[0:00:00] The CEO of Google DeepMind, Demis Hassabis, who has decided to cooperate with us, has actually created the artificial intelligence era we live in and is opening up the future industrial artificial intelligence era. Demis Hassabis, who won the Nobel Prize in Chemistry for being a scientist and a businessman, is now here. Let's listen to how the future of artificial intelligence, Korea, and us should face the change.
[0:00:28] Welcome, Mr. Hosavis. It's great to be here. You came to Seoul with AlphaGo and came back to Seoul after 10 years. How do you remember Seoul at that time? And how do you feel about the Seoul you're back in now? Well, Seoul and Korea has a special part in my heart, place in my heart because the match was so important. We played with Lee Sedol 10 years ago now with AlphaGo.
[0:00:56] I think in many ways it marks the beginning of the modern AI era. So it's amazing to be back here 10 years later. So much has happened, but it still feels like yesterday when the match was playing.
[0:01:22] Well, I think we're on track with that prediction for 2030. And what I hope is and why I started this journey on AI is we can use AI as the ultimate tool to advance science and medicine. So for me, that's the most profound use of AI. But of course, it's going to impact every area of life, including productivity, our programs like Gemini for productivity. So it will help in our work as well as in our everyday life.
[0:01:51] Can you give us some examples? Well, for example, I actually use Gemini to help me with planning my sleeping schedule for the jet lag. For coming from here, from the US, I've been doing a lot of traveling, so it's working pretty good so far.
[0:02:10] Recently, the AI Tako robot has beaten humans in Japan. But in fact, most artificial intelligence is not yet able to come out of the screen. So it only has a head. Then, for example, Google Gemini, when will it be able to have a body? For example, if you want to go to the stage where you can wash the laundry, cook, and replace the hard work that people do, what more process should you go through?
[0:02:36] That's one of the most exciting, I think, areas in the next few years for AI is to come out of the digital into the physical world. With Gemini, our foundation model, we made it multimodal from the beginning so it can understand images, videos, not just text. And the reason is so that it can understand the world around us because it's very important for robotics and also for assistance like on glasses or your phone that can help you in the real world and understand the context around you.
[0:03:05] So I think we're on the threshold of a really big breakthroughs on robotics using Gemini and then adding some specific robotics data on top. So we work with many top robotic companies, including Hyundai with Boston Dynamics. Yes, I think in the next couple of years, you will see some big breakthrough moments with physical AI.
[0:03:34] Robotics maybe in the factory to help with production line, maybe with automotive factories, but also eventually in the home to help with the chores and administration in the home. So I think that will be coming in the next few years. Maybe the thing I'm most excited is automated lab. So you have a research lab with many robots working on experiments to help test out hypotheses.
[0:04:03] Oh, you even do robots in the lab. Based on the success of receiving the Nobel Prize, you also founded a new drug development company. You're running it now. How much time does it take to develop a new drug that usually takes 10 years with artificial intelligence? And will the first result of the new drug development we're making now be a cancer cure?
[0:04:26] Yeah, so it was very exciting. We spun out a company called Isomorphic Labs two, three years ago to work on speeding up drug discovery by 10x.
[0:04:35] Instead of taking one decade, maybe it can take a few months, even weeks, to design a new drug. So we are building on our alpha-fold breakthrough that was understood and could predict the 3D structure of a protein. And that's one part of the drug discovery process. Of course, you also need to design the chemical compound that binds to the protein in the right place. So this is what we're doing with Isomorphic Labs. It's very exciting.
[0:05:03] We're starting with oncology and cancers and also immunology, two very big areas. But eventually, we think the platform we're building at Isomorphic, the drug design platform, can be applied to almost any area of disease.
[0:05:32] AI will help people and will be used to improve the quality of life. And there is also the aspect of raising expectations, but it is also true that it gives you anxiety right now. I'm especially worried about the job. How do you think AI will replace this job in the future?
[0:05:49] I think it's going to be a big revolution. Sometimes I say it's going to be 10 times the size of the industrial revolution in terms of impact, maybe 10 times faster. So, you know, one decade instead of one century. So it's like 100x of the industrial revolution. So there's going to be big changes, but I'm very confident in human ingenuity and adaptability. We've always adapted to previous technologies, the computer, the internet, mobile. This is the next one. It's bigger.
[0:06:18] So we'll have to see about the differences this time around. But I think there'll be many new areas, new jobs, new kinds of businesses that can be created with AI once you learn how to use the tools to the maximum level. So I think that is coming. But of course, there are also going to be some challenges about how to make sure the benefits reach everybody and all countries in the world, not just a few. Yeah.
[0:06:45] However, there is certainly a worrying aspect. It is also true that it is likely to be used militarily. How do you think the government should use AI technology?
[0:06:56] I think we have to be very thoughtful about how we deploy these technologies. The two things I worry about is misuse and misalignment. So misuse is when a bad actor or rogue state uses the AI systems for harmful ends. Maybe it's cyber attacks, something like this. So we have to build up the defensive systems first so they're much stronger than the offensive systems. That's one thing. Second thing is misalignment. So as the systems become more autonomous, as we get towards AGI in 2030,
[0:07:26] We have to make sure that these systems stick to the guardrails that we give them and make sure they do what humans want and they're beneficial to us, not go off the rails. Yes, we are trying.
[0:07:56] Do you think this is the right direction for Korea?
[0:07:59] I think Korea has so much potential to be one of the world leaders in AI. Partly it's building the AI systems themselves, maybe building on open source or converting one of the existing frontier models. But also the applications, I think, is incredible. That's what's going to come next. We mentioned robotics earlier. This should be an area, a very strong area for Korea, very strong in industrial, in manufacturing, and also robotics. So I think combining this with AI,
[0:08:28] It should be very powerful for economic gain and economic growth.
[0:08:45] Well, we are hoping to work very closely with government on many areas. So for this MOU, we're focusing on advancing science and using our AI technologies to help with like KAIST and other research institutes here in Korea to push forward many areas of science, things like material science, for example, and the weather prediction. And then we're also advancing training and education of the youth
[0:09:14] Basically use the latest AI systems in their daily life. And then finally, touching on safety, working on international standards and guardrails for secure and safe operation of these AI systems. Something that Korea has been in the forefront on and they hosted the summit last year, the AI summit last year. So we want to continue and collaborate on these topics.
[0:09:38] You said that artificial intelligence is the renaissance of science. But isn't this going to everyone, not just a certain group? How do you see this problem?
[0:09:52] Well, I think first thing is to make sure that the systems are efficient and can be distributed to everybody. So we try very hard with our Gemini systems to make very efficient smaller models that can be distributed everywhere. So everyone benefits from these amazing technologies. And we also think about that with our products, things like the Gemini app, which is very popular here in Korea to give into the hands of ordinary people.
[0:10:18] In their ordinary lives, in their everyday lives, so they can benefit from AI helping with administration tasks and things they don't want to do.
[0:10:43] Well, I think this is going to be the big challenge and interesting question for the AI world in 10 years time when we, let's say we have the technical things worked out and the economic abundance and productivity comes.
[0:10:56] First of all, we have to make sure everyone benefits from that. And then secondly, we need to think about philosophical issues like meaning and purpose. And I have so many ideas about this, but this is something that society needs to think about now. In terms of training the youth, I think what my recommendation would be for the kids of today to become incredibly proficient with these new technologies. And I think over the next five, 10 years, it will be like they have superpowers to be able to do many things, create things, games.
[0:11:26] The development of AI is very fast right now. I wonder if it's as fast as you expected, or if it's over the top. Aren't you afraid that it's too fast?
[0:11:52] I think we're kind of on track if you think of the predictions I made back 2016 and even earlier, 2010, when we started DeepMind, very early in the AI journey. But we sort of thought about it as a 20-year mission when we started in 2010. So I think we are on track.
[0:12:10] In order to guarantee a prosperous future,
[0:12:37] So I think the way we're going to get that kind of world of abundance is to use AI again for science in areas like energy sources and materials. So you could imagine future AI systems helping us to crack fusion.
[0:12:58] Or maybe superconductors, new types of superconductors, better batteries. So there are many different ways I think AI could help with material science and cutting edge physics to give us free, renewable, clean energy. So I think that would be the number one thing that will then start this new era of abundance.
[0:13:27] What do people like me need to prepare to live in the era of AI for decades?
[0:13:35] I think this is, again, to use all of these new tools to kind of enhance your productivity. So I would recommend that in the first place. But I think also it's for our generation to think about these philosophical questions and help to guide society from all different walks of society, not just the technologists, but also other areas of society, academics, humanities, about what should the next era look like. Because it will be a new era.
[0:14:04] Is there a specific clinical trial period in the development of AI medicine?
[0:14:27] We are starting with preclinical at the moment. We're testing a few of our early stage drugs. We're working actually on nearly 20 different drug programs already, which is amazing for a company that's very young still. But that shows you, I think, the potential of our platform. One day, I hope we can be designing hundreds of drugs at the same time, which is kind of unheard of today. It seems science fiction.
[0:14:57] But I think in the next 5 to 10 years, I think this will become possible with the help of AI.
[0:15:14] I think Korea has many amazing industries that will be very relevant for the AI era. So we have many partners actually, of course, here in Korea. Samsung on the mobile and with Gemini. We work with Hyundai on the self-driving cars and also robotics, as I mentioned earlier, with Boston Dynamics. And then, of course, you have incredible chip manufacturers like SK Hynix, which are all important.
[0:15:44] for the chips that power the AI revolution. So I think Korea is incredibly well positioned to take advantage of an AI first future industries. So that's why I'm very excited to be here. And over the next couple of days, I'll be having many business meetings too with our partners in the country, our very important partners. So I think Korea is on a good track to be at the forefront of the AI revolution and should embrace it further with things like robotics.
[0:16:13] You chose Korea as an AI partner. Can you tell us again why you chose Korea?
[0:16:28] I think because Korea has, first of all, so much talent, technical talent, being trained here in the top universities. But also as a society, I think that Korea is very positive about new technologies in general. Of course, evidenced by our original AlphaGo match, which is one of the reasons we had it here in Seoul. And then because of all the talent and the big companies that work on different aspects that are important for
[0:16:58] AI has to exist for humans. But humans themselves don't know what they want. What should AI do to know what humans want and control them?
[0:17:23] I think that's one of the things AI can help with in our personal lives is be like a trusted assistant, maybe to give advice, to help in your daily life, give you some information to make a better decision. So this is what I hope Gemini and this kind of assistant technology will do for us in our daily lives to improve our everyday life.
[0:17:53] Well, you know, I think it's fantastic, our history with Korea. I love this place and its place in the AI story with things like AlphaGo. So I think Korea, I expect it to be in the forefront of this new AI revolution and the exciting world we're going to. Thank you.