github-actions[bot]
commented
2 years ago Language: english
Transcription: The Rise of AI When I think about the rise of AI, I'm reminded by the rise of literacy. A few hundred years ago, many people in society thought that maybe not everyone needed to be able to read and write. Back then, many people were tending fields or herding sheep, so maybe there was less need for written communication, and all that was needed was for the high priests and priestesses and monks to be able to read the holy book, and the rest of us could just go to the temple or church or the holy building and sit and listen to the high priests and priestesses read to us. Fortunately, we've since figured out that we can build a much richer society if lots of people can read and write. Today, AI is in the hands of the high priests and priestesses. These are the highly skilled AI engineers, many of whom work in the big tech companies. And most people have access only to the AI that they built for them. I think that we can build a much richer society if we can enable everyone to help to write the future. But why is AI largely concentrated in the big tech companies? Because many of these AI projects have been expensive to build. They may require dozens of highly skilled engineers, and it may cost millions or tens of millions of dollars to build an AI system. And the large tech companies, particularly the ones with hundreds of millions or even billions of users, have been better than anyone else at making these investments pay off, because for them, a one-size-fits-all AI system, such as one that improves web search or that recommends better products for online shopping, can be applied to this very large number of users to generate a massive amount of revenue. But this recipe for AI does not work once you go outside the tech and internet sectors to other places where, for the most part, there are hardly any projects that apply to 100 million people or that generate comparable economics. Let me illustrate an example. Many weekends, I drive a few minutes from my house to a local pizza store to buy a slice of Hawaiian pizza from the gentleman that owns his pizza store. And his pizza is great, but he always has a lot of cold pizza sitting around, and every weekend, some different flavor of pizza is out of stock. But when I watch him operate his store, I get excited, because by selling pizza, he is generating data, and this is data that he can take advantage of if he had access to AI. AI systems are good at spotting patterns when given access to the right data, and perhaps an AI system could spot if Mediterranean pizzas sell really well on a Friday night, maybe could suggest to him to make more of it on a Friday afternoon. Now, you might say to me, hey, Andrew, this is a small pizza store, what's the big deal? And I say to the gentleman that owns this pizza store, something that could help him improve his revenues by a few thousand dollars a year, that would be a huge deal to him. I know that there is a lot of hype about AI's need for massive datasets, and having more data does help, but contrary to the hype, AI can often work just fine, even on modest amounts of data, such as the data generated by a single pizza store. So the real problem is not that there isn't enough data from the pizza store. The real problem is that the small pizza store could never serve enough customers to justify the cost of hiring an AI team. I know that in the United States, there are about half a million independent restaurants, and collectively, these restaurants do serve tens of millions of customers, but every restaurant is different, with a different menu, different customers, different ways of recording sales that no one-size-fits-all AI will work for all of them. What would it be like if we could enable small businesses, and especially local businesses, to use AI? Let's take a look at what it might look like at a company that makes and sells T-shirts. I would love if an accountant working for the T-shirt company could use AI for demand forecasting, say, figure out what funny memes they print on T-shirts that would drive sales by looking at what's trending on social media. Or for product placement, why can't a front-of-store manager take pictures of what the store looks like and show it to an AI, and have an AI recommend where to place products to improve sales? Supply chain. Can an AI recommend to a buyer whether or not they should pay $20 per yard for a piece of fabric now, or if they should keep looking because they might be able to find it cheaper elsewhere? Or quality control. A quality inspector should be able to use AI to automatically scan pictures of the fabric being used to make T-shirts to check if there are any tears or discolourations in the cloth. Today, large tech companies routinely use AI to solve problems like these, and to great effect. But a typical T-shirt company, or a typical auto mechanic or retailer or school or local farm, will be using AI for exactly zero of these applications today. Every T-shirt maker is sufficiently different from every other T-shirt maker that there is no one-size-fits-all AI that will work for all of them. And in fact, once you go outside the internet and tech sectors, in other industries, even large companies, such as the pharmaceutical companies, the car makers, the hospitals, also struggle with this. This is the long-tail problem of AI. If you were to take all current and potential AI projects and sort them in decreasing order of value and plot them, you get a graph that looks like this. Maybe the single most valuable AI system is something that decides what ads to show people on the internet. Maybe the second most valuable is a web search engine. Maybe the third most valuable is an online shopping product recommendation system. But when you go to the right of this curve, you then get projects like T-shirt product placements or T-shirt demand forecasting or pizzeria demand forecasting. And each of these is a unique project that needs to be custom-built. Even T-shirt demand forecasting, if it depends on trending memes on social media, is a very different project than pizzeria demand forecasting, if that depends on the pizzeria sales data. So today, there are millions of projects sitting on the tail of this distribution that no one is working on, but whose aggregate value is massive. So how can we enable small businesses and individuals to build AI systems that matter to them? For most of the last few decades, if you want to build an AI system, this is what you have to do. You have to write pages and pages of code. And while I would love for everyone to learn to code, and in fact, all of us, I would love for everyone to learn to code, and in fact, online education and also offline education are helping more people than ever learn to code, unfortunately, not everyone has the time to do this. But there is an emerging new way to build AI systems that will let more people participate. Just as pen and paper, which are a vastly superior technology to stone, tablet and chisel, were instrumental to widespread literacy, there are emerging new AI development platforms that shift the focus from asking you to write lots of code to asking you to focus on providing data, and this turns out to be much easier for a lot of people to do. Today, there are multiple companies working on platforms like these. Let me illustrate a few of the concepts using one that my team has been building. Take the example of an inspector wanting AI to help detect defects in fabric. An inspector can take pictures of the fabric and upload it to a platform like this, and they can go in to show the AI what tears in the fabric look like by drawing rectangles, and they can also go in to show the AI what discolorations in the fabric look like by drawing rectangles. So these pictures, together with the green and pink rectangles that the inspector's drawn, are data created by the inspector to explain to AI how to find tears and discolorations. After the AI examines this data, we may find that it has seen enough pictures of tears but not yet enough pictures of discolorations. This is akin to if a junior inspector had learned to reliably spot tears but still needs to further hone their judgment about discolorations. So the inspector can go back and take more pictures of discolorations to show to the AI to help it deepen its understanding. By adjusting the data you give to the AI, you can help the AI get smarter. So an inspector using an accessible platform like this can, in a few hours to a few days, and with purchasing a suitable camera setup, be able to build a custom AI system to detect defects, tears and discolorations in all the fabric being used to make T-shirts throughout the factory. And once again, you may ask, you may say, hey, Andrew, this is one factory, why is this a big deal? And I say to you, this is a big deal to that inspector whose life this makes easier. And equally, this type of technology can empower a baker to use AI to check for the quality of the cakes they're making, or an organic farmer to check the quality of the vegetables, or a furniture maker to check the quality of the wood they're using. Platforms like these will probably still need a few more years before they're easy enough to use for every pizzeria owner, but many of these platforms are coming along, and some of them are getting to be quite usable to someone that is tech-savvy today with just a bit of training. But what this means is that rather than relying on the high-precise and precess to write AI systems for everyone else, we can start to empower every accountant, every store manager, every buyer and every quality inspector to build their own AI systems. I hope that the pizzeria owner and many other small business owners like him will also take advantage of this technology, because AI is creating tremendous wealth and will continue to create tremendous wealth. And it's only by democratizing access to AI that we can ensure that this wealth is spread far and wide across society. Hundreds of years ago, I think hardly anyone understood the impact that widespread literacy will have. Today, I think hardly anyone understands the impact that democratizing access to AI will have. Building AI systems has been out of reach for most people, but that does not have to be the case. In the coming era for AI, we'll empower everyone to build AI systems for themselves, and I think that will be an incredibly exciting future. Thank you very much. Thank you. Thank you. Thank you.
Translation: The Rise of AI When I think about the rise of AI, I'm reminded by the rise of literacy. A few hundred years ago, many people in society thought that maybe not everyone needed to be able to read and write. Back then, many people were tending fields or herding sheep, so maybe there was less need for written communication, and all that was needed was for the high priests and priestesses and monks to be able to read the holy book, and the rest of us could just go to the temple or church or the holy building and sit and listen to the high priests and priestesses read to us. Fortunately, we've since figured out that we can build a much richer society if lots of people can read and write. Today, AI is in the hands of the high priests and priestesses. These are the highly skilled AI engineers, many of whom work in the big tech companies. And most people have access only to the AI that they built for them. I think that we can build a much richer society if we can enable everyone to help to write the future. But why is AI largely concentrated in the big tech companies? Because many of these AI projects have been expensive to build. They may require dozens of highly skilled engineers, and it may cost millions or tens of millions of dollars to build an AI system. And the large tech companies, particularly the ones with hundreds of millions or even billions of users, have been better than anyone else at making these investments pay off, because for them, a one-size-fits-all AI system, such as one that improves web search or that recommends better products for online shopping, can be applied to this very large number of users to generate a massive amount of revenue. But this recipe for AI does not work once you go outside the tech and internet sectors to other places where, for the most part, there are hardly any projects that apply to 100 million people or that generate comparable economics. Let me illustrate an example. Many weekends, I drive a few minutes from my house to a local pizza store to buy a slice of Hawaiian pizza from the gentleman that owns his pizza store. And his pizza is great, but he always has a lot of cold pizza sitting around, and every weekend, some different flavor of pizza is out of stock. But when I watch him operate his store, I get excited, because by selling pizza, he is generating data, and this is data that he can take advantage of if he had access to AI. AI systems are good at spotting patterns when given access to the right data, and perhaps an AI system could spot if Mediterranean pizzas sell really well on a Friday night, maybe could suggest to him to make more of it on a Friday afternoon. Now, you might say to me, hey, Andrew, this is a small pizza store, what's the big deal? And I say to the gentleman that owns this pizza store, something that could help him improve his revenues by a few thousand dollars a year, that would be a huge deal to him. I know that there is a lot of hype about AI's need for massive datasets, and having more data does help, but contrary to the hype, AI can often work just fine, even on modest amounts of data, such as the data generated by a single pizza store. So the real problem is not that there isn't enough data from the pizza store. The real problem is that the small pizza store could never serve enough customers to justify the cost of hiring an AI team. I know that in the United States, there are about half a million independent restaurants, and collectively, these restaurants do serve tens of millions of customers, but every restaurant is different, with a different menu, different customers, different ways of recording sales that no one-size-fits-all AI will work for all of them. What would it be like if we could enable small businesses, and especially local businesses, to use AI? Let's take a look at what it might look like at a company that makes and sells T-shirts. I would love if an accountant working for the T-shirt company could use AI for demand forecasting, say, figure out what funny memes they print on T-shirts that would drive sales by looking at what's trending on social media. Or for product placement, why can't a front-of-store manager take pictures of what the store looks like and show it to an AI, and have an AI recommend where to place products to improve sales? Supply chain. Can an AI recommend to a buyer whether or not they should pay $20 per yard for a piece of fabric now, or if they should keep looking because they might be able to find it cheaper elsewhere? Or quality control. A quality inspector should be able to use AI to automatically scan pictures of the fabric being used to make T-shirts to check if there are any tears or discolourations in the cloth. Today, large tech companies routinely use AI to solve problems like these, and to great effect. But a typical T-shirt company, or a typical auto mechanic or retailer or school or local farm, will be using AI for exactly zero of these applications today. Every T-shirt maker is sufficiently different from every other T-shirt maker that there is no one-size-fits-all AI that will work for all of them. And in fact, once you go outside the internet and tech sectors, in other industries, even large companies, such as the pharmaceutical companies, the car makers, the hospitals, also struggle with this. This is the long-tail problem of AI. If you were to take all current and potential AI projects and sort them in decreasing order of value and plot them, you get a graph that looks like this. Maybe the single most valuable AI system is something that decides what ads to show people on the internet. Maybe the second most valuable is a web search engine. Maybe the third most valuable is an online shopping product recommendation system. But when you go to the right of this curve, you then get projects like T-shirt product placements or T-shirt demand forecasting or pizzeria demand forecasting. And each of these is a unique project that needs to be custom-built. Even T-shirt demand forecasting, if it depends on trending memes on social media, is a very different project than pizzeria demand forecasting, if that depends on the pizzeria sales data. So today, there are millions of projects sitting on the tail of this distribution that no one is working on, but whose aggregate value is massive. So how can we enable small businesses and individuals to build AI systems that matter to them? For most of the last few decades, if you want to build an AI system, this is what you have to do. You have to write pages and pages of code. And while I would love for everyone to learn to code, and in fact, all of us, I would love for everyone to learn to code, and in fact, online education and also offline education are helping more people than ever learn to code, unfortunately, not everyone has the time to do this. But there is an emerging new way to build AI systems that will let more people participate. Just as pen and paper, which are a vastly superior technology to stone, tablet and chisel, were instrumental to widespread literacy, there are emerging new AI development platforms that shift the focus from asking you to write lots of code to asking you to focus on providing data, and this turns out to be much easier for a lot of people to do. Today, there are multiple companies working on platforms like these. Let me illustrate a few of the concepts using one that my team has been building. Take the example of an inspector wanting AI to help detect defects in fabric. An inspector can take pictures of the fabric and upload it to a platform like this, and they can go in to show the AI what tears in the fabric look like by drawing rectangles, and they can also go in to show the AI what discolorations in the fabric look like by drawing rectangles. So these pictures, together with the green and pink rectangles that the inspector's drawn, are data created by the inspector to explain to AI how to find tears and discolorations. After the AI examines this data, we may find that it has seen enough pictures of tears but not yet enough pictures of discolorations. This is akin to if a junior inspector had learned to reliably spot tears but still needs to further hone their judgment about discolorations. So the inspector can go back and take more pictures of discolorations to show to the AI to help it deepen its understanding. By adjusting the data you give to the AI, you can help the AI get smarter. So an inspector using an accessible platform like this can, in a few hours to a few days, and with purchasing a suitable camera setup, be able to build a custom AI system to detect defects, tears and discolorations in all the fabric being used to make T-shirts throughout the factory. And once again, you may ask, you may say, hey, Andrew, this is one factory, why is this a big deal? And I say to you, this is a big deal to that inspector whose life this makes easier. And equally, this type of technology can empower a baker to use AI to check for the quality of the cakes they're making, or an organic farmer to check the quality of the vegetables, or a furniture maker to check the quality of the wood they're using. Platforms like these will probably still need a few more years before they're easy enough to use for every pizzeria owner, but many of these platforms are coming along, and some of them are getting to be quite usable to someone that is tech-savvy today with just a bit of training. But what this means is that rather than relying on the high-precise and precess to write AI systems for everyone else, we can start to empower every accountant, every store manager, every buyer and every quality inspector to build their own AI systems. I hope that the pizzeria owner and many other small business owners like him will also take advantage of this technology, because AI is creating tremendous wealth and will continue to create tremendous wealth. And it's only by democratizing access to AI that we can ensure that this wealth is spread far and wide across society. Hundreds of years ago, I think hardly anyone understood the impact that widespread literacy will have. Today, I think hardly anyone understands the impact that democratizing access to AI will have. Building AI systems has been out of reach for most people, but that does not have to be the case. In the coming era for AI, we'll empower everyone to build AI systems for themselves, and I think that will be an incredibly exciting future. Thank you very much. Thank you. Thank you. Thank you.
URL
https://www.youtube.com/watch?v=reUZRyXxUs4