Live on Bloomberg TV

00:00I had a chance to catch up with the Godfather of A.I., the Nobel Prize winning physicist Geoffrey Hinton, now professor emeritus over at the University of Toronto. We talked about the meteoric rise of the technology that he helped pave the way for and, well, some of the warnings that he's giving folks for the future. Take a listen. There's a bunch of people, mainly former academics, worried about safety. So many of the real experts in this stuff really believe we're getting superintelligence coming. And it's very scary unless we do it very, very carefully. You. Almost a year ago, I believe, October of last year. Of course, you were awarded the Nobel Prize largely for your work that effectively laid the foundation for a lot of where we are today when it comes to artificial intelligence. When you look at the general models that are being used and I know that don't necessarily mean the technological models, but the business models that are being used by various companies. Do you see some of those companies as being more responsible actors than others? Yes. First, I want to correct something. There were quite a large number of people contributed to artificial intelligence. I was just one of them. I think some companies are more ethical than others. Anthropic was set up to be more ethical than open air. It was set up by people who left open AI because they didn't believe open. I was living up to its mandate to develop it with safety is the first principle. Anthropic is very concerned with safety. DeepMind, which develops Germany for Google, is fairly concerned with safety. Demis Hassabis, who runs DeepMind, has been concerned with safety for a long time, and he's very well aware of the issues of how dangerous a superintelligence could be. Some of the other companies are much less ethical, in my opinion. In my opinion matter. And you know, most company X, I seem to be much less concerned with developing stuff safely, even though people like Elon Musk do understand the risks. When we talk about those potential risk. I am curious about your past relationship with Google and what is effectively now. DeepMind, you sold your company to them more than a decade ago. However, you also left quite some time ago as well. Why? Okay, so I didn't leave in order to criticize Google. I got along very well with Google while I was there. I believe they were operating about as ethically as a company could. For example, they developed these large chatbots before anybody else. The ones that use Transformers in 2007 developed in 2017, and by a few years later, they had very good chatbots. They didn't release them to the public. But when open I made a deal with Microsoft and they were released to the public for search for being open. AI Sorry, Google had no alternative but to compete and make them public. I am curious about your thoughts on the reliability of a lot of a lot of what we're seeing with regards to AI and specifically the tech side of it. Obviously, for the consumer and for a lot of us, our first sort of exposure to air through chat bots, there've been a lot of discussions about whether those chat bots are actually delivering us real information or whether they're just simply reinforcing our own personal thoughts and beliefs as we sort of enter prompts and whether they are truly thinking for themselves. Can we rely on what they're telling us? No, you can't rely on what they're telling you any more than you come for a person. So one of the problems with chat bots is what was called hallucinations for a language model. They should be called confabulation. And it's been studied by psychologists for a long time in people. So we know that when people remember things that happened some time ago, they actually make up a lot of details and they're perfectly confident in these details, even though the details are wrong. You don't often see that. But it was very clear, for example, when John Dean testified in the Watergate trials, he didn't know there were tapes of what went on in the Oval Office, and he confidently described meetings in the Oval Office that never happened. And it doesn't look like he was lying. It looks like he was trying to tell the truth. He just said what seemed plausible to him retrospectively, given his experience in the Oval Office. And he characterized what went on very well by making up little scenarios that were typical of what happened but weren't actually what happened. Now, Chappell's through that, too. And the fact that they do that just makes them more like people, not less like people. Would you like to see a little bit more coordination amongst the world governments, whether it's through a body like the U.N. or something maybe new and similar to that, to maybe provide, if not guardrails, at least a little bit more guidance as to how we develop. Yes, I would like to see actually a lot more guidance. The problem is that on many of the dangerous misuses of AI by people. The government's misaligned. So, for example, the United States and China are going to collaborate with each other on how you do the best cyber attacks. The interests are exactly opposed on that. They'll never collaborate on that. They're not going to collaborate with each other on how you build better autonomous lethal weapons. The interests of misaligned on that. There is one issue, though, on which all countries have aligned interests will cooperate, and that's preventing superintelligent AI from taking over from people. No country wants that. The American government doesn't want that. The Chinese Communist Party doesn't want that for sure. So they will collaborate on that. But there are a lot of people out there, as you know, Professor, that believe that that's probably inevitable, maybe not in a whole as in 100% of all of our jobs and tasks. But I mean, you even go back to some of the early writings of Keynes and a few others, this idea that maybe some form of artificial intelligence replacing or technology, I should say replacing our day to day jobs, if you will, that basically frees us up to be creatives and artists or whatever else out there. Is that a fantasy? It's not entirely a fantasy. That would be great on the face of it. If I can greatly increase productivity, then there should be more goods and services for everybody. So everybody should be better off. We would have to work less. We'd be more free to do the most creative things we do or the most human things we do. But we live in a particular kind of society. It's served most of us very well so far. But in that capitalist society, unless it's well regulated, what's going to happen is artificial intelligence is going to make the rich a lot richer and a lot of people will lose their jobs. That's what I believe. It's not certain. Some economists disagree. I think they're wrong.

No schedule data available.

BTV Channel FinderWatch BTV in your area

'Godfather of AI': Need a Lot More Government Guidance

Geoffrey Hinton, University of Toronto professor emeritus and Nobel Prize winner in Physics, says many academics are worried about superintelligence. He stresses, “It’s very scary unless we do it carefully,” and tells Romaine Bostick and Matt Miller on “The Close” that some companies are more ethical than others, with Anthropic among them. (Source: Bloomberg)

More From Bloomberg Markets: The Close
All episodes and clips

Bloomberg Technology
The only daily news program focused exclusively on technology, innovation and the future of business from San Francisco. Hosted by Emily Chang.
More episodes and clips