{"id":1291,"date":"2019-06-17T21:37:10","date_gmt":"2019-06-17T21:37:10","guid":{"rendered":"http:\/\/blogs.harvard.edu\/toshietakahashi\/?p=1291"},"modified":"2019-06-17T21:42:05","modified_gmt":"2019-06-17T21:42:05","slug":"interview-with-yoshua-bengio","status":"publish","type":"post","link":"https:\/\/archive.blogs.harvard.edu\/toshietakahashi\/2019\/06\/17\/interview-with-yoshua-bengio\/","title":{"rendered":"Interview with Yoshua Bengio, Pioneer of AI"},"content":{"rendered":"
\"\"<\/a>

Prof. Yoshua Bengio at the MILA<\/p><\/div>\n

On June 7th<\/sup>, 2019 at the MILA (Montreal Institute for Learning Algorithms)<\/a> in Montreal, Canada, I conducted my interview with Professor Yoshua Bengio, who is one of the pioneers of AI (Artificial Intelligence).\u00a0 He is well-known as the \u201cfather of AI\u201d for his great contribution to developing so-called deep learning.\u00a0 He has received the 2018 ACM A.M. Turing Award <\/a>with Geoffrey Hinton and Yann LeCun for major breakthroughs in AI.<\/span><\/p>\n

In my interview, I asked him about the possibilities of AGI (Artificial General Intelligence), biased data, people\u2019s concerns about GAFA (Google, Amazon, Facebook, Apple) and China, the opportunities and risks of AI and the future of AI.\u00a0 All these questions are based on my previous experiences in the University of Cambridge as well as many international summits and conferences on AI I have been invited to recently.\u00a0 <\/span><\/p>\n

Bengio is also noteworthy because he chooses to remain as an academic, staying at the University of Montreal<\/a> as head of the MILA, while other AI leaders such as Geoffrey Hinton have left academia and now work for Google.\u00a0 Bengio continues to contribute to teaching students as well as engaging with local communities. He believes the education of future generations and people\u2019s engagement with AI is crucial for the creation of a better society including AI.\u00a0This is because he is aware of not only the opportunities but also the risks of AI. \u00a0As he owns his startup, so-called Element AI<\/a>, he is instrumental in building a bridge between academia and the business world. \u00a0<\/span><\/p>\n

This is my interview with Yoshua.<\/span><\/p>\n

The Road to AGI <\/span><\/b><\/h3>\n

Yoshua Bengio\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Did you have some questions for me?<\/span><\/p>\n

Toshie Takahashi\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes, of course.\u00a0 Thank you for taking the time.\u00a0 I’d like to ask you about AGI.<\/span><\/p>\n

YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Okay.<\/span><\/p>\n

TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 I watched some of your videos and I understand you are very positive about AGI.<\/span><\/p>\n

YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 No.<\/span><\/p>\n

TT \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 No? I thought you showed a…<\/span><\/p>\n

YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 I’m positive that we can build machines as intelligent as humans, but completely general intelligence is a different story. I’m not positive as to how humans might use it because we’re not very wise. <\/span><\/p>\n

TT \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 Okay. So can you show a road map of how you could create AGI?<\/span><\/p>\n

YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes, the one I have chosen to explore.<\/span><\/p>\n

TT\u00a0 \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 I spent some time in Cambridge, and some scholars, for example Professor John Daugman, the head of Artificial Intelligence Group at the University of Cambridge, said that AGI is an illusion created by science fiction because he said that we don’t even understand a single neuron so how can we create AGI?<\/span><\/p>\n

YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes, I disagree with him.<\/span>\u00a0<\/span><\/p>\n

TT \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 Okay, so could you tell me about that?<\/span><\/p>\n

YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Sure. Having worked for decades on AI and machine learning I feel strongly that we have made very substantial progress, and in particular we have uncovered some principles, which today allow us to build very powerful systems. I also recognise that there’s a long way towards human level AI, and I don’t know how long it’s going to take. So I didn’t say we’ll find human level AI in five years, or ten years or 50 years. I don’t know how much time it’s going to take, but the human brain is a machine. It’s a very complex one and we don\u2019t fully understand it, but there\u2019s no reason to believe that we won\u2019t be able to figure out those principles. <\/span><\/p>\n

TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 I see. Mr. Tom Everitt at DeepMind said that he could create AGI in a couple of decades, maybe 20 or 30 years.\u00a0 Not too far in the future.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 How does he know?<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 I don’t know. I asked him but he didn’t answer it.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Nobody knows.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Nobody knows. Yes, of course. When I met Professor Sheldon Lee Glashow, a Nobel Prize winning American theoretical physicist, he told us that we won\u2019t have AGI.\u00a0Or even if we have it, it\u2019d be very far away.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Possibly, so we don’t know. It could be ten years, it could be 100 years. <\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Oh really?<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes. <\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Okay.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 It’s impossible to know these things. There\u2019s a beautiful analogy that I heard my friend Yann LeCun mention first. As a researcher, our progress is like climbing a mountain and as we approach the peak of that mountain we realise there’s some other mountains behind.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes, exactly.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 And we don\u2019t know what other higher peak is hidden from our view right now. <\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 I see.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 So it might be that the obstacles that we’re currently working on are going to be the last ones to reach human level AI or maybe there will be ten more big challenges that we don\u2019t even perceive right now, so I don’t think it’s plausible that we could really know when, how many years, how many decades, it will take to reach human level AI.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 I see.\u00a0 But some people also say that we need a different breakthrough to create AGI. We need a kind of paradigm shift from our current approach. \u00a0Do you think that you can see the road to reach if you keep on with deep learning? So this is a right road?<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 We have understood as I said some very important principles through our work on deep learning and I believe those principles are here to stay, but we need additional advances that are going to be combined with things we have already figured out. I think deep learning is here to stay, but as is, it\u2019s obviously not sufficient to do, for example, higher-level cognition that humans are doing. We’ve made a lot of progress on what psychologists call System 1 cognition, which is everything to do with intuitive tasks. Here is an example of what we’ve discovered, in fact one of the central ideas in deep learning: the notion of distributed representation. I\u2019m very, very sure that this notion will stay because it’s so powerful. <\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Wonderful! I’m happy to hear that.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes.<\/span>
\n<\/p>\n

The Future of AI <\/span><\/b><\/h3>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 You are really ahead of this…So what can you see of the future?<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 I don\u2019t see the future.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Really? I thought that you could see a future we cannot see yet.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 No, I have research goals and I have chosen to explore particular directions because I believe in them, so the vision I have is that a big missing piece in our current machine learning systems is what people call common sense. So think about the intelligence of a two-year-old or the intelligence of a cat, we don’t even have that in machines right now. And that’s not even starting to think about things like language where we’re seriously lacking, especially when you look at the mistakes made by current state-of-the-art systems. That sort of common sense understanding includes things that current machine learning doesn’t do, like understanding cause and effect, and discovering cause and effect. It includes a broad understanding of the world, not just one specialised task. It includes the ability to discover this model of the world through unsupervised exploration. We rely today heavily on supervised learning where all of the high-level concepts have been defined by a human teacher or human labels. There are lots of aspects of intelligence that are currently at the frontier, and I’m not the only one exploring these things, which could make a big difference in a few years, but it’s hard to be sure.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Is your goal to create a human or a superhuman?<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 No, my goal is to understand general principles of intelligence, how an agent can become intelligent. I and many others would like to discover the equivalent of the laws of physics, but for intelligence.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 And presumably those principles would apply to humans, to animals, to aliens, who might be intelligent, to machines that we can build, so these would be very general principles and machine learning has already established some of those principles, but we’re still missing some important ones, I believe. <\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 But you’ve found a simple principle?<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes, several. But behind this, there’s a meta-principle, a scientific hypothesis, that intelligence could be explained by a few simple principles. We don’t know if that hypothesis is true, but the success of deep learning in the last few years is a good validation of that hypothesis. It’s consistent with that hypothesis because deep learning is built on a few very simple principles. Most of the complexity of the systems that are trained with deep learning is not in the learning mechanismsp; it’s in the data. The data contains the overwhelming share of the information in a current trained AI system, while only a little bit of information, relatively speaking, is in those principles, which are like the learning procedures.<\/span>\u00a0<\/span><\/p>\n

Biased Data and AI for Humanity<\/span><\/b><\/h3>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 You said that the intelligence is from knowledge and knowledge is acquired from data?<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 That’s right.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 But if data is biased, what happens?\u00a0Some social scientists criticize most data for being based on male, Caucasian middle aged… <\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 That\u2019s right. <\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 So if a young woman of colour applies, for instance, for an insurance policy, AI might say no because they don\u2019t have enough data for those applicants.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes, absolutely. I think there are technical solutions and social solutions to this problem. We have to change our social norms, for example, so that companies building products use technological solutions and logistical solutions, for example, in the way that the data is collected, in the way that it’s described and managed, and in the particular learning algorithms that are used because we know techniques that can mitigate the bias and discrimination. So we can probably include those techniques, but more importantly we need to make sure that companies and governments use them. <\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Is that why you think it’s important that both social scientists and natural scientists work for AI together?<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 I love the idea of \u201cAI for humanity\u201d as you have in the Mila here.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Right, because the AI researcher might not realise some of the social issues that could be involved in the deployment. I think it’s particularly important for people who are doing research or development of products that is close to something that people will use, in large-scale deployment for example.<\/span>
\n<\/p>\n

People\u2019s Concerns about GAFA and China<\/span><\/b><\/h3>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 People are concerned about big companies such as Google or Apple, or a country like China because they have a huge amount of data so they can do what they want.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes, but they probably also want to be considered as positive, responsible agents in society, and so if we do the right communication, to explain the issues, and engage in social discussion about these things I’m optimistic that we can have those social norms be improved. And of course, it means changing how to do things.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 I met Dan Klein, chief data officer from Valtech in Cambridge, UK. He is also concerned about China.\u00a0 Because China has a huge amount of data which computer scientists can use in order to develop the AI, while UK and EU have the limited data access because of the Data Protection Act.\u00a0 Also Chinese companies pay high salaries for computer engineers outside of China, so great European engineers are going to leave for China.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 I don’t think our European engineers are going to China.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Oh really? Maybe American?\u00a0 Oh, I don’t know.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Not much. No but they don’t need that. They have plenty of good scientists and engineers. The issue with China is that it’s difficult for many of us to have confidence that the current political system of China will behave responsibly, but it’s true of many countries that governments are not very responsible. If you consider, for example, climate change, the US has been behaving very badly. <\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes. <\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 So even if it’s a democracy it doesn’t mean that governments will be doing the right things. So I think every country has an interest in being part of the global consensus for obvious economic reasons, but also, to feel good about themselves. So I think we should not pit countries against each other and peoples against each other.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes.\u00a0 I totally agree with you.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 It\u2019s not going to help.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 If you develop an algorithm in Canada, would it also work in Japan or should we adapt it to our society?<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 No the algorithms are very generic. It\u2019s like math. \u00a0Addition is the same in Japan as in Canada.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes, but data itself has some cultural meanings\u2026<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 No, but that’s data. That’s not the algorithm. <\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Okay. I see. <\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 So the learning procedures are going to be the same, but the data will be different, and the systems that are trained using the learning procedures and the data, of course, will be different in different countries.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 So once we have the algorithm, we can use it with our own data.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 That\u2019s right.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Then it will work very well. <\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 It works better.<\/span>
\n<\/p>\n

The Opportunities and Risks of AI<\/span><\/b><\/h3>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Okay, wonderful. You might be getting tired of talking about the opportunities and risks of AI because people always like to ask you as one of the leading experts on AI.\u00a0 But I also think it\u2019s very important for us to understand both.\u00a0 Then we can maximize opportunities and minimize risks in order to get social benefit from AI.\u00a0 So what are the biggest opportunities and the biggest risks? I know, there are many, many risks, but from your point of view, what are you concerned about the most?<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 So with AI in terms of opportunities, I think they’re a huge potential for social good, in healthcare, in the environment, fighting climate change, which is a very important question for the planet. Maybe a little bit further down the road in education as well. And on the risk side, I think the biggest risk really is a threat to democracy, a threat to the stability of our social fabric because of things like killer drones, because of things like political advertising and the influence that one can buy on social networks, because of things like concentration of power, in a few hands, in a few people, a few companies, a few countries, and because of potential social unrest that could come from rapid automation. So all of these could be disruptive to society and we have to be careful where we cross the red line between what is acceptable and what is not acceptable in the applications of AI.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Who determines the red line?<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 It’s a very good question. Humans define their social norms through a global discussion, and in different countries it might be different types of people. Scholars usually have more impact on the result, and scientists, I think, should be part of the discussion, but regular citizens should be part of the discussion as well. <\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes, I agree.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 At the end of the day, in decent democratic countries it’s going to be democratic decisions where we put those lines. Where I think it’s trickier is that many of these decisions cannot be taken in isolation in each country. There has to be global international coordination.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes, definitely. I met Mr. Irakli Beridze, the head of Centre for Artificial intelligence and Robots, the United Nations Interregional Crime and Justice Research Institute (UNICRI), and he said that he goes to Russia, Syria and other countries because the governments of those countries also have to cooperate. <\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Right. Yes. It’s very important.<\/span><\/p>\n

\u00a0<\/span>TT\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 But it\u2019d be very difficult. <\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes, unfortunately, we don’t have a good international coordination framework. The UN is very weak.<\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Oh really?<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Oh yes. It doesn’t have any power. <\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Really? I thought it had power. No?<\/span><\/p>\n

\u00a0<\/span>YB\u00a0 \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 No, the UN doesn’t have nearly enough power. One issue I’m a little bit more familiar with is the killer robots and lethal autonomist weapons. The Secretary General has been saying for a while now, this is both morally repugnant and dangerous for global security, but the problem is that a lot of the decisions in UN decision-making committees and treaties happen by complete consensus. So if just one country in the committee says no, there’s no treaty. <\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 That cannot work. The problem is that individual countries have been too scared of losing some sovereignty, some power, to a higher level, which would be, for example, international government, but we have to do that otherwise we will not solve the climate change problem. We will not solve fiscal issues across the planet. We will not prevent dangers from misuse of AI. So there are lots of issues for which we have to have global coordination. <\/span><\/p>\n

\u00a0<\/span>TT \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Yes. <\/span>I also believe we need more discussions about living with AI both locally and globally. \u00a0<\/span>Thank you very much.<\/span><\/p>\n

\u00a0<\/span>YB \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 You’re welcome.<\/span><\/p>\n

\u00a0<\/span><\/p>\n

Acknowledgement<\/span><\/b><\/p>\n

I would like to express my gratitude to Myriam C\u00f4t\u00e9, director of AI for Humanity at the Mila for her kind invitation and great support on my cross-cultural research on AI for good. \u00a0This research is supported by the KDDI Foundation.<\/span><\/p>\n

\u00a0<\/span><\/p>\n","protected":false},"excerpt":{"rendered":"

On June 7th, 2019 at the MILA (Montreal Institute for Learning Algorithms) in Montreal, Canada, I conducted my interview with Professor Yoshua Bengio, who is one of the pioneers of AI (Artificial Intelligence).\u00a0 He is well-known as the \u201cfather of … Continue reading →<\/span><\/a><\/p>\n","protected":false},"author":2464,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[117681,13076,117682,117676,117657,117678,134,117677,3022,23092,117668],"tags":[],"class_list":["post-1291","post","type-post","status-publish","format-standard","hentry","category-agi","category-ai","category-biased-data","category-cross-disciplinary-approach","category-culture-and-communication","category-deep-learning","category-education","category-future-society","category-opportunities","category-risks","category-robot"],"jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/archive.blogs.harvard.edu\/toshietakahashi\/wp-json\/wp\/v2\/posts\/1291","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/archive.blogs.harvard.edu\/toshietakahashi\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/archive.blogs.harvard.edu\/toshietakahashi\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/archive.blogs.harvard.edu\/toshietakahashi\/wp-json\/wp\/v2\/users\/2464"}],"replies":[{"embeddable":true,"href":"https:\/\/archive.blogs.harvard.edu\/toshietakahashi\/wp-json\/wp\/v2\/comments?post=1291"}],"version-history":[{"count":14,"href":"https:\/\/archive.blogs.harvard.edu\/toshietakahashi\/wp-json\/wp\/v2\/posts\/1291\/revisions"}],"predecessor-version":[{"id":1306,"href":"https:\/\/archive.blogs.harvard.edu\/toshietakahashi\/wp-json\/wp\/v2\/posts\/1291\/revisions\/1306"}],"wp:attachment":[{"href":"https:\/\/archive.blogs.harvard.edu\/toshietakahashi\/wp-json\/wp\/v2\/media?parent=1291"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/archive.blogs.harvard.edu\/toshietakahashi\/wp-json\/wp\/v2\/categories?post=1291"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/archive.blogs.harvard.edu\/toshietakahashi\/wp-json\/wp\/v2\/tags?post=1291"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}