DataFramed
DataFramed

Episode · 6 months ago

#96 GPT-3 and our AI-Powered Future

ABOUT THIS EPISODE

In 2020, OpenAI launched GPT-3, a large language AI model that is demonstrating the potential to radically change how we interact with software, and open up a completely new paradigm for cognitive software applications.

Today’s episode features Sandra Kublik and Shubham Saboo, authors of GPT-3: Building Innovative NLP Products Using Large Language Models. We discuss what makes GPT-3 unique, transformative use-cases it has ushered in, the technology powering GPT-3, its risks and limitations, whether scaling models is the path to “Artificial General Intelligence”, and more.

Announcement

For the next seven days, DataCamp Premium and DataCamp for Teams are free. Gain free access by following going here

You're listening to data framed, a podcast by data camp. In this show you'll hear all the latest trends and insights in data science. Whether you're just getting started in your data career or you're a data leader looking to scale data driven decisions in your organization, join us for in depth discussions with data and analytics leaders at the forefront of the data revolution. Let's dive right in. Hello everyone, this is adult data science educator and eventelist at data camp. One of the most exciting inspiring developments of the past few years in data science has been the rise of large language models like gpp three, in case you've been living under a rock. Generative models like GP three for text Vu two for images have shown the incredible potential for what an AI powered future would look like, whether it's automatic image generation from prompts, sophisticated code autocomplete, the possibilities are endless and that is why I'm so excited to speak with Shubam Sabu and Sandra Kublic, authors of the Ry Book Gpt Three, building innovative and LP products using larger language models. Throughout the episode we talked about the rise of large language models. The underlying technology and how it's different. Why gpt three's API interfaces revolutionary in machine learning, potential use cases, it's risks and limitations and much, much more. If you enjoyed this episode, make sure to rate and comment, but only if you enjoyed it. Also, I wanted to let you know that this week, access to the data camp premium and data camp teams is completely free. What does this mean? It means that all you need to do is registered to gain access to all our learning content, access to data camp certifications, workspace competitions and much, much more. Make sure to take advantage of the offer with no strings attached. Now on to today's episode. Shubam Sandra, it's great to have you on the show. Yeah, pleasure to be here. Thank you so much for having us. I am excited to talk to you about GP three, your book on it and what it means for the future of AI and data science. But before we get started, can you give us a bit of a background about yourselves? Yeah, sure, my name is Shabam and I started as a data scientist initially, and during my time as a data scientist, I got to work with a Fintech firm where I established the entire machine learning and data practices for the technology infrastructure, all from scratch. Then I thought of doing something for the community and moved into the role of evangelists, where I got to foster the ideas and thoughts of the community members throughout the spectrum, and right now I'm working as a senior AI evangelist at Gina Ai, which is a new research company around twenty twenty. When the opening I API got released, I was in the early members who got the access to the API. I was literally amazed by the things that it can do and have experimented with the API A lot. have been posting a lot about the use cases that we could do, started writing blocks on it, and that's how it all started, and that's how it all converted into an orderly book, where I got in touch with an orderly editor and she was really excited to have something out on this topic. That's how it all got into the place. Yeah, I have on typical background when it comes to AI. I was a liberal arts major and I was always drawn to creative projects, so I used to think that I will become an academic or writer. I was experimenting a little bit with movies as well and then I pivoted to start up ecosystem because I always loved tech. I always loved how it enables us to improve our lives, to make them as friction as as possible. So I was always in love with it and I just wanted to be closer to it. And also so I felt that the being an outsider at the time, that...

...the breakthroughs happening in ai are something to be observed and to watch closely, and I wanted to just get involved in whatever form I could. I started with setting up a Hackton community for enthusiasts. That's how deep learning laps was established, and then it organically grew into an incubator for a startups, ne spret. I also started a youtube channel just to feed my curiosity and give myself some space to discuss and think through these a breakthroughs that were the most appealing to me and obviously, like everybody here, I guess was mind blow when Gpt thrills launched, I was likely enough to also become the early tester in the BETTA and yeah, I guess around the time Shabab reached out to me. That's how he found me. I think through through the videos and he offered me this awesome project to to write a book about it, to learn more about it. So of course I just dived into it and for the past year or so I was working for a Neptune Ai and right now I want to continue on this mlp path. So I just enjoyed here and we are also launching the book, so it's super exciting. So I'm excited to talk to you about GPP three and your book on it. There's a line from Ernest hemingways the sun also rises, where one character asked another how did you go bankrupt, and the other character responds with two ways. Gradually, then suddenly, I'm always reminded of this when looking at a lot of the results from large language models like GP three. It feels to me like an outside observer like that the AI community has been doing a lot of gradual improvements and ope systems and it has suddenly resulted in awe inspiring systems and outputs. I wanted to set the stage for today's conversation by first understanding what makes gp three so interesting and how is it different from other machine learning systems that were used to? So when we talk about GP three, it always makes sense to look up a little bit of the three from it all started on what's the origin of all these language models? So it goes back to seventeen when transformers got introduced, and it has changed the direction of n LP, how the field has been looked, how the field will progress and how things will work in it. So transformer was one of the revolutions reduced of attention, which is basically looking at certain things similar to how a human does. So an AI model which can exactly replicate how a human brain works. That's where it all started. Then researchers at Google, open AI, Microsoft, they started experimenting with how we can take this forward and make something that is usable for general public or audience or someone's an engineer or a data scientists, and how all of these use cases can be put to use in real world use cases and how businesses can be formed on top of it. And that's how the GPT series got originated. So we didn't directly lended a GP it was a part of generative predaen series. We had gpt one and we had gpt two, then we had gpt three. But what changed with GP three and what makes it so exciting and what makes it so revolutionary as this was the first time that we saw that an AI model and LP model or a language model can do task, can do any number of tasks, is not limited to a specific task what we have conventionally seen in n LP. So, just to give you some context, like learning model, how it works is you give it a training set for a specific task and then you train it on the training data set and that's how you influence it or that's how it performs a task, a specific task on which it has been trained on. But gpt three, because the data it has been trained on comes from such a big universe of Internet, it can literally perform any number of tasks that you can think of, and so it is the first time that we have seen a task agnostic or a task independent a model are truly generalized,...

...a model that can perform any number of tasks. And the other good thing about gp three is the kind of is the way you interact with the model. Right previously it was the thing that, if you have to interact with model, you need to have a technical pre exciety, you need to have an understanding of a programming language, you need to have data set, you need to understand how training works, how infencing works, how you can be deploy it. But GP three just there is all all these conventional paradigms and gave you a simple user interface where it is as easy as talking to a human. So you just go to the playground, you give us instructions in simple English and the model will come up with an output. It's like collaborating with a human or collaborating with a body or a subject matter expert. So you can also think of GP three as a subject matter expert for a number of tasks. So that's what makes that's what makes gp three so special, and I'll pass it over to Sandra to throw some light on that. These great points. What I would add at the top of that was that introducing GP three in the form of an API and giving a broader access to it to developers and, as Java mentioned, the interface cuts so simplified that people, without necessary, necessarily heavy m use cases. So after a short period of time, once the initial access was released, you could see all these use cases just emerging from the community testing it, and they were just like mind learn just translating legal documents into a simple language or analyzing the recipe of the product and then translating it into what are the ingredients that are harmful? What are the ingredients that are good for you? All sorts of things really and I think this is this was one of these like radically new things that got to interact with this powerful model via a very simply designed API. And you've got to actually explore different use casses at the top of it. And we talked with open API creators in the interviews for the book and they admitted themselves, for example Peter Welling, theer VP of product there. He admitted that they themselves, when releasing the API, had no idea what it is capable of and they wanted to give the access to the community so that they can show them actually the boundaries, the limits and explore further. And that was just a brilliant idea and I got us where it is right now. That's really exciting and I want to unpack all of these elements with you. So let's start off with the technology underlying GP three. I mentioned a bit here the gradual work that the eye community has been doing to improve the technology underlying d three. What are the changes that have happened over the past few years that led to these high performing results. Chadoon has already mentioned that we had this major paradigm shift in the NLP once that transformer architecture was introduced. So we started in two thousand seventeen with the famous paper at tension, is all you need where this architecture was launched, and the backbone for transformers is a sequence to sequence architecture. Basically, transformer model processes a sequence of text all at once instead of award at the time, and also has this powerful mechanism that shown has mentioned, called attention and transformer architecture is definitely a key thing to to highlight when you think about the changes that have led to the bar of gpt three. Another one was that with time in the NLP space language models. Initially they weren't so big, they weren't so big and impressive, but they started to become bigger and bigger and the data sets that they were being trained on were becoming bigger. With more and more data availability. Was Open source projects where researchers put together this massive data sets and was just easier to train these models on. Another thing that sort of parallelly emerged was...

...more and more computing power at the hands of the guys that have the computing power, and so it allowed them as well to train more and more powerful models and experiment with bigger and bigger architectures. Another one was that, okay, you have this powerful computing power, but at the same time you want to find ways in which you use this computing power in economic efficient way so that you don't run out of it, so to speaks, and simply put and one of the one of the techniques used in GP three was pre training the model, and this basically helped to reuse all the initial training, this very lengthy process of earning of the model, to be applied to other use cases, with a little bit of fine tuning or a little bit of tweaking to a particular use case that you have in mind. So that was like a big one as well. Chev I'm also mentioned that there were iterations of gpt before we arrived at gpt three, and there was gpt one, I think it was introduced in like eighteen where it had I think, like around One d twenty million parameters, then gpt two, ten times bigger, with a bigger data set as well, and then eventually gpt three, which was a hundred times bigger and also had a hundred times heavier data set. So they were constantly working with opportunity to have bigger data sets and also with opportunities to have more computing power and also seeing that scaling leads to the emergence of more powerful language capabilities, and these models were actually being able to do more and more performed better at a variety of tasks. They went this path and that's how we arrived at gpt three, with which hit the sweet spot. That's really awesome and I'm excited to talk about the scaling aspect here of gpp three. And where does the world do we hit a wall when it comes to scaling? But you mentioned earlier in your chat here center, especially the API model and how transformative it is democratizing access to working with such powerful models. The API model of g three is definitely interesting and I think it does introduce a paradigm shift and how we interact with complex AI systems. Even as someone who hasn't necessarily codd a law in the past two years, found it very intuitive to work with gpt three. How does the API model change how we interact with M L systems, and can you walk us through the concept of prompt engineering. So we have seen like in twenty it was the first time somebody has introduced as a based approach. Previously it was all hosted locally. You train your model, you hosted, you get your own data sets and that always lets you hit up ball on the past that you want to perform. So you can only lease to a certain limit or certain level of accuracy when you're training on your own data set, when you're hosting on your own, because there are technology in technology limitations, infrastructure limitations, posting limitations and whatnot. But what openly I decided and what I think made it revolutionary was giving GP three's access in the form of API. It allowed people who don't understand coding, as you've mentioned right you haven't coded for quite some time, but when you use GP three it is very intuitive. You don't feel like you're coding something or even you're interacting with a sophisticated language model. So all of these things doesn't come into picture. When you interact with GP three, it's as simple as talking to someone and getting the output by giving the input, and input is as simple as natural language or simple English, and you get the output of whatever you want. Want. It can be generative output, can be searched output, it can be a number of cases, classification and it recognition all the things that is possible with conventional energy and the process of giving this kind of input and getting the outputs, which is a kind of organic or natural process, which is as close to giving an input...

...in English is from design. So it is an intuitive process for people without any Emil expertise or Emil background that they can give a textual input to model in simple English and get the desired outward and whatever you want. So let's say if you want to write a paragraph on n LP, what you have to do is simply ask GP three, can you please write a paragraph or n LP, or please write a paragraph on n LP with this and it comes up with a paragraph. It is as simple as that. Some tips on prompt design and engineering that you have to keep in mind while interacting with GP three is to understand what gpt three knows about the world and giving the input in such a manner to leverage the knowledge of gpt three. So GP three is not great at giving you the factual answers it can create because it generates things on the fly. So it is really good when you have to complete something, when you have to create some when you want to go creative and when you want to put abstract things out in reality. Right, because we have seen a lot of artists, a lot of illustrators and people from design background getting attracted to GP three and getting and using GPCRE, because all of these people had a lot of abstract ideas going through their mind and they didn't have any idea of how to represent it or how to put it, to put that into execution. Then they came to GP three, they gave it as input in the form of prompt and got all those output for those abstract ideas. So it's basically like acting as a sounding boat for these kinds of people and making it easier for them to understand what they actually want to do and helping them with their creative and design process. Yeah, maybe what I think it's a great take on front design. Maybe what I would add to some tips when it comes to interacting with the model would be that to realize that gypdig is like super, super good a storytelling and it's going to continue in the same fashion as you would prompt and if you start with like science fiction novel with a few lines of a science fiction novel, it will continue in the same way. If you will start with a line that looks more like a love letter, it will continue in the same way. It's just it's incredible at being able to move between different styles and mimicking and continuing in the same fashion. So the most important would be to making sure that the initial input that you give it hits, that checks of that requirement. If you're going for a center certain genre, just make sure to give it enough of an input so that it can continue in the same way. Yeah, and another thing would be that if you find yourself getting inconsistent messages, inconsistent outputs from gypp three, just make sure that you give it enough of a context to make it consistent. An example that comes to my mind would be like a question and answer use case where you try to get sort of trivial style questions and answers. You give it a question, gives you an answer, or you ask it to create both questions and answers and without enough of a context, it might get it might give you some answers that are non factual, that are just made up because it has all these all these data at its hand and it doesn't necessarily think in logical factual ways. However, if you ask it to be factual, if you say okay, write a few trivial questions with factual responses, then you're going to get the factual responses. So it's as simple as that, just like giving it enough information of what you're trying to achieve in order to arrive at the desired output and thinking about it as if it's a one metaphor. Is just like talking to a friend in a bar and trying to be as simple and concise in your messages as possible so that the other side understands what you're going for, and...

...then you should be good. I love that last part, especially on the question and answer style prompt engineering. You know, one example that I've seen which is one a testament for the intuitiveness of the question and answer style prompt engineering, as well as for the emergent capabilities of models like gpt three, is creating jeopardy style questions and answers. I saw this example recently. Five, seven, eight, ten years ago, you would need to train a model specifically on jeopardy questions to be able to reach that level of parity, but just with a few prompts. For example on GPP three, it has been able to blow that specialized model out of out of the park right just through that prompt engineering. It's really interesting in a sense because it showcases those emergent capabilities of GP three. It's not trained on that ask but does really well just with two or three prompts. Yeah, exactly. As we mentioned, it's extremely good at very quickly figuring out what you would like to achieve and as long as you give it enough of a context, it should get there. At this point it's like extremely generalized model that can be applied to so many language based tasks. I think there's a reason for why we are still waiting for, say, the next iteration GPD for let's see what will happen in the future, but at the moment gpt three, it's already so usable and so appliable to different types of tasks that you can really achieve a lot with it with just a little bit of a nudge in a certain direction. So we talked about scale and a lot of ways, scale in the terms of the data ingested and the number of parameters of GP three has been a massive factor in why it's so good and why it's so easy to use. How important is scale as part of GP three success, and what I'm trying to get specifically here is reaching generalized intelligence truly a matter of scale. First of all, when you look at opening ice mission that what they are striving for is all the projects basically that they engage in is arriving or facilitating the development of a g I that is benevolent and beneficial for as many people as possible. So with their experiments with the models, they certainly are trying to arrive to as general intelligence as possible. So with initial expect to Gyp d three, the scales extremely important. It was crucial. They were being taken aback by how much the model capabilities change when you add scale to it, when you leave the same sort of architecture, when you leave the transformer as a backbone, but when you just make it fatter and bigger. And the iterations that followed gptwoe two, gypy three, they weren't that different in terms of the architecture. It stayed the same, but what they were doing was they were increasing the number of parameters, increasing the data sets, and this is how they were trying to see whether it changes and whether it gets better at certain benchmarks or general language based tasks, and it's proved to be true. So that's why they were incentivized to go in that direction. Having said that, with scale and with certain and with scaling of Computing Power, they're also come costs and they're also come concerns, for example related to the environment and as open areas. Was Scaling their models, there were more and more research showing that we should actually be more aware, more careful of how we are using this computing power because of the economic and ecological most of all, footprints. So one of the one of the research papers that I saw that basically compared how much carbon footprint gpt three generated compared to let's say cars. It showed that the initial training phase, that last couple of months, was comparable to a lifetime of five cars, five passenger cars, that generate...

...a certain carbon footprints. So it's massive when you think about it right like. It's a lot. They are aware of it and they are trying to address it and we no longer think that only scaling, blowing things sort of proportion and, you know, reaching bigger and bigger levels is necessarily the answer to arriving at more generalized intelligence. I think we are looking at experimenting with more techniques that are trying to achieve the same level of performance, but on a lower scale, which is like some tweeks on the architecture, and I think we also are not only scaling language models, but also involving other modalities, like audio visuals, in order to arrive there, and I think this will be more of a direction where we will go to in order to achieve this more and more generalized intelligence. Following up on what Sandra said right, the other way to look at generalized intelligence, apart from scale, would be how we can make changes in the architecture and use the same number of parameters or maybe, yeah, along the same lines. So the other way to look at it would be how we can combine different modalities and how multimodality can be brought into picture. Because G P three, if you see, even in all its glory, it still works on text. It's just a text based model which use texts as input and gives texts as output, as simple as that. But if you think about combining different modalities, right, combining texts with image, audio video, how we as humans perceive things right. It's not just images, it's not just audio, so it's a combination of what we see and what we hear, and that's how we perceive things and that's how we make sense of things. So to get to generalize intelligence which is similar to humans, we need to take into consideration the component of multimodality, and I do think that we are moving in that direction in the next iterations of future fagistic language models because, if you see, recently dally too got introduced, which basically takes text and images and it generates images much better than any artists can do in that given time. So within seconds it comes up with brilliant images given the text prompt. And again, your text problem can be as abstract as possible to still come up with images because it has been trained on billions of images. So futuristic language models can be a combination of dally and GPG three, where it combines different modalities like text and images, and then it can make sense out of it. We also look at the other research in the same area, as we will get to know that Google, step mind, has released Cato. So Gato is a generalized agent which again combines multiple modalities, and not just text. It combines text with audio and video and it is a multimodal multitask like which model. So what it can do is it can use the same ways and it can play Atary, it can capture images, can chat and even use a robot Um to do a number of tasks. So this is the direction that we are moving towards in the future. And again, a very dressing example that come to my mind is second so second what it does is it combines the advancements in language model with robotics. So you have this understanding of language model universe and you combine that knowledge with robotics and then it's as simple as giving command to a robot and the robots become smarter in doing all the task that you want them to. So multimodality is definitely the direction that we're moving forward and this is where the generalized intelligence can come I definitely agree with that notion, especially on multimodality and in some sense reaching a form of generalized intelligence, and I use here like air quotes, for generalized intelligence is both a research problem but...

...also a system architecture problem of how can you combine different tasks oriented AI systems together, and I do think that even if the tow a certain extent on the research side, like the goal post for what defines generalized intelligence moves, we will, to a certain extent in the future, see useful generalized models be actually used in real life, and I think this marks a great segue to discuss some of the greatest re use cases that you've seen GP three produced. There's a lot of actual startups and tools right now that are built on top of gpt three. Can you walk me from of your favorite use cases of GP three so far? Definitely, and it's very exciting to see like how the next threew of startup has started on top of GP three or built on top of GP three. So while researching for our book on GP three, we have the section where we discuss about the start up ecosystem, the coporate ecosystem and the entire effect on economy, on for language, models like GP three can have. So we we did came across a lot of different use cases and it would be right to say that GP three actually acted as a launch pad for these startups. Some of the use cases that I really like that I have to point out would be viable what it does is it is a feedback aggregation tool and intelligent feedback aggregation tools, which can combine all the sources that you have, your customer feedback, your internal documents, all the insights that you're getting from different sources, and it puts it all together and gives you a proper user interface, or simple user interface, where you can just ask questions and get simple answers on how it works. So you can see the questions like what's frustrating our customers about the checkout experience, and the application may respond like customers are frustrated with the checkout flow. Fload, as simple as that. So data and analyzing it, like what went right, what went wrong, and then coming making decisions or growing out inferences out of it. That entire curve has been reduced to a simple question that you can ask to this a model, to this TP three based application. So it has really simplified the life of product managers, founders, customer success teams and all the front in side of the teams and the people who are working in these startups or these companies. So prohas one of the very interesting use case that I had that I think had a lot of value in real world and can be definitely used for conventional data analysts instead of doing that. The other interesting use case that I came across, it was a very recent use case that I came across, is super means. So it basically uses tbt three to generate memes. It was one of the most interesting and funny use case that I came across. So what it does is it takes it takes input from you. What do you have in your mind? It allows you to select a template and it runs CP three and comes up with different kinds of themes in a matter of seconds. So you just select a template, gives what whatever is in your mind. It comes up with a lot of memes. So now it's very easy for anyone to be a meme lot in the world of tpt three. I'm actually looking at examples right now and they're pretty hilarious. MEME assistance as a golden a golden use think. But yeah, I mean, I mean generally, my my favorite use cases are also like around the creative use of GP three. I think it's just such an incredible storyteller that it's just made for these use cases. It's very natural for it to create a story behind a certain character. You have all these examples, like a dungeon, for example, or you have just like text based adventure game, where gpt three is the powering engine behind all these characters and stories that you engage in and you create as you go. One use kids that we've got to actually explore deeper in the book was fable studio. Fable...

...studios like this pioneering VR studio that is creating a new genre of stories using new technologies, using virtual reality, but also using AI, and they have experimented with GPD three to basically create the messaging that the content behind its character, Lucy. They created this amy our twinning movie called the wolves in the walls, and they have this character that is eight year old girl. Lose many appearances, for example on twitch, where Lucy was like pan just engaging with the viewers or was singing a song or just like telling a story and they could engage with it, and they told us that of all these Lucy appearances, who are powered by GPP three. So that's just incredible that you can basically create a character with the help of the model. I think there is a big potential there and so I'm very excited that about use cases like that. Another one, also related to writing, was basically creating a copy that allows you to, I don't know, sell a product, create a nice social media post. GPT here is also incredible at that. So there are many use cases such as copy ai or Jasper or copysmith platforms, were you are able to literally, with Thein Seconds, generate very nice social media posts, articles, what have you, whatever you need, youtube video titles, Youtube video scripts. It's just incredible how much it can be helpful with a variety of texts when it comes to the digital realm. So that would be also one of my favorite use cases and also we typed into it in the book. That's really great. And what do you think are use cases that will be truly transformative in the short term? Short term, I think it's key here. I think we can already see how how all sorts of assistance related to the creative work are transformative in the sense that they allow it to do more, faster and maybe in a more fun way. GP Three is incredible for writer's blog when you're writing something and you just feel stuck and you want to generate a handful of paragraphs to choose from or just just too keep the creative juice flowing. So that's a great one and it's already available now and I think it will impact our comfort and our creativity of writing texts, that's for sure. Another one would be coding assistance. So GPD three is not only trained on human language but also on programming languages, and so it can be used to power, for example, coding assistance where, just like with github copilot, and I think they use codex, which is a descendant of GP three, it's basically like a younger brother of GP three. They are using it to help you either learn how to code or to fix certain problems that you arrived at when you're coding. I know already people that are using you have the pilot, and I think this use case will be growing in the future and will definitely again change the comfort and sort of the creative process of coding. Yeah, I definitely agree on that assistant type use case. There's also an adjacent use case that I've been excited about, which is very educational. Like one thing that I've seen is explain your my code or explain this piece of code type use case, which I think is going to be really great for democratizing education and programming, data science, Etcetera. On the flip side, it's also very important to also acknowledge the limitations of GP three. Large language models are definitely far from perfect and can make it best some pretty basic mistakes and, at work, some very harmful ones. Can you walk us through what the limitation and risks for systems like GPP three are? Before we look at the limitations and that TP three persons, it's...

...very important for us to understand that chip e three is not a truth teller, it's a storyteller and if you think of chipt three as a minute replic of human brain, just like all other humans, it is also poised to make some mistakes and it can be perfect right. We also do a lot of basic mistakes in our day to day lives. Similar with GP three, if the prompt or the input given to model changes of varies or contains some harmful keywords, I can come up with a response that can be harmful in nature, that can have pretty basic mistakes. If you ask for a factual questions that goes beyond the time of the training, that goes beyond the time when the model training has stopped, it will definitely come up with the wrong answers or make mistakes, and des Poised to make mistakes there right. So it is perfect when we talk about the generative of the creative capabilities of chipt three, but when it comes to a factual answer, we cannot think of it as something that can be perfect. And other potential rates that it possess, and the biggest one that I can think of is misinformation, because it is capable of generating a vast amount of data, a vast amount of information, just through simple prompting and the fact that anybody can do it. It was a very big risk of a huge level of misinformation that we can see on the Internet. It can give rise to a lot of propaganda boats which can just spread information within seconds and that which can spread a large amount of information within seconds. It can value the quality of information that is already present because in today's well, it's very difficult to verify the sources and before you verify the source, that information spreads like fire. So it is very difficult to control the quality of information that's available online. And even with GP three, we have seen this example where somebody has created a click weight blog post and pushed it on hacker news and then it just spent viral. It came on the top of Hacker News. Um, yeah, so this like the spread of misinformation, bias and all these things you can do. Yeah, I would add to the misinformation. But we actually did some research on the research out there involving GP three and expicient. There is this really cool report by Georgetown researchers released I think within the past two years. They are basically taking gpd her and looking at all these different use cases that are well my level. You know, let's put it let's put it like this, all these possible ways in which GP three can power misinformation and in which in which they can go wrong. And this report I was really struck by how easy it is to generate stuff like, for example, tweets that propagate a certain idea in the certain lights, targeting a certain group of people. Gpt Three is actually very good at that and you can do it easily, you can do it fast and then, as shaven mentioned, you can have an army of boats on twitter that just took your agenda forward. So it is an actual risk. It is scary how good it is at giving sort of the powerful voice to anybody, and in this group we include people that have some sort of agenda, political agenda, what have you, behind it. Misinformation is a huge one, I would say, another one by us. Here's the thing. When your Dataset was created, the guys weren't exactly thinking that all the Reddit or the or the reddit posts, all the reddits that they were scrapping from the Internet to curate this data set, to create this sub sample of humanity or what have you, they will have all these ideas that are extremely racist, extremely sexist, extremely X Y Z that they will be propagating. Started certain stereotypes, political, social, economical, all sorts of...

...stereotypes, and what happened was that, because the data was in the Dataset, GPD three, for some reason is able to now amplify them in its responses. So it needs to be curated, it needs to be filtered in order to prevent these responses. Initially, it was a big challenge for the opening it team and they were working super hard on addressing it. They introduced the content filter for the responses so that they are safe to use, and they also introduced this it's called process for adapting language models to society. That's the name. So essentially what they did was, how do we make these models that have these crazy amplified biases that we did not expect and we didn't and we do not want to have nicer, more adapted to society? But their models basically they came up with this process where you create a set of values for the model to follow and it's actually is able to adapt and to follow and become more more usable in the society sotistics. So that's another limitation of the model which comes from the fact that in the data set these two tips were included. That's really great. I really appreciate the holistic answer. I also share a lot of the concerns that you have, especially on the potential for misinformation and bias and also creating personal bubbles for people on social media. If a lot of the content tailored to your penarated and it's tailored to your preferences, we could have the risk of supercharging social media's capability of creating political bubbles as well as social bubbles, but even to the concept of personal bubbles where all the content is sailored to you and it's also generated for you, and I'd love to discuss that at the end of our chat. A lot of the concerns that you mentioned and also connecting back to the concern of environmental economic costs as well. What do you think are some of the research solutions or safeguards that are being developed right now to be able to fix these problems in the long term, and how do you think in the short term, teams using GP three will have to reconcile or work their way around these limitations? The one good thing about the gpt three was that opening. I was aware from the very beginning that these can be the potential limitations and can potential a potential risk that a language models with the scale of GP three can have. So they had a dedicated team working on AI, tickes responsible ai and defining an AI policy to safeguard the end users from these kind of potential risks and harms. So the thing that they had is, along with the language models that they have built the different variations of GP three, they also built in parallel a content hintering model. So whenever you give an input to GP three and it comes up with an output, there's a clear lining where if the output is safe, it is highlighted and Grean but if the output contains some harmful keywords, it is contains sensitive content or contains content with this sexiest, with this races. It highlights and it gives a warning that the content is harmful and not good to use. So that's the quick fix, that opening I came up with. I'm not saying it is perfect, but it is at least something and we are moving in the direction where we are thinking about responsible ai, we are thinking about Ai thics and how we can tackle those challenges and in the short term, what teams can do, or what end users for using GP three can do to avoid these risks is too go about their problempt smartly. When I say smartly, they can be careful about the problem that they're giving to gpt three. They can see that whatever keywords that they are giving to GP three does not from the model to generate some harmful response or some sensitive content. One of the very good example, or we can also call it as a misuse of GP three that end users third was of air dungeon. So it was a stick storytelling experience, entirely virtual experience where you give certain inputs and the stories and different worlds get created and was a kind of very realistic game, but then people started using it...

...for sexist things, racist things, and because the model doesn't have a ball. There is no thin line where model can deferentiate what is good and what is bad. It is again a machine which will do whatever you ask it to do. So there needs to be some safeguards that need to be put from opening. I who has designed the model, but also the end users need to understand their moral obligations and basic duity when they are using these kinds of models or techniques. That's where the air policy comes in and that's where the air thinks as a subject comes in, and I do believe these will be bursioning field going forward and in long term we'll see a lot of research on AI thinks and air policy and how these models can be used. One good example which I can give you right in current times you have seen a lot of talks about value targeted data sets. Right. So, as Sandra correctly mentioned, no, none of the AI or language model or on an en mpy model by its origin is biased or has yeah is biased or have misinformation or capable misinformation about the data set it has been trained on. So it's a data set that has been generated by humans. It's the data set we get from Internet. Those contents bias and that bias gets propagated to the language model. So recently we have seen this concept of value targeted data sets, where data sets are adopted to how the values of society are. So data sets adapted to the values of society, what we think is good, what we think is bad, and adapting those when in training those models on the value target data set. and Gb three has the good feat tuning where you can just take few samples, like hundred, five hundred samples and call it a small data set and can be and you in the model. That's where you can use a value target data set. So for your domain specific application, it makes it highly reliable and it assures that you don't get was kind of sensitive harmful response in the end when you clean off, when you find you on the gipt three models on a value targeted data set, and to know more about it, you can definitely check out the book. We discussed there in detail of how we targeting data sets look like and how you can find you on gpt three for your use case to avoid these limitations and risks. Yeah, I just want to add one one more point in which it shows that open is like continuously working on its on its models and it's improving the API. A few months back, they have released a series of models called instruct Gpt, and what they do is basically there are models that are trained to be much better at following your instructions, they are much better at giving factual answers and they are much better at filtering the unintended, abusive, violent would have your content. So I think they also not only they give tools the community to be able to herb these negative potential, negative outputs coming from the models, but also they are working on making the apis safer, is that when you're using the API in order to on some sort of application, you have built a product, you want to give it to the world, you are going through a process where you need to explain what this is for. They are looking at it in depths, they're looking at the type of use case that you're using it for and then they decide whether it's safe to be released to the world or not. So they give themselves the opportunity to put a stop to something. They just wouldn't like this tool to be used for. An example that shapar mentioned with a dungeon was that when both the a dungeons creators and open air folks realized that the model is being used for creating like sexs, races, content and so forth, they have used much bigger filters to the content and there was like a big push coming...

...from the open up this well stop because they are monitoring how the Apia is being used and they are being able to I would say they are definitely carrying a lot when it comes to the safety and they are. Of course, the models aren't perfect, but they're continuously working on it and we can expect, a SA mentioned, a lot of research coming making these models better safer to use. That's really great and we're reaching almost the end of our episode, but we've talked a lot about the short term use cases as well as the value of models like GP three. But I'd also talk about the future a bit as well, the Paradigm Shift ussured by large language models and the transformer architecture. I think it's truly something. You know, we saw this with the recently released catto system by deep mind, the many different large language models developed by Google, Microsoft and Meta, and the same vein of how the advent of the smartphone ushered in the end tools that we didn't thought or APPS that we didn't thought were possible before the smartphone. I think Uber, AIRBNB, etcetera. Where do you see the future of NP systems heading and what are some of the ways, were unexpected ways, that you think they will change our lives? So language models like GP three has completely changed the way we see and perceive the words. So, if I have to put it in simple words, it has just opened the imagination of what is possible and it has just change the realms of what is possible. We are living in very exciting times and we have a very exciting future ahead of us because GP three has the capability of models like GP three has the capability to replace the way people find and search for their information on Internet. So it can allow you to access customized and concrete information that is to the point for whatever you're looking for. And it's similar to replacing what we actually do with Google today. Right we search for information, we get a lot of results, then it's on up to us to make sense, to go through all these different web pages and then find what what's the information that we're looking for. So let's say, finding something or researching about something takes thirty minutes to find the complete examples and make notes of it. What GP three can do us as can give us an accept to the point information in a matter of seconds. So the thirty minutes of yours get converted into seconds and that's all the time you need to get the relevant information. Another important concept that I want to touch here is procedural web. That's something that I think we will be adding in the future. So what procedural Web is? It is a kind of internet where content will be adapted to the users. Content will be personalized to the needs and users queries. So what it can have is, let's say, instead of me going to google and searching for different results and it comes up with the rank number of pages and I go to different pages, what it will do is I'll search for something and rather than us searching from a select set of databases, it will generate things on the fly. That's like a human does. So if you search for something, it will generate that thing on the fly and you'll get concrete, to the point information. So it's as simple as asking questions and getting the answers. So that's the kind of future we can experience and we can get with the progress in n LP systems and large language models and moving towards more generalized Ai, getting information on the fly and just removing the time that we spend on research, because research is a part of it is a very big part of every job, right. It's not limited to engineering, data science or data professionals. Everybody who does any kind of job has to invest a lot of typee to research, and the only medium of research is Internet, and it can just completely change how we look at things conventionally and can create a future where things are more sortid, more when defined and more streamline and we can get information on our fingertips. That too, really concrete and to the point. I think...

...these are incredible points. And adding to that, I think not only large language models will change our relationship to information, how we consume and how we benefit from it, but also we'll make it more fun, basically, if to engage in searching for information. What I'm thinking of is, for example, being able to talk to virtual assistance that are powered by these large language models, that are able to have a really nice small talk with us about, or so all sorts of topics and then moving on to like certain more effective information exchange, but like having this really nice sort of human touch in interactions with the machines would be one of my would be one of my bets that. I think this will definitely increase our comfort with talking with childboats with virtual voice assistance will get better and better. Right now it's already fun, but I would say it's still pretty limited and can feel that when we're talking to our Alexas and so forth. Actually, my Alexa just woke up, but yeah, it's just it's going to be. It's going to be better and more fun. That's one thing, and another thing. Following up on this coding assistant use case, I think L L MS will allow us to create also more effective sort of communication with the computer where it will be much easier to translate from human language into coding, in which it will be also possible to translate from voice commands, vote natural language voice commands, into coding. And I can definitely see a future where I am talking to my machine and my machine is creating a game for me by someone I just subscribed that I want a certain world with certain characters in it and with a certain storyline. I think it's definitely going to be possible and also I think it's going to make people without coding skills, for example, more and more engaged in this process and just it will just democratize access to it. So that's myself coming from non coding background. I will be able, I am already able, to create like super basic games with Codex, for example. I think the opportunity there is incredible and we are actually exploring this bigger trend of combining the no code approach with large language models. In the book we are talking to bubble in the book, to the Bubble Co founder, who tells us how he sees this this aspect moving forward. Yeah, I'm really excited about the swam as well. Yeah, these are super exciting use cases and adding on top of that, I think there's also a potentiality for even people with low digital literacy or not really high ability to use computers to leverageless voice command to be able to use machines generally. You know, there's one example of a stept ai which is I think had a former open ai engineers working on it as well, and what they do is that they let you use your computer through voice commands. Say, download exl, do this, do that, and I think the potentiality of having a Jervis like assistant is going to be super exciting. So, as these models get better, what are you most excited about and what are you most worried about? So I think the excitement is obvious. Like you, you have this super powerful tax. You see all these different ways in which it can be like awesome, and what I'm just looking for is to seeing it more and more in their real life. Right now I'm talking about the potential, the different like early signs of where it can be really good at, but what I would really want to see is to be surrounded by all these applications, to be talking to my Alexa and having my blast, having an awesome conversation, or to be able to create this game. So I'm really looking forward into actually moving from discovering the possibilities into actually bigger, more wider adoption coming from not only startups but also enterprises, is that we use these products on a daily basis. I'm...

...just looking forward to that for sure. What really excites me is how this language model, like GP three, has the capability to bring in people from different backgrounds into the s ecosystem, which wasn't a possibility, which wasn't even near possibility before, because I used to be such a big puzzwords. People hear it and they were like, oh, that's not for me, I don't have the technical career because I am not a technical person, I'm not a research scientist. But with GP three it is getting much and much easier and more and more organic and natural for people to understand what it can do and come up with different applications or use cases in this ours, like we already saw in case of replica and Fable Studio, of how they actually came from a design background, film background and they combine GP three. So what replica did is it combines GP three with a virtual assistant, personal personalized virtual system giving you a personalized experience of chat experience. And what payable studio did is it combines GP three with metabors, like creating one storytelling and all these things, and how it work flawlessly that a person with an engineering background or a data science background would have found it hard to do. So it is very exciting to see how people who are not even related to technology can come and put their ideas into execution and the products that we get to see right. So I do feel, and I'm very positive about it, that in the next one or two years we'll see a wave of these startups, a wave of these products which will just blow our mind. And coming to what worries me a lot about gp three is, again going back to the bias and data sets that we have that the model has been to be on, because it is nearly impossible to eradicate those biases completely from the data sets because those have been existing for decades, for years, and they have been there and we don't have an option to create data sets from scratch. So that has been there. But what we can do and what I am really looking forward to, is coming up with research solutions like content filtering model or something which against streamlines whenever there is a bias or highlights whenever there's a bias or there's a sensitive content or comments on the output. So I am positive that we are going to resolve and take all these challenges with the research that's going on in the field of ethics and a policy. So, yeah, I hope to see this problem getting resolved soon, very soon. Yeah, I realized I forgot to mention a towards me as well. So I would definitely second shot in terms of the bias, but also what we have mentioned before about the potential for using this tool for political propaganda. And myself in based in Poland right now, so it's really it's really close to Ukrainian conflict, to Russian and Essian and I am seeing on a daily basis the campaigns which are designed at finding all these boats that are spreading misinformation within the context of the war and taking down these accounts and just imagining how gpts we could power these accounts just Carres me, honestly, that that would be my biggest concern at the moment. Definitely. That's a really great list and to your points, should especially on the combining the metaphors with a lot of AI generated worlds. To a certain extent, if VR matures and multimodal model matures as well, we can have the potentiality of a lot of ai generated worlds where people have their own unique, personalized experiences. And to your concerns here as well, the risk of misinformation and biased are really huge, especially once you combine that vision of Ai Generated worlds and what that could what could that need? My biggest concern when it comes to large language models, as well as like image creation models like Dali, is the potential for I mentioned this slightly in our conversation, for personal bubble filters where P will really lose reality of a...

...collective is reality to a certain extent, and don't have a shared experience anymore simply because their feed is curated for them by autog generated content that is created in personalized for them. Do you think see that as a risk in the future by any chance? I think personally that it's already happening. A friend of mine has a couple of twitter accounts in order to be able to tap into different communities and, based on this little experiment that any of us can do, you can see how different your feed experience will be like. We are basically existing in these ECO chambers already. I think we are already there. It's still not in the context of the metaphors, but it's metavers will be basically translation of the reality into the Meta. So going to Meta, I think we will face the same problems as we are facing now, unfortunately. But yeah, we just need to be better at designing the algorithms that are taking us. Maybe purposefully out of these bubbles, being able to create person life experiences but at the same time having a dose of just throwing us out of the comfort zone and experiment with that SIF it works, and keep on improving, because I don't think we have much choice, other otherwise we'll just end up in our own echo chambers, in these close communities that only are able to talk to each other and don't have the language to talk to people outside of the certain group where they are finally, Sandra and Chubam, I really enjoyed our conversation. As we're closing out, how can people access gift three and how can people read your book? So, of course our book will be out in mid July. It will be available in the in book version and it will be available in the physical paper version at the end of July. You can already actually order it via Amazon. We have released a couple of digital resources along with the book. One of them is a sandbox power advice tremment where we help users to create their own application, and we have resources on that poform get up and also on Youtube. We have also started releasing the conversations that we have mentioned here in the podcast that we had with the many different sides of the ecosystem, coming from start uppers, com founders of startups start by GP three, people that built the API, influencers in the space. We we try to talk to as many people as possible that are involved in creating this emerging ecosystem. So we are really seeing these conversations as well via youtube. You can check them out already. Just to add, we also how the TPT three club away. You can just go to GP three dot club, like premiere into get an upgradid premiere into what GP three is. How you can quickly get started with other steps to take in this simple too, to three steps is all you need to get started with GP three. And it also highlights what what do we covert in the book? So a basic allverview of what GP three is, how we can get started, how the ecosystem looks like. Uh, well, we are moving forward in the future. So that's what G B three club covers as well. All right, that is awesome, true. Bomb and Sandra, thank you so much for coming on data framed. Thank you for having us. Thank you for having us. This fund stuck too. You've been listening to data framed, a podcast by data camp. Keep connected with us by subscribing to the show in your favorite podcast player. Please give us a rating, leave a comment and share episodes you love. That helps us keep delivering insights into all things data. Thanks for listening. Until next time,.

In-Stream Audio Search

NEW

Search across all episodes within this podcast

Episodes (121)