 Thank you. So just to start off, especially for the people, all the folks will not assume a lot of data science knowledge. In fact, it seems no data science knowledge. And though I see a bunch of you that I actually recognize, that you actually have a really good data science knowledge, we're going to answer that. That when I look at the collaboration, the chance that you're going to be able to create a bunch of soil online like data science is here, this is a big place for a lot of people to understand that. That's really what we're trying to do in this department. So I'm going to talk to you a little bit. The unit data science is, and I'm going to talk about how does that compare to these terms AI and machine learning? Because there's two parts that really show it's AI, but what does that really mean? How does that really compare to data science? And then in the afternoon, parts of the unit data science. We're going to explain how this kind of spin and give you a lot of a little bit. So I'm going to start with data science. So we're going to have a discussion about what data science is. We did my 67th introduction to data science class in June, and it's good to have a review because it's been Xamarin a little bit. So, I think that's not so bad. So the problem with data science is it's a new data science. The problem with data science is it's really this actionable insight. If you take data, if you come around and you don't find it, you don't have to jump into something that's large that can be reserved for the people that come later. So the problem is to generate actionable insight. To use data and have it in a way that the organization can bring it to that insight will have to give it to you. The disadvantage of just getting data or having information. If you have information that's useful, but you have to have a few of those synthesized things. You have knowledge that you can get out of trends, like sales are going up over time. So there's no other way to do it. The actionable insight, we'll talk about different examples of what actionable insight is. That's really valuable to find a sheet. Using data to create a science. And here you can see that data science is being a combination of three different disciplines. So we're going to start with data science. We're going to be in high school and we're going to do a lot of high school and junior science. So data science is programming. You don't need to have a knowledge. You're going to need to talk to people that really know. You don't talk to people that really know. You might throw a certain source of insight that is fully irrelevant to the actual prospect of development. And then finally, we can do some math. Data science is not going to be a high grade. Even the other level, how do we do this? You need to understand 80% accurately what the rat needs. So I thought I'd give you some examples of what I know just to kind of give some context to what it is. But the first one, I don't know if it's everyone, maybe I won't talk about this every day anymore. So how great it is that there's lots of different questions and other outbreaks as well. And there's a whole lot of data science to understand how that will spread. So how people will say it's important that it's important that it's critical that people are still studying it now. And we all know that it isn't just to say again, there's many people involved that might be helpful for some people. But it's what the projection is. How many people will get this disease next month. And that's actually much more useful because then you can do things in terms of accounting for the last bullet or whatever it is. You can try to have something useful for the person who doesn't get it. So five elements. Five elements is focused on understanding, for example, where they're going to. And again, we have to actually know it's not just well, this may not be related, it's basically one of these synergies in terms of what it should be. Which is true. But that's not an action, but action must be made in a period of time and what do this mean to the environment going forward? What do you mean by that? What do you mean by that? But that is a broader level of illusion and things like that. So we're probably going to need to find just data, which is it's warm, warm, warm. What does this mean going forward and turning that into actualness? And I want to understand what the representation is. So thinking about it from another perspective, we have a whole conversation about it this afternoon. So what is open data? Being able to look at the people and how might that be useful? So I won't talk about that a lot because we're going to have a whole discussion on that later. How that works, I think that's all about values and which ones are useful and which ones are not useful. The thing about different treatments and different diseases, there's been about symptoms of what disease you might have. And again, it's not about having any particular diagnosis about having all of this data and you need to have some sort of information to make a prediction. So that prediction might be or can go to your home and you have to keep it to be answered. That's actually what we say. Other ones, we're going to use so you're going to be going to be going to be at 3 o'clock or 3 o'clock for baseball games. So another one is to use all of sports. It's used pretty much across from sports to changes to playing with somebody's shoe runner. How do you evaluate people? That's actually not just in sports. You think about how to evaluate people to hold you over HR. It's about how to be used in kind of any field. In various analysts, it's probably the most broad use, broad area of focus. So this can be anything from predicting cells, understanding what will generate better cells, understanding what aspects of your product you're going to be able to tell what ways people are going to know about it or not. So having an accurate insight is definitely a good thing. You've all seen it in previous, so I'm sure you've already got that picture in your favorite streaming video. And all of these platforms including things like Tic Tac suggest what to do next. Right? And the best way to do it is you can choose different things in different platforms. So engage in an event and buy something. Engage in that place and just watch something. You can take that and also actually watch something or just a different type of watching. And you can also actually provide insight. Now if you take a step back and say is that actually an insight useful for you as a consumer, or a consumer, then you have to engage in a platform. So there's a great representation. And to some extent you prefer to be suggesting things that are interesting. Like if it's going to be a bad suggestion to do nothing, it's a platform at all and you won't like it. Now you have to find out if providing you something that's interesting, providing you something that's different. So I think that the challenge is a big model of what you're useful and what you're being used and you may not have a product in which case that's an example of that. But at a higher level this is actually the insight and it's not used, it's helpful. But I appreciate it because I actually have an effort to address something that's helpful for the person who's trying to figure out what I actually find. So this class anything? Like I said, it's not the other way. It's not the other way. It's not the other way, but it's probably the other way. So if everyone has the ability to collect data where many things happen and something can be used to collect data you can really try to figure out what actions you can generate. It's a class of four examples of what is the business problem, what are you trying to achieve. And the connection is this action always. And that, the invention is what data science is. But if I mentioned my goal here in a very short amount of time is to give you a higher understanding of data science, but also a machine learning in AI. So there's a lot of terms that people are similar to each other but they're not fully interchangeable. But you can see that there's a lot of data. And that is the part of what people think of, and that's what they're using interchangeably sometimes. But there's a lot of data on that but I will also try to explain what some of the differences are. So one of the key things that you can see is that the data science you can see is the end-to-end process of kind of collecting data and generating actionable data. The machine learning really connects to the AI you think it's that way. And that focuses on a very disparaging model. But if you have the data and you're trying to really focus on how you can build an algorithm or use an existing algorithm to generate this predation. So it's like a key part of data science. But the part of that we're not going to go hate-knowledge is not about machine learning, but hate-design. So that's one way to think about the difference. And then there's the AI which, obviously the AI isn't going to be used in the world at least. I mean, I'll talk about that. If the AI uses machine learning, it can use other contexts. So a algorithm that walks is definitely considered part of the AI. This machine learning because the way that we're about to learn is to walk is machine learning. It's springing out that if it doesn't lift up a leg, it's going to fall down. But it's not really part of the science because it's not one of the mechanisms that is actionable. It's really different. The vast majority of people will think of like a way about walks. They're self-driving cars. It's being very much AI, very much machine learning and not in real science, but not as much here. I kind of want to say a couple of words. But here you kind of see that the third thing we have here, so using the right algorithm is generated in your action. Well, it's idea of how you can create that information. But I think that machine learning is one of the amazing arguments that is similar to your actions to your intelligence. But if I want to actually work inside of your network, I think that that is human intelligence. I think that that is a very good model for the human intelligence, but not really about human intelligence. But here are seven differences. And one of the things about it is you think about machine learning as being free, as self-serve, or as AI. You think about it as part of that. But what I'm going to talk about is machine learning. So what kind of machine learning is called deep learning? So deep learning is a little bit of a thing, that means, for a long time. But it's only going to be used, if I'm going to say it, for about five years. And we've had a huge impact. So what I show on this track here is a speech recognition. So Alexa, we don't know if they did device. But how many systems were there? And it took a long time for them to get here. So for many, many years, you can see down here, the machine could get like 75% of the conversation. Which is pretty good, but not actually good enough. Here it is, about 95%. So what that means is I've been talking here about learning five percent of the time, but you've not been able to do anything that you've been doing most of it. And often, until deep learning happened, the machine would be able to listen to me, it got like 75%, whatever you say. That's not really useful in the tree. That's the way we introduced the field of speech recognition for the change, where over two or three years it equaled tunes, so machines using deep learning techniques got like 95%. And from 2018, I think that before we actually got better, now it's just, back in 2020, it was about 97%. So machines can understand now my conversation better than any of you. So that's key to learning. And I'll talk a little bit about what it is, but it's a type of machine learning, that's usually where my work's kind of in some sense trying to simulate how your brain works. Basically, learning and learning things you can learn about with deep learning. Deep learning is really useful. Deep learning is also really useful in self-driving cars. So here's some examples of where deep learning is used. In addition, I mentioned skin cancer protection, so that's also a deep learning. So it's a way that this makes us for a specialized machine learning algorithm, called deep learning. So deep learning is used in machine learning algorithms. It's specialized. It's more than simply using what's kind of new. There's no rest which is how your brain works. So I want to talk about real art. Real art, because that's how you do it. What are the others? How many people go for the Gini? Or like, yeah, don't worry about it. I did my hand right here. We're charging P2. Many people are going to charge P2. Actually, we are. So we've all heard of charging P2. So charging P2 is an implementation of general AI. There are other implementations as well. It's the one that people have been started by recently. People have used the most people in the world against. But the way it works is through general AI. So the algorithm that Gini uses in the Gini algorithm Gini algorithm is a deep learning algorithm. It's part of machine learning. So when you're using charging P2 you can think of it as a visual target but it's stated in the conversation. So it's like a person if you know. It's definitely a machine learning because that kind of knows the answer to it. And it's not using a general machine learning technique. It's using a new language that we can assert between general AI. General AI can use across many different areas. So that's generation. That's charging P2. In this generation we can actually use value. How many people have to change value? So in the center sense. I'll give you an example of that. It can actually be used for really a generation. It can be used for tasks. So it can be used in many different ways. I'm not going to get about where it is. So I just want you to understand that general AI is probably machine learning. It's probably the design and it's also part of AI. So we can be talking about what is the AI and we can be talking about the data science. We can talk about it in the area of time. It makes it very similar. So there's charging P2. So we have a model that uses a model that uses models are very large. These machine learning algorithms these neural networks, you want to think of it that way. There's the area of the work. They're so large they're only in companies with vast amounts of computer power to create them. So we can create, you know, in a laptop or in a smaller machine you can create a smaller model but you can create a smaller model like that. It kind of looks okay. But if you want to be as charging P2 does you need lots and lots of computer power. Like lots of it. And so we have companies that use hundreds of millions of dollars to make them that much. So that's why Google has it in Microsoft and many of our companies can work to invest lots of money to create these machine learning models. But in some ways they're just machine learning models that are machine learning models like that. And all of these are generally AI. Similarly, there's generally AI because the way the models work like in the language conversation part I'm going to say let's say that all of us, if it's with the next one should be in a conversation. So based on the entire internet that I always do just text data a lot of the internet data is text data. And especially with that we kind of realize that oftentimes we're going to be following the problems we're seeing in the entire world we're walking that forward that. So, what's that part of what we followed what we're doing in one concept follows the next concept. Because we need to ask chat, if you need a question they want to understand what you ask in the way that we should think we understand. What it does is it knows that question knows that set of tasks and it knows big how to do them what are the types of hits So it's not always any question person, but it's no question enough to maybe realize what an inertia is. But some people think of that as understanding the question, because it's very insightful or anything. But it certainly doesn't understand it like we think in conversation. So here's an example. So it says, some people like to picture some people. This was generated in a generative way. So this one actually, I don't think it was value, I think it was a different program. But value is what you can get from each other to be in the same company. There's several others. And what you do is you describe the picture here as creative. And it increases. It creates a lot of pictures as well. And the question is, is this art that person created with human instructions, using art that the value of every other model is created? Or is it actually owned by all the different pictures on the other hand that were used as input data to create this picture? But then who knows how good that is created is not an easy question to answer, actually. Some of these questions make a lot of perspective. And then this one might be a conceptual question. How do you think it moves this? So in the next few years, people will kind of face this a lot. But then actually, this picture will actually come in the artist that you will want in the world. And the question that's, is that okay? The competition was in part number 21. I don't think that's the question. It's actually, in general, a computer-augmented system is important to continue creating this question. I think people have faced this before. Almost everywhere. That's good. Finally, it's probably similar to which ones are real and which ones are not. So it's a five-period. So why is that not easy? There are things after this and lots of new ideas. If you look at the years, most of these are not real. If you watch these, or if you pick that most of them, if you watch them, you will get which ones are real and which ones are not. So the problem that you have, if I give you any more questions, the great problem is it's not going to have to find, do some interesting things. Have people maybe do things that they need. But the problem is, like, maybe some crews, let them want to say some of the things that deepfakes are having to say. No, definitely, you don't know which ones are real and which ones aren't. Just because some crews in the video sets use the real found crews, the better one is the real found crews. So in the video, you just view a deepfake found crews. You'll find a whole bunch of them. And then you can find them wrong by actually looking at movies, for example, to treat them. And you can say, like, they're going to be unrealistic. So you can do it completely, you can do it for politicians, you can do it for sports players. So you put ways in your mouth of your political opponent. Find a discreet with somebody who says it's not find a discreet with somebody who said they didn't even say it. But if you know why, just say it to them. But you need to understand what they are. So that's a little hint about why it's hard. But I've talked a little bit about the difference between them. But I don't know if that felt like why. Why is it interesting to do it in a speedy way. And I think it's a very simple little example. But here's a statement. I started out in the middle of the public space. Pretty simple statement. It's two, four, six, I think nine or ten words. But I don't use this. I started out in the middle of the public space. And I, if you can tell, is good. That's pretty simple. Back in the middle of music. I started out in, that's how people say it here. No, no, no. The official phrase could be either. You have to know about your context around it. That's how J.I.R. works. You need to know your context around it to figure it out. You can't just say the same sentence. It's way more like, you know, you just figure it out yourself. And I did it to kind of, find these two things, one inside a full volume. Which is the chapter you can write down for ages. You can use it. And then, there are values like, which one is better interpreted. And you can see the kind of, this one here, this is the one I'm filming on a hill in the town of Skil. This one. I don't know. I think you put a Muslim in the town of Skil. You've got that here, It's a good man on the hill, so it shows one scenario of the others, probably because the English language across the entire web of internet, that's the context that makes the most sense. It's used more often. So the other things to kind of be thought about as we use these techniques and technologies and approaches, lots of examples of big companies and things that they didn't have done. And a lot of point out about the other points, not the specific article itself, sort of about we use an animal to find a lot of other things as well. It's not a point out that I guarantee you it really does not want that kind of thing. I guarantee you it really does not want that kind of thing. It's not like they say it's a big sentence, but actually it's not a sentence. That's not what happened. They say it's just big things, and they didn't only say it's something to deal with about how biased they can be. And if you're really talking about generally, there are a lot of different machines that all work on data. And what I didn't say, what I should have said, is that data is biased in any way that the model would be created by. And so understanding where things have challenges that create a very general environment where they couldn't get really good feedback or maybe it's a little bit complicated like me, they didn't put that thought about how they think about understanding that my data is biased. In addition to bias, there are other things that we can do about it. Is it fair? I'm just going to focus on this one over here. It was hard to be fair, you know. I hope people are able to see what it's all about, stuff they're understanding on the point of view. And they placed things differently based on where they live. So when you look at the model, it's 5% right if they were in high up. Not because transportation is more or less expensive, it's just they think people in New York will pay more. Is that fair? Good question. I want you to come through. And if you think about it as fairness in a couple of different dimensions, then we've just been fair for pricing this English. I'll give you a shout out for being a lawyer, but it kind of helps you understand different definitions of what needs to be fair. So we can make plans and we can say, for example, that we're going to be fair to different groups, people in Ohio and New York, for example. But what I mean is, laws today are the same with different states, I expect. We're going to do this 67% for people in Ohio, 67% for people in New York. So it's going to be a fair fair. Another definition is, I'm going to say, if there is an 80% chance that person's going to be fair to everyone in the world. So that's fair. Because everybody has a point of action when they live alone. There are not many guys. If it so happens that if you're in Ohio, and you're going to have a better record of what they're going to pay your loans. Well, I don't know. It's a result of information that we're going to share because we're going to just use the expected payment. But this one. The better one is, well, I don't know, just because I'm from Ohio, it should give you a better revenue or a richer advantage. But maybe I won't pay back the expected rates. But I don't know, just ignore the fact that where you're from, obviously it's where you're from, Ohio versus New York. It's other attributes that might be more distinguished by this. So anyway, two of the definitions are fair. They're all fair. Just in some contexts, some are more appropriate than others. Some are more useful than others. So if you think too much what scenario it is, it's not just a question of, one of these is that I thought we should talk about. It's not there. Any questions? Yes. Back to your hierarchy chart. Yes. So between the stream learning and deep learning, is that deep learning really a subset of neural networks? The question was deep learning and machine learning and running neural networks today. So we're going to kind of run through the question. Yes. So I would say that's very much all of this machine learning and that neural networks is a class of machine learning techniques. And then the thing that deep learning is a class of neural network techniques. And JVRI is a class of deep learning. So what are the nested, if you want to think about it, the fire, which I'm waiting for. If they live in the library, they live in the library. And that sets out each other. I have a question from online. Person wrote, given the extreme probability of a deep base, it's Gen AI is improving rapidly. Is there any work on non-recudation techniques so the real people can be verified about benefit when seen or heard? Yeah. So that's a good question about everybody. But I do have a question about deep-based and potential issues. And it's not just visual, right? It's even just trying to be included. All three of those things, they might not be true, but might not have said it. So you've got a lot of work to find and understand how to use AI to develop AI. That's one of the topics. The third topic is learning, because you can use a box where you can kind of check on box. And so I don't want to say what it's like, but you need to be aware of that when you're in the research area. Yeah. So I'm just wondering about the diagram where you have three studies, data science and AI, on which you learned one of the studies. And I was wondering if we have an overlap between all three. What is that part that is not AI or what you're learning in data science? So if you actually look at the definition of data science as a phenomenon, that is, what is that? I think that's a fair question. So that part is what people typically think that this has to be our part of the data science life cycle. So things like understanding the business context. If you're just learning a machine learning algorithm, you're building it across more than one user. You're not really focused on a particular domain context. But when you're learning something in data science, you have a machine that's specific to many contexts. You come to a machine learning algorithm and you might use like multiple of your machinery algorithms. But all of that is because data in the data science project is often here. Then you use a machine learning model or a multiple machine learning model to make some predictions. And then once you have a machine learning model and you have to put that into a way that more and more people will understand. And that actually validates how accurate it is. It's not just that it works or whether it works but that it's not an accuracy, because it's useful. But all of that works around looking at how accurate the model is in the context of how it's being used. But how do you explain that to kind of like the users of the model? But how do you actually respond to that in some real old scenario? All of that stuff is definitely followed in data science where the machine learning has actually got created with this model. But that's just a very long time to start with this. Yeah. We also face the opposite problem that, for instance, something that was human-creative can be said that it was AI-generated. Yeah. So that's actually a big problem, which is that's the hardest time on the entire world. So the hardest time for me is when I say I know my boss or my professor said he would be right there to do that but I really would. So it's that kind of issue that they come in two sides of the coin. Which is, AI is language what kind, human-generated versus what a human-generated but where is it important to understand that distinction? And where is it not so important? But when AI is coming together and really asks me if I use the spell checker if I use the spell checker, right? The first one, if I do the AI-generated I just take a tour of everything else. And that's generally how you use this spreadsheet to do some equation. And at the same point, you're going to effectively say what is acceptable if we use these two scores? And the other is not acceptable but we have to like set speeds. Very loosely, I don't understand that yet. So it's pretty much on both sides. It's definitely an issue and all they do is you didn't actually create that. So you don't have that issue. So we're following up on questions. So both of us that are here in the middle what is our reinforce today? Does it disclose everything or what approach we use? I just had an article that somebody sent to me and they said we need to cut this down to a thousand words because that's all they said. Right after Jeff, he said we need to cut this down a lot of this and probably just make it right back. Yeah, so that's very nice. We're going to hear that in Zoom or whatever. So somebody had a challenge of a lot of other people who needed to cut down to a thousand words. It's hard to do that but you know what, touching between these people really means that. Especially if it's not acceptable. In many cases it's probably acceptable because taking your word is just inside you. It's not like it's clear where they're going. So I think in many situations that's going to be fine I don't even need to be weird because the other thing that is there are bad models often is creating things and they're not telling you they're creating things. They kind of just pretend it's bad. So if you're giving it often that's what you're doing but if you're doing the kind of example I'll create an example. Let it happen that doesn't exist. You've got back in the world you didn't ask the reference system to reach your fingers that they were but I don't even know that paper I don't think I read that I'm talking to you and you're just going to stand and say I'm not going to cut it off because I'm not going to buy it at all. That's actually what it has to be and it's actually doing this now I think this is true that everyone has it's back. So I wouldn't be telling you that But I would definitely have to read it over with the fact that it's actually in the case of one of these submitted things. So read them, and like at least read them personally, then it's okay. Yeah, last question. Yes, from what you talked about in the real world. And you spoke about actually inside it. Yeah. And you also said this means like they gave you something and they recommended, for example, an ambulance. So how do you control the addiction for those things? And then like things of, for example, you know, because they do recommend you to know that you watch a lot of, you know, things of videos about how to create things. You click next, it takes you to another thing where you walk more and you just keep going. Yeah. Yeah, so the question was, I don't know if that's kind of a word or anything, but I understand kind of the suggestion part of the suggestion. That's a kind of, that's a hard question. That's sort of the side of this. In the intro, I think that's an active research question. Maybe I would say that. The first part is to acknowledge that that's the world that I live in. Because we've been trying to do this together and it's not doing knowledge. They're analyzing people with knowledge. I thought it was a lot of questions all the way about it. Large language model is so bad at facing math. There'll be the right audience. So no matter what you're trying to do, especially when you're pushing people like basic math, it has to be kind of just the way the model is constructed. So what I would say is those challenges are being improved. So if you ask the math question today, it's better to do it. I'm sure you can say that. I see that in the math plan and I want to kind of explain the next topic, which is, why don't we do that to raise some of the questions that this came up? We'll make a transition, but thank you very much.