What do you make of the sort of preamble that we're fragmented? I mean, we certainly are fragmented. I think when we think about global governance, there's actually an advantage to the companies in having global governance, and that is that fragmentation. And you think about how expensive it is to train these models. You saw earlier today Microsoft buying, you know, the rights to the energy of Three Mile Island 100%. It's very expensive to train these models. And if you have to do that differently for every jurisdiction because they each have their own rules, that would be an insane amount of power, especially when we're talking about maybe making bigger models 100 times a thousand times, 10,000 times bigger than existing models. So fragmentation is actually a problem for the companies. It's also a problem for the world. I mean, we really do want to have some uniform approach to all of the downsides things like non-consensual deepfake porn and misinformation and so forth.

There are many downsides, and we want a coordinated response to it. So, I think it's actually in everybody's interests to have some coordination, but fragmentation is the default. Let's talk about coordination. The United Nations has put up this thought process report, and now they want an independent scientific panel on artificial intelligence in the same way they almost have with climate change. Is that going to help? Is that going to make sure that more people are at the table other than basically the G7 and some of the biggest companies in the world? I mean, God is in the details, but I think it's a good start. I think absolutely we need more independent scientists weighing in. We also need authority, if something doesn't work, to do something. So, one of the most interesting things that happened the last couple of weeks is OpenAI put out a new model, and they said there's a bioweapons risk.

here. Now, they might be right about that. They might be wrong about it, but let's suppose they're right about it. They decided despite that risk, they would go ahead and release it. And that was really just ultimately Sam Altman's decision and nobody else's. There was Senate testimony earlier in the week where a former employee and a former board member said basically they're not they didn't feel listened to. Right. And so open air. Sam Altman doesn't have to listen. His employees say the bioweapons risk here is serious. Maybe not for this model, but the next model, he can go ahead anyway and say, But I need to make back the money for the hundred $50 billion valuation, which, by the way, is not a trivial thing, and there's also no outside input. So even if this IPC like panel that says, you know, this is dangerous, there's nothing in the U.N.

that can immediately do anything about it. There's nothing in the US government that can immediately do anything about it. So it's a very good, I think, thought experiment, like when this stuff gets worse, who's making these decisions? I mean, but this has happened time and time again and we always go, They should self-regulate. And in a way, someone did just that. You just told me that you're coughing a bit because you're basically exhausted, having written this book in two months out of sheer anger in a way. What's the what's the response, The anger? What's the answer here? Well, I mean, the anger is partly I went to the Senate. I testified next to Sam Altman, and he talked about not just self-regulation, but real regulation. The Senate bipartisan, all talked about this, as was May of last year. Nothing has actually happened. There's been no, you know, major piece

of legislation in the United States to deal with any of this. And self-regulation is not going to happen. I mean, even when they make promises, they often backed down, and they're not there. They're going to agree to things like information sharing. They're not going to agree to, you know, deferring the decision about whether to release this new product that maybe finally makes them some money to the general public. And they're not making money right now. I mean, OpenAI had a $5 billion operating loss here last year. Let's go back to what you just said on OpenAI. It's not a trivial amount of money, but the stakes are also very high. How much does that keep you up at night? That we have a single company with quite a confusing governance structure that has the world's biggest technology companies and investors fighting over around. And it's releasing technology at a cadence that you outlined. I mean, everybody should have been

worried at the drama last year. It sort of made for an amazing new Netflix special or something like that. When Sam, we remember fires and rehired and all of this drama, but it also says like, who is running this company and what are the checks and balances. Right. And the ultimate answer is Sam is running it. There are no checks and balances. And, you know, we should be worried. The company can't even govern itself internally. A lot of people left after that incident. We should be worried. It's a lot of power and we should be worried not just about open air, but the whole industry, like responsibly, excuse me, was a watchword for the entire field for a few years. And now it's almost like nobody really cares because there's so much money on the table. We had the president of Latvia here yesterday from the European Union. You have UNGA next week.

Do you have any hope that actually this sort of roundtable, state-level discussions will yield some kind of framework in which the world is kept safe? For me, I have a mild bit of hope, but I think that we really need the public to be involved and say this matters. So, you know, politicians always have to weigh competing interests. They get a lot of money from the tech companies. They're involved with the tech companies in various ways. If the public doesn't speak up. I think we will see very light regulation, if any, at all. The reason I wrote this book is, I think making these decisions the most important thing that we can do. But one of the most important things we can do in the world in the next few years think about what happened with social media. We waited too long and we entrench the set of practices that now lead to all kinds of polarization, bad for teenage girls and so forth.

And it's hard to fix these things. We have to get a policy right now.

NYU's Gary Marcus on the Need for Tech, AI Regulation

Welcome to the Bloomberg Technology YouTube channel. Subscribe for original series and mini-docs on game-changing technologies, profiles on innovators, live conversations with experts, and breaking technology news & analysis from Bloomberg TV.

NYU's Gary Marcus on the Need for Tech, AI Regulation

NYU's Gary Marcus on the Need for Tech, AI Regulation

Scripts