter Notebook, which comes with the distribution. Anaconda Differentiation Rules I will go through some differentiation rules first, as a quick refresher to some Calculus topics that you probably have forgotten a long time ago. Power Rule The power rule states: Which is self-explanatory if you have ever taken a calculus class. So let's take a simple example, and go through the steps by hand; This is a good, easy example to test out with the SymPy library. To implement this in python, first import the library, and declare a variable that you will use within your functions. The snippet below shows how to declare a single variable function: sympy sp

x = sp.Symbol( ) import as 'x' The final step is to get the derivation by running the code below: sp.diff(x** ) 3 Which outputs: You will notice that the output equation comes out in a nice print format. If you want to get the second order differential, , you can simply include twice within the command: f''(x) x diff sp.diff(x** ,x,x) 3 Where is the second order differential, or put in a hand-written way: 6x Product Rule Let's go through a quick example by hand: So running this in python will give you the following: sp.diff(sp.sin(x)*( *x** + )) 2 2 2 Quotient Rule The quotient rule helps allows us to efficiently find the derivative of one function divided by another function, such that: So, let's go through a quick example by hand: and this is what it looks like when we do this in : Python sp.diff((sp.sin(x))/x) Try this by hand, and then run it in python: The quotient rule is very similar to the product rule, except with changing the plus to a minus, and the extra step of dividing by the step. g(x)^2 Chain Rule As mentioned earlier, I have chosen to use Langrangian notation to go through these rules. The chain rule for some reason gets more over-complicated than it needs to be. The formula is shown below: So let's go through an example by hand: The inner function, g(𝑥) is 𝑥^2 +1 which when differentiated is 2𝑥. The outer function is f(𝑥) is ( )^7, which when differentiated turns to 7*( )^6. stuff stuff So following this logic, and here is the hand calculation: Which is pretty straghtforward when using the chain rule formula. This is what this looks like when run in python: sp.diff((x** + )** ) 2 1 7 And that is pretty much it. Try to calculate by hand the following examples if you want good practice: and, For more indepth explanation of the chain rule, check Aaron Schlegel's post on the . Chain Rule Exponential Derivatives and, using the chain rule, while giving an exponent of , e g(x) Let's go through a quick example: and the code for that will look like: sp.diff(sp.exp( *x)) 3 This covers all the differentiaiton rules you would have memorised in your calculus class. With all that, let's look at something called Partial Derivatives: What about Partial Derivatives? A partial differential equation (PDE) differs from the Ordinary Differential equations we have looked at so far, as PDEs contain multivariable functions. The difference being is that you take derivatives of one variable at a time. Let's start with a two variable function and find their partial derivatives: Now how do I do this in Python you might ask? The first step is to declare your variables 𝑥 and 𝑦 like so: x, y = sp.symbols( )

f = x** * y 'x y' 4 In the above, I have assigned the function as a variable, so as not to rewrite this variable every time I take a derivative. Taking the partial derivative in respect to x: sp.diff(f, x) And taking the partial derivative in respect to y: sp.diff(f, y) You can also take the derivatives with respect to many variables one after the other within the same line of code: sp.diff(f,x,y) Another example with 3 variable functions x, y, z = sp.symbols( )

f = x** * y * z** 'x y z' 3 2 The remaining part now is as before: sp.diff(f, x) sp.diff(f, y) sp.diff(f, z) And that's pretty much it on the derivatives side. However, these aren't actual functions useful in Python code, this is more of a homework thing. Actually putting this in Python Code! What is the point of calculating differential equations in my Python code if I can't plug in the numbers? SymPy gives us a function to do just that! This very handy function is called , where you substitute numbers where you have placed symbols. lambdify Let's do a quick example: f = *x** + *x
f_prime = sp.diff(f)
f_prime 2 3 4 So I have set up f_prime, but I want to substitute 𝑥 with the number 2. We now use our handy little function here: f_prime = sp.lambdify(x, f_prime)
f_prime( ) 2 >> 28 And there you have it. Let's do another example with multiple symbolic variables f = x** * y * z** F = sp.lambdify([x,y,z], f)

F( , , ) 3 2 1 2 3 >> 18 And with this, you now have a python library for finding derivatives and turning them into functions. Have fun! Future Articles Integrals deserve an article of their own, and will be the part 2, followed by Series Expansion and SymPy plots. After that, I will only do articles on the applications of calculus such as the persuit curve, rocket launch equations, orbital mechanics, or anything useful for Kerbal Space Program's kOS mod, which allows you to automate specific tasks. This is probably the easiest way to immediately apply calculus functions with Python. Referenced Websites These are the the websites which I have heavily referenced to make this revision cheat sheet. I would recommend visiting these sites, especially which goes through the same rules. Taking Derivatives In Python Taking Derivatives In Python SymPy.org Calculating Derivatives: Problems and Solutions Derivatives of Exponential Functions

How To Do Calculus with Python: Derivatives Cheat Sheet [Part 1] 

Things that I saw in person In one of the coffee shops where I worked, there was an employee who had been with the company for many years. She mastered all the opening and closing procedures of the store, knew how to prepare all the drinks and snacks on the menu, and had also worked in other franchised stores of the company. In this coffee shop, we worked in three shifts: opening, intermediate and closing. The coffee shop was in a shopping mall, opening before the stores and closing at 11 PM during the week and midnight on weekends. This employee had a son who was about 7 or 8 years old, and she was the only one responsible for him because the father had disappeared or something like that. She had a sister who, when necessary, would pick up the boy from school and take him home until she got off work. One day, she told me that her son’s teacher was very concerned because he was always tired in class, sleepy, and inattentive. The teacher asked her to pay attention to his sleep schedule. The problem was that the boy could only sleep when his mother got home. He stayed awake waiting to do his homework and spend some time with her. Although she had asked him not to wait for her, he said he needed help with his homework and couldn’t sleep until she arrived. She was very sad because she didn’t know what to do, especially since the following week she would have to work the night shift again. So, I suggested that we talk to the bosses, who were responsible for making the schedules, and proposed swapping shifts with her so that she would always have the morning shift and I the night shift. I lived just three minutes from the mall, and my wife also worked late, so it wouldn’t be a problem for me. In fact, I would be happy to help. We talked to the bosses, explained the situation, and requested the shift change. They said that, unfortunately, they couldn’t change it because the shifts had to be rotating to be “fair” to everyone. In another restaurant where I worked for about six months, we had three bosses who took turns in command. Each was very, very different from the other, and I remember one day in particular when one of the more irritable bosses was in charge of the store. Clearly, he wasn’t having a good day. From what I understood of the situation, the pasta machine had jammed, and according to him, it was because one of the cooks hadn’t put enough milk in the dough. He simply threw a box of milk in her face. I clearly remember seeing that woman, black and strong, all soaked in milk, with tears in her eyes. She didn’t quit because she couldn’t; she had a child to raise. There was another case where the restaurant owner was quite sadistic with the employees. When someone resigned and gave the one-month notice, he made that person’s life a living hell, assigning the worst tasks, at the worst times, and humiliating them in front of others in the most cruel way possible. And of course, there was that other incident where a boss pushed me because I was arranging the refrigerator with my shirt sleeves rolled up. He had previously warned me that I shouldn’t leave my tattoos visible to customers, even though the restaurant was still closed and there were no customers. Things that I saw online I’ve seen many reports on Medium, LinkedIn, and YouTube from people working in the tech sector, mainly, talking openly about how they feel purposeless in their remote, well-paid jobs with flexible hours, good work-life balance, company stock options, annual bonuses, and even office toys… Not to mention the free coffee and snacks. Just a small parenthesis: the coffee in the coffee shops where I worked wasn’t free for employees. These people have talked about how they want to work in a place that gives them more purpose in life, such as in a supermarket as shelf stockers. They really believe that these jobs are stress-free, but guess what: work is not stress-free. Let’s take the example of someone who quits everything to work in a supermarket. This was the case that recently appeared on my timeline. Work in the market is also divided and rotating shifts, to be “fair” to everyone. This means you won’t have time to organize yourself to take a course, practice a sport, or even pick up or drop off your kids at school at the same time, because you won’t have a fixed routine. Your weekly organization will need to be done according to the rotation of shifts, which means you can forget about medium-term commitments. Additionally, you rarely earn more than the minimum wage, have to deal with cleaning the supermarket (yes, it’s not just the cleaning team that cleans, everyone cleans), and you have to control what leaves the stock first and the order of expiration dates on the shelves. And all this while customers are on top of you, wanting to grab the items you just placed, often literally over your head, as if you didn’t exist, as if you were just an obstacle to the product. I find it very difficult to understand how this job, as dignified and necessary as it is, can bring purpose to someone who previously had all the freedom, money, and benefits I mentioned before. The intersection I believe people have this out-of-touch thinking because, for the most part, they have never worked in other areas. They don’t know that when the restaurant closes, the employees have to stay there for at least another hour cleaning and restocking everything for the next day. I currently work at a tech company as a software engineer, and most of my colleagues have never had another job besides IT. They often complain that they don’t have enough freedom, even though we don’t experience micromanagement, and we have paid days off and many other perks. Many believe they would be happier doing something “simpler”. And I think that’s where the problem lies: work should not be your main source of happiness. It’s not work that brings purpose to life. Work is there to support you, so that you can provide what you and your family need to live, so that you can make choices. Money may not bring happiness, but the lack of it to the point of not being able to put food on the table brings a lot of unhappiness, that I can guarantee. My thoughts I really like what the character Tina from the series The Bear says in episode 6 of the third season: “ I’ll clean a dish. I’ll wash a floor. I’ll sell some bullshit. I gotta cover this ass, right? Just look, I don’t need to be inspired. I don’t need to be impassioned. I don’t need to make magic. I don’t need to save the world, you know? Just - I just want to feed my kid, you know? ” https://www.youtube.com/watch?v=2wK4GVfSqeQ&embedable=true This doesn’t mean we should become a bunch of conformists who hate work. Quite the opposite! Most people will need to work for the rest of their lives, so if you’re unhappy and can make a change, do it (“unfortunately, most people can’t”)! I was miserable working in restaurants and I could change. I can guarantee it was the best decision of my life. Today, I have freedom, a job I enjoy doing, and the peace of knowing that I can provide everything my family needs. I didn’t change because I needed more purpose, I changed because I needed better living conditions. People with great positions, salaries, and benefits want to abandon their fields to do something with more meaning, with purpose, with impact on the world, maybe… Without realizing that the only purpose of work is to enrich the boss. In case it wasn’t clear, the purpose of work is to generate wealth for the company, not to create a life of happiness for you. Happiness and purpose change from person to person. For my former co-workers, the purpose was to take care of their children. For others, it may be helping their parents, spending more time with friends, traveling, playing an instrument, or sleeping 8 hours a night in a warm bed. Work wasn’t THE PURPOSE; it was a means to get there. So, don’t romanticize other people’s professions so much, and try to see your work with a bit more optimism. As NBA player Giannis Antetokounmpo responded to a reporter: “There’s no failure in sports. There are good days, bad days, some days you are able to be successful, some days you are not, some days it is your turn, some days it’s not. That’s what sports is about. You don’t always win.” https://www.youtube.com/watch?v=9mXGSjnUvSM&embedable=true And you can trust me when I say that this applies to far more than just sports. If you enjoy my content and want to support it, you can share and subscribe or you can buy me a coffee! Things that I saw in person In one of the coffee shops where I worked, there was an employee who had been with the company for many years. She mastered all the opening and closing procedures of the store, knew how to prepare all the drinks and snacks on the menu, and had also worked in other franchised stores of the company. In this coffee shop, we worked in three shifts: opening, intermediate and closing. The coffee shop was in a shopping mall, opening before the stores and closing at 11 PM during the week and midnight on weekends. This employee had a son who was about 7 or 8 years old, and she was the only one responsible for him because the father had disappeared or something like that. She had a sister who, when necessary, would pick up the boy from school and take him home until she got off work. One day, she told me that her son’s teacher was very concerned because he was always tired in class, sleepy, and inattentive. The teacher asked her to pay attention to his sleep schedule. The problem was that the boy could only sleep when his mother got home. He stayed awake waiting to do his homework and spend some time with her. Although she had asked him not to wait for her, he said he needed help with his homework and couldn’t sleep until she arrived. She was very sad because she didn’t know what to do, especially since the following week she would have to work the night shift again. So, I suggested that we talk to the bosses, who were responsible for making the schedules, and proposed swapping shifts with her so that she would always have the morning shift and I the night shift. I lived just three minutes from the mall, and my wife also worked late, so it wouldn’t be a problem for me. In fact, I would be happy to help. We talked to the bosses, explained the situation, and requested the shift change. They said that, unfortunately, they couldn’t change it because the shifts had to be rotating to be “fair” to everyone. In another restaurant where I worked for about six months, we had three bosses who took turns in command. Each was very, very different from the other, and I remember one day in particular when one of the more irritable bosses was in charge of the store. Clearly, he wasn’t having a good day. From what I understood of the situation, the pasta machine had jammed, and according to him, it was because one of the cooks hadn’t put enough milk in the dough. He simply threw a box of milk in her face. I clearly remember seeing that woman, black and strong, all soaked in milk, with tears in her eyes. She didn’t quit because she couldn’t; she had a child to raise. There was another case where the restaurant owner was quite sadistic with the employees. When someone resigned and gave the one-month notice, he made that person’s life a living hell, assigning the worst tasks, at the worst times, and humiliating them in front of others in the most cruel way possible. And of course, there was that other incident where a boss pushed me because I was arranging the refrigerator with my shirt sleeves rolled up. He had previously warned me that I shouldn’t leave my tattoos visible to customers, even though the restaurant was still closed and there were no customers. Things that I saw online I’ve seen many reports on Medium, LinkedIn , and YouTube from people working in the tech sector, mainly, talking openly about how they feel purposeless in their remote, well-paid jobs with flexible hours, good work-life balance, company stock options, annual bonuses, and even office toys… Not to mention the free coffee and snacks. Just a small parenthesis: the coffee in the coffee shops where I worked wasn’t free for employees. LinkedIn YouTube These people have talked about how they want to work in a place that gives them more purpose in life, such as in a supermarket as shelf stockers. They really believe that these jobs are stress-free, but guess what: work is not stress-free. Let’s take the example of someone who quits everything to work in a supermarket. This was the case that recently appeared on my timeline. Work in the market is also divided and rotating shifts, to be “fair” to everyone. This means you won’t have time to organize yourself to take a course, practice a sport, or even pick up or drop off your kids at school at the same time, because you won’t have a fixed routine. Your weekly organization will need to be done according to the rotation of shifts, which means you can forget about medium-term commitments. Additionally, you rarely earn more than the minimum wage, have to deal with cleaning the supermarket (yes, it’s not just the cleaning team that cleans, everyone cleans), and you have to control what leaves the stock first and the order of expiration dates on the shelves. And all this while customers are on top of you, wanting to grab the items you just placed, often literally over your head, as if you didn’t exist, as if you were just an obstacle to the product. I find it very difficult to understand how this job, as dignified and necessary as it is, can bring purpose to someone who previously had all the freedom, money, and benefits I mentioned before. The intersection I believe people have this out-of-touch thinking because, for the most part, they have never worked in other areas. They don’t know that when the restaurant closes, the employees have to stay there for at least another hour cleaning and restocking everything for the next day. I currently work at a tech company as a software engineer, and most of my colleagues have never had another job besides IT. They often complain that they don’t have enough freedom, even though we don’t experience micromanagement, and we have paid days off and many other perks. Many believe they would be happier doing something “simpler”. And I think that’s where the problem lies: work should not be your main source of happiness. It’s not work that brings purpose to life. Work is there to support you, so that you can provide what you and your family need to live, so that you can make choices. Money may not bring happiness, but the lack of it to the point of not being able to put food on the table brings a lot of unhappiness, that I can guarantee. My thoughts I really like what the character Tina from the series The Bear says in episode 6 of the third season: “ I’ll clean a dish. I’ll wash a floor. I’ll sell some bullshit. I gotta cover this ass, right? Just look, I don’t need to be inspired. I don’t need to be impassioned. I don’t need to make magic. I don’t need to save the world, you know? Just - I just want to feed my kid, you know? ” The Bear https://www.youtube.com/watch?v=2wK4GVfSqeQ&embedable=true https://www.youtube.com/watch?v=2wK4GVfSqeQ&embedable=true This doesn’t mean we should become a bunch of conformists who hate work. Quite the opposite! Most people will need to work for the rest of their lives, so if you’re unhappy and can make a change, do it (“unfortunately, most people can’t”)! I was miserable working in restaurants and I could change. I can guarantee it was the best decision of my life. Today, I have freedom, a job I enjoy doing, and the peace of knowing that I can provide everything my family needs. I didn’t change because I needed more purpose, I changed because I needed better living conditions. People with great positions, salaries, and benefits want to abandon their fields to do something with more meaning, with purpose, with impact on the world, maybe… Without realizing that the only purpose of work is to enrich the boss. In case it wasn’t clear, the purpose of work is to generate wealth for the company, not to create a life of happiness for you. Happiness and purpose change from person to person. For my former co-workers, the purpose was to take care of their children. For others, it may be helping their parents, spending more time with friends, traveling, playing an instrument, or sleeping 8 hours a night in a warm bed. Work wasn’t THE PURPOSE; it was a means to get there. So, don’t romanticize other people’s professions so much, and try to see your work with a bit more optimism. As NBA player Giannis Antetokounmpo responded to a reporter: “There’s no failure in sports. There are good days, bad days, some days you are able to be successful, some days you are not, some days it is your turn, some days it’s not. That’s what sports is about. You don’t always win.” https://www.youtube.com/watch?v=9mXGSjnUvSM&embedable=true https://www.youtube.com/watch?v=9mXGSjnUvSM&embedable=true And you can trust me when I say that this applies to far more than just sports. If you enjoy my content and want to support it, you can share and subscribe or you can buy me a coffee! If you enjoy my content and want to support it, you can share and subscribe or you can buy me a coffee ! buy me a coffee buy me a coffee

The Pursuit of Purpose: Why Are So Many People Quitting Their Jobs?

Disclaimer: The author works at Microsoft. The article is a general exploration of user trust and AI, but it does call out some references to the author’s company’s products, such as Copilot/Azure Open AI. “I knew I had just seen the most important advance in technology since the graphical user interface.” — Bill Gates, about ChatGPT acing an AP-level biology test Generative AI’s surge in popularity has opened up new and innovative possibilities, transforming the way we live and work. According to a study conducted by PwC, AI could contribute up to $15.7 trillion to the global economy in 2030. A Forbes Advisor report shows that a staggering 97 percent of business owners believe ChatGPT will help their business. These are exciting times for new startups specializing in Generative AI. However, there is a dichotomy in how end users feel about the Generative AI revolution. For instance, 46 percent of Americans are equally concerned and excited about AI (Artificial Intelligence). This type of user mistrust and skepticism has the potential to encumber AI adoption. What are some ways to mitigate inaccurate or misleading content generated by Large Language Models? How can companies build products that users not only love but trust? Hallucinations Large Language Models (LLMs) have been labeled stochastic parrots — a fitting metaphor for entities that regurgitate convincingly plausible answers without really understanding the meaning behind them. As Stephen Wolfram put it, a model like ChatGPT is simply adding one word at a time. LLMs are trained on massive amounts of data from various sources but aren’t by themselves doing any lookups or referring to external sources of information. This means that an LLM can make misleading claims or factual errors — while sounding fairly confident about the mistake. These counterfactual statements are, somewhat anthropomorphically, nicknamed hallucinations. To some extent, LLMs are meant to hallucinate. It’s Generative AI’s version of “it’s not a bug — it’s by design.” The same capability that makes LLMs hallucinate is also what makes them powerful. On the flip side, there are situations when hallucinations are not desirable and where it is critical to get truthful and factual responses. Working around the non-determinism of GenAI products continues to be a persistent challenge. However, there are ways to ensure that we can still maintain user trust while harnessing all the creativity and potential behind AI. Adding a Grounding Source of Truth One way to mitigate hallucinations is to augment the knowledge an LLM possesses with a grounding source of truth. Imagine you’re making someone take a test on a complex, previously unseen topic. This person is a fluent English speaker, but they know next to nothing about the topic they’re being tested on. Now, they could simply fabricate answers for all the questions. Or, you can give them an open textbook about the topic they’re being tested on so they have reference material from which they can answer. This example is a simplified explanation of what grounding represents in the context of LLMs. Grounding is table stakes while creating any kind of custom LLM application, especially for internal data (such as customer service chatbots that rely on non–public-facing knowledge bases) or applications that rely on a corpus to retrieve answers (a.k.a., search engines). As far as implementation goes, Retrieval Augmented Generation is probably the most ubiquitous pattern currently used for grounding LLMs. There are numerous out-of-the-box RAG solutions in the market that take in a custom corpus to do retrieval and generation. Azure Open AI on your own data is one such example, where users can upload a dataset or document from blob storage and “chat” with it. The feature also offers citations to the source text, along with the response. Making an LLM cite an authoritative source as the reason behind a particular answer automatically increases transparency and adds credibility. Grounding can also help with evaluation and validation of response quality, a key component to testing Generative AI applications before making them production-worthy for public release. In the absence of a grounding dataset, there is no way to “compare” the responses of an LLM to see if they are based on truth or not. So, choosing the right implementation of RAG and the right grounding dataset is critical for improving the factualness of the GenAI application. Increasing Explainability Surveys have shown that people think providing too little context on AI decisions can erode trust. Deep Learning models are trained to get accurate and fast predictions, but the end users are often clueless as to why the model generated a particular response. A famous example of this black box behavior occurs in my favorite work of fiction, Douglas Adams’ brilliant Hitchhiker’s Guide to Galaxy. In the book, the supercomputer Deep Thought is tasked with working out the Answer to The Ultimate Question of Life, the Universe, and Everything. After 7.5 million years, Deep Thought responds that the answer is 42. The scientists are puzzled, but Deep Thought is quite confident about the answer: “I think the problem, to be quite honest with you, is that you’ve never actually known what the question was.” In order to then calculate the Ultimate Question, a special computer is built, and hilarities ensue — but I won’t spoil the book any further. Decades after the book was written, we now have ways to peek behind the curtain and understand why an AI-generated a particular response. While traditional Machine Learning (ML) models use techniques like LIME and SHAP for explainability, Chain-of-Thought (CoT) prompting is a popular technique to add more transparency to Generative AI’s reasoning process. In essence, CoT consists of meta-prompts asking the AI to explain itself or interpret why it arrived at a particular output. In fact, researchers have noted that just adding a phrase like “Let’s think step by step” also increases the accuracy of generated responses, even in zero-shot scenarios. Employing Guardrails Apart from grounding, there are also various other techniques to reduce hallucinations, including advanced prompt engineering with CoVE (chain of thought verification), feedback loops with iterative improvements (or “humans in the loop”), using a combination of fine-tuned models with RAG, asking an LLM to self-critique its own output, as well as playing around with the “creativity” and “temperature” of LLM APIs. However, none of these methods are infallible since an LLM is essentially non-deterministic and unpredictable. There could be grave ethical as well as practical implications from Artificial Intelligence making decisions without any oversight. Incorrect predictions could lead to potentially life-threatening implications in fields like medical sciences, pharmacology, autonomous vehicle design, law enforcement, or disaster management. In the end, the safest option is for humans to have the final say in accepting or rejecting any AI-driven predictions. Having the end user in charge of vetting AI decisions also serves as a great feedback loop for developers to learn from and iterate continuously. Efforts to reduce hallucinations by improving factualness and transparency will increase user confidence in AI products. In addition, accountability is a key Responsible AI that must be at the center of all AI design. AI can be an amazing tool to revolutionize the way we all live and work — but ultimately, not to replace human judgment. Disclaimer: The author works at Microsoft. The article is a general exploration of user trust and AI, but it does call out some references to the author’s company’s products, such as Copilot/Azure Open AI. Disclaimer: The author works at Microsoft. The article is a general exploration of user trust and AI, but it does call out some references to the author’s company’s products, such as Copilot/Azure Open AI. Disclaimer: The author works at Microsoft. The article is a general exploration of user trust and AI, but it does call out some references to the author’s company’s products, such as Copilot/Azure Open AI. “I knew I had just seen the most important advance in technology since the graphical user interface.” — Bill Gates, about ChatGPT acing an AP-level biology test “I knew I had just seen the most important advance in technology since the graphical user interface.” — Bill Gates, about ChatGPT acing an AP-level biology test “I knew I had just seen the most important advance in technology since the graphical user interface.” Bill Gates, about ChatGPT acing an AP-level biology test Bill Gates, about ChatGPT acing an AP-level biology test Generative AI’s surge in popularity has opened up new and innovative possibilities, transforming the way we live and work. According to a study conducted by PwC, AI could contribute up to $15.7 trillion to the global economy in 2030 . A Forbes Advisor report shows that a staggering 97 percent of business owners believe ChatGPT will help their business . These are exciting times for new startups specializing in Generative AI . AI could contribute up to $15.7 trillion to the global economy in 2030 97 percent of business owners believe ChatGPT will help their business new startups specializing in Generative AI However, there is a dichotomy in how end users feel about the Generative AI revolution. For instance, 46 percent of Americans are equally concerned and excited about AI (Artificial Intelligence). This type of user mistrust and skepticism has the potential to encumber AI adoption. What are some ways to mitigate inaccurate or misleading content generated by Large Language Models? How can companies build products that users not only love but trust? 46 percent of Americans are equally concerned and excited about Hallucinations Hallucinations Large Language Models (LLMs) have been labeled stochastic parrots — a fitting metaphor for entities that regurgitate convincingly plausible answers without really understanding the meaning behind them. As Stephen Wolfram put it, a model like ChatGPT is simply adding one word at a time. stochastic parrots adding one word at a time. LLMs are trained on massive amounts of data from various sources but aren’t by themselves doing any lookups or referring to external sources of information. This means that an LLM can make misleading claims or factual errors — while sounding fairly confident about the mistake. These counterfactual statements are, somewhat anthropomorphically , nicknamed hallucinations . somewhat anthropomorphically hallucinations To some extent, LLMs are meant to hallucinate. It’s Generative AI’s version of “it’s not a bug — it’s by design.” The same capability that makes LLMs hallucinate is also what makes them powerful. On the flip side, there are situations when hallucinations are not desirable and where it is critical to get truthful and factual responses. meant Working around the non-determinism of GenAI products continues to be a persistent challenge. However, there are ways to ensure that we can still maintain user trust while harnessing all the creativity and potential behind AI. Adding a Grounding Source of Truth Adding a Grounding Source of Truth One way to mitigate hallucinations is to augment the knowledge an LLM possesses with a grounding source of truth. Imagine you’re making someone take a test on a complex, previously unseen topic. This person is a fluent English speaker, but they know next to nothing about the topic they’re being tested on. Now, they could simply fabricate answers for all the questions. Or, you can give them an open textbook about the topic they’re being tested on so they have reference material from which they can answer. This example is a simplified explanation of what grounding represents in the context of LLMs. Grounding is table stakes while creating any kind of custom LLM application, especially for internal data (such as customer service chatbots that rely on non–public-facing knowledge bases) or applications that rely on a corpus to retrieve answers (a.k.a., search engines). As far as implementation goes, Retrieval Augmented Generation is probably the most ubiquitous pattern currently used for grounding LLMs. There are numerous out-of-the-box RAG solutions in the market that take in a custom corpus to do retrieval and generation. Retrieval Augmented Generation Azure Open AI on your own data is one such example, where users can upload a dataset or document from blob storage and “chat” with it. The feature also offers citations to the source text, along with the response. Making an LLM cite an authoritative source as the reason behind a particular answer automatically increases transparency and adds credibility. Azure Open AI on your own data Grounding can also help with evaluation and validation of response quality, a key component to testing Generative AI applications before making them production-worthy for public release. In the absence of a grounding dataset, there is no way to “compare” the responses of an LLM to see if they are based on truth or not. So, choosing the right implementation of RAG and the right grounding dataset is critical for improving the factualness of the GenAI application. Increasing Explainability Increasing Explainability Surveys have shown that people think providing too little context on AI decisions can erode trust . Deep Learning models are trained to get accurate and fast predictions, but the end users are often clueless as to why the model generated a particular response. providing too little context on AI decisions can erode trust why A famous example of this black box behavior occurs in my favorite work of fiction, Douglas Adams’ brilliant Hitchhiker’s Guide to Galaxy . Hitchhiker’s Guide to Galaxy In the book, the supercomputer Deep Thought is tasked with working out the Answer to The Ultimate Question of Life, the Universe, and Everything. After 7.5 million years, Deep Thought responds that the answer is 42. The scientists are puzzled, but Deep Thought is quite confident about the answer: “I think the problem, to be quite honest with you, is that you’ve never actually known what the question was.” “ I think the problem, to be quite honest with you, is that you’ve never actually known what the question was. ” I think the problem, to be quite honest with you, is that you’ve never actually known what the question was. question In order to then calculate the Ultimate Question, a special computer is built, and hilarities ensue — but I won’t spoil the book any further. Decades after the book was written, we now have ways to peek behind the curtain and understand why an AI-generated a particular response. While traditional Machine Learning (ML) models use techniques like LIME and SHAP for explainability, Chain-of-Thought (CoT) prompting is a popular technique to add more transparency to Generative AI’s reasoning process. In essence, CoT consists of meta-prompts asking the AI to explain itself or interpret why it arrived at a particular output. In fact, researchers have noted that just adding a phrase like “ Let’s think step by step ” also increases the accuracy of generated responses, even in zero-shot scenarios. Let’s think step by step Employing Guardrails Employing Guardrails Apart from grounding, there are also various other techniques to reduce hallucinations, including advanced prompt engineering with CoVE (chain of thought verification), feedback loops with iterative improvements (or “humans in the loop”), using a combination of fine-tuned models with RAG, asking an LLM to self-critique its own output, as well as playing around with the “creativity” and “temperature” of LLM APIs. However, none of these methods are infallible since an LLM is essentially non-deterministic and unpredictable. There could be grave ethical as well as practical implications from Artificial Intelligence making decisions without any oversight. Incorrect predictions could lead to potentially life-threatening implications in fields like medical sciences, pharmacology, autonomous vehicle design, law enforcement, or disaster management. In the end, the safest option is for humans to have the final say in accepting or rejecting any AI-driven predictions. Having the end user in charge of vetting AI decisions also serves as a great feedback loop for developers to learn from and iterate continuously. Efforts to reduce hallucinations by improving factualness and transparency will increase user confidence in AI products. In addition, accountability is a key Responsible AI that must be at the center of all AI design. AI can be an amazing tool to revolutionize the way we all live and work — but ultimately, not to replace human judgment.

Pursuit of Trustworthiness: Increasing User Trust in Generative AI Products

Introduction There are different ways to form an investment thesis in crypto. One is to reject the importance of fundamentals and subscribe to the view that token prices are driven mainly by quick-gain-100x-seeking behavior or "financial nihilism." Is all crypto simply a meme? Should your technically sophisticated, tier-1 VC-backed, and community-vetted utility/governance "meme" coin be valued more than mine? Which narrative will be picked up by the KOLs next and pumped up by the crowd? What are we aping into? These views and strategies have merit but don't work at scale and across market cycles. Disciplined investors and researchers must carefully examine the facts and start seeing the forest for the trees. Let's stick to the first principles and dive into the concept of re-staking. Is it another backed by token emission or a new sustainable economy? As the saying goes, the proof is in the pudding; let's look for it! yield-generating bubble Cosmos Interchain Security Eigen Layer has popularised the idea of restaking (renting ETH economic security); however, it was initially proposed and implemented within the Cosmos ecosystem (renting ATOM economic security). Chains started with the Cosmos SDK provide configurability and sovereignty. However, developers' main challenge is building their own operator set (100-150 validators based on Tendermint consensus), bootstrapping liquidity, and using a native token, which has the risk of inflation and security value loss. enables other Chains (Consumer Chains) to be secured by the Cosmos Hub's validator set and the ATOM token's total market cap. In other words, it allows Consumer Chains to rent the economic security from Cosmos Hub. Cosmos Interchain security With 248MM ATOM staked, Cosmos Hub provides 3.2Bn$ in Economic Security to Consumer Chains. This first version of CIS is called “Replicated Security.” We Can Identify the Within the CIS Model: Following Stakeholders Validators – Run Cosmos Hub nodes and additional nodes for Consumer Chains Stakers – Delegate Atom tokens to Validators Consumer Chains – Rent Security from Cosmos Hub In Q3 2023, Chorus One the actual economic data from the two Consumer Chains on Cosmos – Stride and Neutron and made the following findings: reviewed Cosmos Hub Rewards: Cosmos Hub minted 367M$ per year or 1M$ daily in ATOM tokens as rewards for Stakers. Out of which 37M$ per year went to Hub Validators. Revenue: Cosmos Hub generated 1.5k$ per day in revenue from transaction fees. Stride generated 1k$ per day, and Neutron 12$ per day for Cosmos Hub. Combined daily transaction fees (2.5k$) would need to be 400 times higher to offset daily ATOM token emission (1M$) Infrastructure Costs (Excluding DevOps Costs) It's important to note that Validators need to set up additional nodes to support Consumer Chains, i.e., the hardware cost scales linearly with the addition of new Consumer Chains: The monthly node cost per Cosmos chain or ICS is 600$ per month or 7.2k$ per year. Cosmos Hub Security infra cost – 180 validators x 7.2k$ = 1.3M$ per year. Single Consumer Chain infra cost is the same – 1.3M$ per year. With Stride generating 365k$ per year and 36.5k$ in commission going to validators, such revenue doesn't come close to offsetting the infra costs for validators (1.3M$ per year). 40% of Cosmos Hub validators are not profitable, and this percentage increases as more Consumer Chains are onboarded. The consumer Chain would need substantially more revenue to provide an to Atom stakers. For example, to generate an extra 2% APR on staked ATOM, ICS would need to distribute $65MM per year to the Cosmos Hub. additional yield Finally, the ICS model removes the utility from the Consumer Chain native token, as it can only be used as a governance token and cannot be used for staking or network operations. , while the original idea of Cosmos Interchain Security offered easier onboarding of new chains, it came with a trade-off of increased infrastructure costs, lack of utility for the Consumer Chain native tokens, slashing risk for Validators, and high cost of economic security and therefore high revenue requirements for Consumer Chains. In summary As a result, Cosmos Hub governance is unlikely to approve more projects to join CIS unless they are set to deliver meaningful revenue to Hub validators and stakers. Moreover, the Replicated Security model is being revised, and it’s proposed to use a more flexible version called “ ,” whereby Validators can decide whether to run the nodes for Consumer Chains and Consumer chains would use only the Economic Security of supporting validators. Partial Set Security Eigen Layer Like CIS, allows Ethereum node operators to opt in to validate the new software modules. Eigen Layer calls these "chains" – Actively Validated Services. Eigen Layer The key difference with the CIS model is that the operators' cost of AVS support is marginal compared to running the ETH node; adding the AVS node to the existing ETH node doesn’t increase server load. Therefore, AVSs only need to generate enough fees to provide additional yield on top of native ETH staking yield. There Are the Within the Eigen Layer Model: Following Stakeholders Operators – Run Ethereum Nodes and additional AVS modules Restakers – Delegate Eth to Operators, which in turn opt into AVSs AVSs – Rent Security from Ethereum Infrastructure Costs (Excluding DevOps Costs): It costs approximately 1k USD per year to run 1 x ETH node (32 ETH), and more ETH validator slots can be added without significantly increasing the server load. The additional cost of supporting AVS is marginal (close to zero) Eigen Layer Economic Security As seen in the CIS model, Consumer Chains need to rent the entire $3.2Bn$ of economic security from Cosmos Hub and are constrained by the number of Cosmos Hub Validators. Eigen Layer implements "elastic scaling of security" and allows AVSs to dynamically control payouts to ETH stakers. 25000 AVS nodes can provide the equivalent of Cosmos Hub 3.2bn$ economic security. Similar to CIS, it would cost 64MM$ annually for AVS to generate an extra 2% APR on staked ETH. However, the key difference is that AVS can start with a smaller operator set and run more sustainable economics. Therefore, while the Eigen Layer brings additional infra costs almost to zero, it introduces the complexity of dynamic operator set and operator payout management. Watchtower networks will play an instrumental role in this process. Other Have Been Currently Worked Out: Open Questions and Challenges AVS Selection Process Who will choose AVSs? Will it be Operators or users of restaking protocols through governance voting? Or both? Operators are motivated to implement internal risk control frameworks to choose more profitable AVSs and manage slashing risks. Restaking protocols aim to conduct similar due diligence on behalf of their users. How will these two groups match AVS strategies and dynamically support the on and off-boarding of AVSs in an automated way? Are there any potential compliance issues with Stakers or Operators supporting AVSs in certain jurisdictions? These are not trivial questions! Potential Operator Centralization As the incremental cost to support new AVSs for operators is marginal – it would be more profitable for operators to support all available AVSs (as long as slashing risk is adequate), leading to centralization. Similarly, if users choose to restake only with the major operators, it will also lead to centralization. Eigen Layer is developing incentives to promote diverse operator sets and solo staking for AVSs to resolve this. Slashing risk Stakers opt-in by granting the Eigen Layer smart contracts the ability to impose additional slashing conditions on their staked ETH. The core idea of the Eigen Layer is to use "pooled security," whereby the same staked ETH is used to support multiple AVSs simultaneously. However, if one of the AVSs has a bug that leads to slashing, it can affect the operations of other AVSs. Eigen Layer addresses this risk by ensuring the system has " "  and a veto committee can reverse the slashing resulting from non-malicious behaviors. training wheels In addition, AVSs can acquire "attributable security," which, unlike pooled security, is specific to each AVS and will only be slashable by one AVS. Business Model Eigen Layer provides more flexibility for AVSs when defining the utility of their native token. In a mode, AVSs can use the native token for staking and network operations and balance its use vs. ETH throughout the project's lifetime. dual staking Conclusion In summary, the comparison between Cosmos Interchain Security (CIS) and Eigen Layer highlights the evolution of economic security concepts within blockchain ecosystems. While CIS pioneered the concept, it revealed drawbacks such as increased infrastructure costs, a lack of utility for native tokens, and high revenue requirements for Consumer Chains. Eigen Layer represents a substantial improvement by introducing elastic security scaling and dual staking. However, this enhanced flexibility comes with added complexity in managing slashing, security scaling, and operator payouts. It's truly fascinating to observe the ongoing evolution of this space as the latest restaking ideas from Eigen Layer permeate into other ecosystems such as Bitcoin ( ), Cosmos ( ), Solana ( ), and . Babylon Ethos Composable Avalanche Thank you to for your valuable feedback and review. Kam Benbrik

Restaking Explored: The Pursuit of a Sustainable Crypto-Economic Model

The United States v Meta Platforms Court Filing October 24, 2023 is part of . You can jump to any part in this filing . This is part 48 of 100. HackerNoon’s Legal PDF Series here COUNT III: FALSE OR MISLEADING STATEMENTS BY META IN VIOLATION OF BUSINESS AND PROFESSIONS CODE SECTION 17500 (BY THE PEOPLE OF THE STATE OF CALIFORNIA) 865. The People of the State of California (California) reallege and incorporate herein by reference each of the allegations contained in the preceding paragraphs 1 through 859 as though fully alleged in this cause of action. 866. From a date unknown to California and continuing to the present, Meta has engaged in and continues to engage in acts or practices that constitute violations of California Business and Professions Code section 17500 et seq., by making or causing to be made untrue or misleading statements with the intent to induce members of the public to use Meta’s platforms when such statements were likely to mislead members of the public about the nature and safety of Meta’s platforms. Meta’s untrue or misleading representations include, but are not limited to, the representations described in paragraph 846. 867. At the time the untrue or misleading representations were made, Meta knew or by the exercise of reasonable care should have known that the representations were untrue or misleading. Continue Reading . Here About HackerNoon Legal PDF Series: We bring you the most important technical and insightful public domain court case filings. This court case 4:23-cv-05448 retrieved on October 25, 2023, from is part of the public domain. The court-created documents are works of the federal government, and under copyright law, are automatically placed in the public domain and may be shared without legal restriction. Washingtonpost.com

The Pursuit of Veracity: California Accuses Meta of Touting Falsehoods

Big tech is charging ahead in LLM development. Around a decade after virtual assistants like Siri and Alexa were introduced, a new wave of AI helpers with more autonomy is raising the stakes. Experimental systems that run on or similar models are attracting billions of dollars of investment as Silicon Valley competes to capitalize on the advances in AI. GPT-4 cloud computing division aimed at helping other companies develop their own chatbots and image-generation services backed by AI. and are also considering selling the underlying technology to other companies through their cloud operations. Copilot, which integrates generative AI in Windows 11 applications, including Photos, Snipping Tool, and Paint with generative AI capabilities such as text-to-image. Amazon’s released a suite of technologies Microsoft Alphabet Microsoft has now launched is quietly working on AI tools to challenge Microsoft-backed OpenAI Inc., Google, and others. The company built its own framework to create large language models and its own chatbot service, which some engineers are calling “ .” Apple Apple GPT Also, it seems is going to release a . In July, the tech giant brought out , which does run text-to-image and image-to-text generation. In May, it also unveiled , a first-of-its-kind AI model that can bind data from six modalities (images and video, audio, text, depth, thermal, and inertial measurement units (IMUs) at the same time, without any explicit supervision. Meta multimodal version of Llama shortly CM3leon ImageBind The speed at which big tech is pursuing success in LLMs makes you feel like calling out ‘steady now.’ But no one wants to be left behind, so no one will listen Meanwhile, OpenAI finally launched which integrates image features into the by integrating Dall-E 3 with ChatGPT Plus and ChatGPT Enterprise. This will usher in many new image-based applications for GPT-4, for example, generating text to match images. multimodal ChatGPT, chatbot This is coming just before Google is launching its Gemini, which is also expected to have multimodal functionality. launched a pretty effective too. LinkedIn AI growth loop In India, in August, Tech Mahindra , which the company is using to create its own advanced language model to improve communication across many Indian languages, starting with Hindi. In the first phase, it aims to cover 40 Hindi dialects (Kinnauri, Kangri, Chambeli, Garhwali, Kumaoni, Jaunsari, and more), and then go on to other Indian languages and dialects, catering to 25% of the world’s population. launched Project Indus https://twitter.com/OpenAI/status/1706280618429141022?embedable=true co-founder Nandan Nilekani-backed is also onto similar development, which is working on creating open-source datasets, tools, models, and applications for Indian languages. Last year, the Indian government also brought out Project Bhashini, in partnership with Microsoft, establishing language datasets and AI applications with crowdsourcing initiatives such as Bhasha Daan and others. Infosys AI4Bharath The race for the best LLM is heating up. The speed at which big tech is pursuing success in LLMs makes you feel like calling out ‘steady now.’ But no one wants to be left behind, so no one will listen. When the likes of to speak freely about the technology’s dangers, after he realized that computers could surpass people in smartness way earlier than he and other experts had expected, it feels like a ‘steady now’ moment. Geoffrey Hinton, an AI pioneer says he quit Google AI models are getting so good that, according to the , they have more than in “listening” to keystrokes to identify typed content. “Typing in a password while on Zoom call could spell disaster,” it says. 2 Guardian 90% accuracy But then, none of these big companies reached where they are by listening to the ‘steady nows’. Microsoft Founder Bill Gates said in July, “ .” “Today’s and tomorrow’s AIs might be unprecedented—but nearly every major innovation in the past has also introduced novel threats that had to be considered and controlled. If we move fast, we can do it again. If we manage the risks of AI, we can help ensure that they’re outweighed by the rewards (of which I believe ),” he says. This makes sense too. The risks of AI are real but manageable there are many This article was originally published by on Navanwita Bora Sachdev The Tech Panda.

Stories, News & Mentions about PERSUIT

Persuit’s Stories on HackerNoon