Open Air is based in San Francisco, California. It is probably the world's leading air name or startup. And they think that federal legislation is a better path than state. Your reaction to that? Well, I will start by saying that another day. I live in San Francisco Anthropic, which currently has the largest model in the world, issued a letter yesterday stating that on balance, SB 1047 is good and that it is implementable and that Congress is unlikely to be fair. You're talking about anthropic. Anthropic. I'm sorry that I not to them anthropic came out and said that on balance the bill is good and that it's implementable and that Congress is unlikely to act. And so I have a lot of respect for open air. But just to be clear, the opening letter does not actually criticize any provision of the bill. It does not criticize anything in the bill. All it says is that this should be handled at the federal level.

And to be clear, I agree it should be handled at the federal level. But here we are in 2024, and Congress still has not passed a data privacy law, has done nothing around social media other than banning TikTok, and has not codified net neutrality. Congress has a very poor record in terms of regulating the tech sector, and I don't see that changing. And so we, California, should lead. Do you think that we should trust? There's been a long list of tech companies, VCs, academics who have criticized this bill. But do you think that we should be trusting some of these voices who are against it? Are there other motivations that you think could be at play here? So, to be clear, there are folks who are criticizing the bill in good faith and people I respect a lot. There are others who are making very inaccurate and extreme and exaggerated attacks on this bill. This is a very light touch, a smart piece of regulation around safety,

regulating only the largest models to just ask these developers to perform safety testing so that we can reduce the risk of things like shutting down the electric grid or attacking the banking system or creating a chemical or biological weapon. And some of the leading minds in AI, like Joshua Bengio and Jeff Hinton, formerly from Google, and Lawrence Lessig and Stuart Russell, are all supporting this bill. In addition to a significant list of startups who are saying that this is a good idea. So, yes, there are some people in good faith who have criticisms of the bill, and we have been working with them to get their feedback. There are others who are just opposed to any and all regulation of tech, and I just don't agree with that. And I want to talk about whistleblower protections, which is also a part of the bill, and it's received less attention today. Two former open-air employees came out in support of the bill, and they have

advocated for more whistleblower protections. Why is that a part of the bill? Why is it important that people who work for these companies, you think should be able to speak out on safety? So I am, just to be clear, a huge fan of it. I think A.I. has so much potential to make the world a better place to solve what we have viewed as unsolvable problems, whether it's around various health conditions or climate change. And so I'm a fan. But as with any extremely powerful technology, there are risks and we should try to get ahead of those risks. And we know how to get ahead of those risks, not to eliminate risk, but to reduce the risk of something catastrophic happening. And so when you have people who are working in A.I. Labs, building some of the most powerful technology in the history of the world, they should be able to speak up and have legal protections to speak up if they

see something inappropriate or dangerous happening. Are you yourself, Senator? Can I ask? Nervous, even scared around artificial intelligence having done this amount of research. I wouldn't I wouldn't say scared. Like, I'm I'm optimistic that I can help us solve some of the hardest problems facing us. At the same time, it is it would be irresponsible to just allow it to spiral and not to in any way even think about the safety issues. And that's all this bill does. It requires safety protocols, a safety evaluation, the ability to shut down a model that is in your possession. These are all extremely reasonable things. I want to be clear that every single large AI lab, including open A.I. and Meta and Google, they have already committed repeatedly and publicly to perform these safety evaluations. We're just asking them to do it in an obligatory way. And it should cause us all concern. Whenever you have an industry that says, We promise we'll do this but don't

require us to do it. Just trust us. Voluntary Commitments Industry have a history of going awry. Senator, you have listened and listened hard to some of the concerns and made changes. What seems to be the ongoing narrative from academia and indeed smaller companies is this open source narrative and the way in which perhaps Lama would in any way suddenly become less available and less innovative. How can you change to enable smaller companies and open source still to thrive under this regulation? Well, I'm a huge fan of open source, including in the air space and so many startups in particular. And by the way, startups are not covered by this bill. It applies only to labs that are spending more than 00 million to train their models. But I actually do. Open source A.I. plays an incredibly important role, and we have worked intensively with GitHub and other open source advocates and made significant changes to the bill to make very, very clear that this bill is not about being negative with open source.

And yes, a lot of people rely on Lama, but I know I sound like a broken record, but I'll say it again. Metta is one of the labs that has repeatedly committed to performing safety evaluations. And so that's what we have put into this bill. And so when you make that commitment and have made that commitment, clearly believing that it was not going to undermine their ability to open source. Lamar Mr. Wena, when Shery Ahn posted on social media, you are coming onto the program. Inevitably, we've got a number of questions from people all over the state. One of them was Professor and IMA, and I'm Kumar, actually previously an employee of Nvidia no longer is is an academic and is in academia. She wants to focus. She basically says in this question to you that you are pretty fixated on billionaires opposing the bill. And she asks, You know how much you take

into account the input of academia. But I also note that she is not alone in that a number of academics came forward when we said, You joined the program. Yeah, I saw that tweet and I have no idea what she's talking about. At no point have I ever said that it's only billionaires who have concerns about this bill. There. I also want to say that there are large venture capital firms, particularly Andreessen Horowitz, that have taken the lead in spreading at times. And Andreessen Horowitz has done misinformation about the bill. But with that said, we know that there are academics who have expressed concern about the open source aspect. And so I have met with various of those academics. We have made significant amendments to the bill to address concerns from the open source community, and there are significant academics like Stuart Russell and Yoshio Bengio, Jeffrey Hood, etc. who support the bill. One of your main defenses of this bill has been that there's already laws,

right, tort law. You've said that that could find companies liable if they do cause mass harm through A.I. You also, as you mentioned, said there's these voluntary commitments. So my question is, if those already exist, why do you think we need this law? Companies can already get in trouble if they cause harm. Well, and that's a really important point, because one of the inaccurate representations that we've seen around this bill is implying or even stating that this bill creates liability for A.I. developers when they had no liability before. Currently today, an AI developer, if you, let's say, open source a model and someone uses that model to, you know, shut down the electric grid, you can be sued today. And so our bill allows the attorney general to get involved. But what the bill does is it creates a very Pacific Obligation not to create liability, but to tell developers, this

is what we expect you to do. You need to test your model. You need to have a safety protocol. You need to be able to shut down the model in a very, very specific way. We also provide the attorney general with a very narrow ability to intervene before a harm occurs to avoid a catastrophic harm in extreme circumstances. So this bill complements existing law. SB 1047 is a state California state bill, and ultimately this might come down to Governor Newsom. Your expectation of what the governor may or may not do in a veto or a signature or signature. I know the governor has not stated a position on this bill. I have learned the hard way over years not to assume whether the governor will sign or veto a bill. He and Governor Brown have surprised me in both directions in the past. You know, the governor is a strong supporter of tech innovation, as am I. The governor has also made clear that when it comes to A.I., that regulation

is appropriate, and we have to strike the right balance. So, I hope that the governor will have an open mind, and I believe he will have an open mind about this bill. And we will continue to make the case. This bill has been getting very, very broad support, particularly among Democrats in the legislature. Your constituency is very big if you include San Francisco in the concentration of AI companies here. Who else will you meet with to try and convince of this bill? We have met with, I mean, we have not met with. We've met with anyone who will meet with us. And we have met with, you know, with folks from startups, from large tech companies, from folks from academia, just frontline tech workers. And by the way, when you pull this bill, it pulls off the charts with the public polls higher among tech workers than it does among the general public. And it was frontline grassroots tech workers who came to me originally with

Some of the ideas in this bill, because these are folks working in the airspace who had concerns about safety.

Sen. Scott Wiener Takes on OpenAI

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.

Sen. Scott Wiener Takes on OpenAI

Sen. Scott Wiener Takes on OpenAI

스크립트