Dr. Arati Prabhakar, Director, White House Office of Science and Technology
(appearing virtually)
In conversation with: Alyson Shontell, Editor-in-Chief, FORTUNE; Co-chair, Fortune Brainstorm Tech 

Brainstorm Tech 2023: Dr. Arati Prabhakar on how the White House is catching up with tech policy

Conferences

Dr. Arati Prabhakar, Director, White House Office of Science and Technology
(appearing virtually)
Alyson Shontell, Editor-in-Chief, FORTUNE


The White House is catching up with tech policy

Dr. Arati Prabhakar, Director, White House Office of Science and Technology
(appearing virtually)
In conversation with: Alyson Shontell, Editor-in-Chief, FORTUNE; Co-chair, Fortune Brainstorm Tech 


In May 2023 the US Senate Judiciary Committee launched oversight hearings into potential ai regulation. The desire was to avoid the mistake of waiting too long to address potential risks. As happened with social media, the proceedings opened with a deep fake using an AI voice clone of Senator Blumenthal, chairman of the subcommittee on privacy technology and the law reading a speech written by chat GPT. He called the experience eerie, even creepy. The committee heard from executives from IBM and open A I as well as academics who spoke in resounding agreement that the AI boom requires safeguards and oversight. In essence, the industry came asking for rules. So what could a national ai strategy look like and how can our government ensure that we are in control of these tools and not the other way around? And on the business side, how can government make sure innovation isn't stifled and that we aren't and that we are staying resilient and competitive in today's complex global setting to answer these questions and more. We are delighted to host Dr Arati Prabhakar, Director of the White House Office of Science and Technology. Dr Prabhakar is a member of the cabinet and is the President's Chief Advisor for Science and Tech. Please welcome Dr Prabhakar and Fortunes editor in chief Alison Chantel. Good afternoon, everyone. Thank you for joining us here at brainstorm Tech and thank you, Doctor Prab for being here with us virtually. It is such a pleasure to see you. Um Thank you all for being here. So we just had the anthropic CEO on stage and we asked him about the meeting that you all had in May at the White House with President Biden and Kamala Harris. All about mitigating risks with A I. The CEOS of Google were there Microsoft Open AI and Anthropic Dr Prab Baar. I'm wondering what your key takeaway was. What did you leave that meeting? Thinking like, oh man, this is the action item. This is how we have to act right now. Uh Well, hello, Alison. Hello, everyone. I, as you can imagine, I wish I were with you live. Thanks so much for the chance to do this. Um And II, I want to start by stepping back from your question for just a minute, Alison, give you a sense of how we're approaching artificial intelligence from here at the White House. Uh This, the president has been very clear that this is a top priority for this administration. And the reason for that is really simple. I mean, just as you've been talking about at your conference, I think we're all aware that a I is one of the most powerful technologies of our time. Uh What we know from a lot of history is that means people are going to use it for good and they're gonna use it for ill. And so the the task before us is to seize A I's benefits. But we know that we have to do to do that where we have to start as managing its risk. And this administration is moving out very aggressively to, to figure out how to do that in the most constructive possible way. Uh I'll just frame our conversation with uh three points. Uh The first is simply to say what I think you're, you know, you've been exploring here, which is the power of this technology is at breadth. And that also means when the applications are broad, it means that the risks are very, very broad. And so I think we need to be very clear that this isn't going to be one and done. There are many different aspects of the different kinds of risks and farms that A I I will bring has already brought and will continue to bring. And so all of those, everything from dealing with risks to truth, trust in democracy, dealing with the risks to safety and security, dealing with the risks to privacy and civil rights and civil liberty and dealing with the risks to jobs and the economy. All of that has to be on the plate. Number one, uh number two, I'll just say that while they, you know, people talk about a I through a lens of science fiction, there's a lot of discussion about, uh what does it mean if the machines are conscious? Sometimes I feel like I'm in a conversation in a freshman dorm room at midnight. Those are all really interesting conversations. Uh but they're not the conversations that are going to drive policy. What's driving policy is just I very clearly focused on the fact that it's human beings who build A I systems who train them on human data and then decide where to use A I. And that's really how we're, you know, that's how we're going to just stay grounded in the real world and the work that we're doing in policy. And the final thing I'll say is you mentioned the CEO meeting, that is one in a series of things that we are take uh where we're moving out and taking action. Um starting with um uh engagement with the companies, we think they have important responsibilities and, and that they are working to step up to. We are encouraging them to take responsibility for these technologies that they're putting out in the world. Uh You'll also see us moving forward with executive action, working with Congress on legislation uh and working internationally. So that's a broad overview, but there's a lot going on as I think is appropriate for this powerful technology. That's a great overview. Thank you so much. Um I'm curious about some of those executive actions like, where do you start when this train is moving so fast? The anthropic was saying, you know, we were recommending plan for two years out. You can't plan for today. It's going to be outdated in like a day plan for a few years out. How are you kind of wrapping your arms around creating policies and thinking about policies that will be future proof when this is really on exponential growth. Yeah. And you put your finger on the whole uh the the challenge here because the technology is moving fast. But of course, we need stable regulatory regimes. And I think, you know, especially when technology is moving fast, it's the time that it is critically important to be clear about the values because that's what, that's what you're going to navigate by. Uh and uh to that point, um the this administration, our office here, the White House put out a report called the Blueprint for an A I Bill of Rights. We did that in October of 2022 just before this uh generative A I generation of A I broke loose and, and really took the world by storm. And I think it was critically important to have that, that clear, you know, a lighthouse to steer by in terms of our values. So I think that's critical, what does it mean in practical terms? One a lot of the harms that people have identified from A I are things that happen to be illegal already. So committing fraud or cyber crime, these are things that we already have laws and regulations to deal with. But in this A I era, it's critically important that we, you know, we make sure we're enforcing that we're able to keep up with whatever acceleration comes from A I, I think there's a separate conversation that's far less clear about managing the technology itself. Is it a matter of understanding what kind of, you know, is there, are there transparency factors that could be helpful? Are there regulations? And that is not yet an answered question. And then the final part of dealing with this, I think really comes to the core, something that I think we keep talking about is how opaque this technology is how hard it is to know why you, that the answer that you got uh and if we're going to have safe and effective A I, that's the goal, that's what we're trying to navigate to. We're gonna need tools and methods for assessing, evaluating testing, benchmarking this uh very powerful technology. And that, that is uh an area where definitely companies are going to be involved. But that's an area that I think we need to take some important government responsibility for as well. So you have a very hard challenge ahead of you trying to wrap your arms around all of this plan for something that we've never experienced before. Um Even if we are able to create great regulation here in the United States, how are you thinking about multinational? Like, what kind of multinational agreements do we need to have in place? How worried are you about? What kind of A I other places like China might build? How are you thinking about all of that? Yeah, the, the the first place to start here is to just recognize that when you have this kind of powerful capability, every nation, every company, everyone on the planet is trying to use A I to create a future that reflects their values. And I think we're all pretty clear, we don't want to live in a future that's shaped, you know, that's driven by technology that's been shaped by authoritarian regimes. And, and because, you know, look the ingredients for this, our data and computing algorithms, they're widely available around the world. So it is absolutely a global race to get this right. And to do this in a way that really reflects the values of, of democratic nations. So that means moving out here in the United States. But doing that in a way that's lined up and uh deeply coordinated with our like minded allies around the world. And that's uh that's an integral part of the approach that we're taking. And how are you thinking about regulation versus innovation when it comes to businesses building A I like the anthropic how are you? What kind of regulations are, you should ceos be aware of that you're kind of thinking through. Yeah, I, again, I think, you know, the, the, the place to start is to just, there's so many of the harms that, that we, everyone's identified from artificial intelligence that are malfeasance and that for which we already have laws and regulations and, and managing the, the, the ways that applications of A I can go wrong is a critical place to start. Uh And, and so people should definitely expect just as an example, um the uh equal opportunity Employment Commission, the Consumer Protection, uh Consumer Financial Protection Bureau, the Department of Justice and the Federal Trade Commission all came out a couple of months ago with a statement just asserting that, that they are still enforcing uh against violations for which we have laws and regulations on the book and that the fact that you're using artificial intelligence technology does not get you off the hook. So I think that's a great example of the kind of beefing up of regulation to manage the harms when A I is used, I don't see a fundamental trade-off between regulation and innovation. Uh if you get regulation, right? And that's exactly what we are working on right now. That's what's actually going to unleash the power of this innovation. Because if you step back and you look at where A I is today, there is great excitement, but there is also great trepidation and it's not until people really come to trust this technology, that's when we're going to get the full power of what, what a I can bring and, and regulation and the putting the, um, the control points in so that we can, we can, you know, it's sort of like having brakes on a car. It actually lets you go faster once, you know, that you're in control and that's really where I think we have to get to this. Um, you know, there's a multitude of risks, I guess. One question I would have is what's the one that keeps you the most up at night? I mean, you can rattle off a very long list and all of them could probably keep you up at night. But is there one that you're just like, this is the one we really need to focus on first? Um, you know, I'll tell you, I hear a lot of people in companies or in civil society organizations who are absolutely fixated on their one risk that they are working on and, and, you know, somewhat to my amusement, they argue with each other about what's the most important risk. I would just like to reject that framing. We are the United States of America. We do not have the luxury of just solving one of these problems. And so our job in government is to make sure that we keep that full plate in front of us. Because I do not want to live in a world that is filled with biosecurity risks because of a I, I equally don't want to live in a world in which a black man is getting arrested for a crime he didn't commit. Because the facial recognition system can't recognize black faces appropriately. Neither of those is acceptable. And so you'll see us continue to work on all of the different aspects of this. I agree with you. I don't want to live in that world either. So I support that answer. One thing that President Biden has said is this is another risk is we have to prove that democracy works in the 21st century. What impact are you bracing for that? A I could have that could further deteriorate our already fragile democracy and really just trust, you know, one of the reasons I was so thrilled that the president asked me to come serve in this role was exactly that quote because I think that is the moment that we're in. And the president specifically talks about how artificial intelligence is part of what is shaping. It's a very complex world that we're navigating in. But what does it mean to prove that democracy works well, first of all, the operations of democracy, the truth and the trust that we all have to have as a society for a democracy to function that has to be in place. And of course, you know, many people have highlighted the potential for this new gene for generative A I to accelerate misinformation and disinformation and, and just the, you know, the psychological reality distortion that can happen would be a very powerful system. But I think that's a clear part of making sure that democracy works. But it's actually deeper than that because what it means for democracy to work is that the advances that come out of a powerful technology like A I uh that they benefit people across the country and that, and that we, we it, it, it, this, this is part of opening the door to opportunity. The most fundamental American idea you can possibly imagine is that every single person in this country gets to participate, gets a path to choose how they, how they make their contribution to the future. We've got some really important choices to make because that is the future that A I can enable. But we have a lot of work to do to make sure we get it on that path. And, and you know, that that's why I and my colleagues come to work every single day. Um Cyber security is of course another issue. It's been on ceos minds for years and you know A I acceleration is bringing that even more importance around that conversation. Um What can be done between government and companies to help mitigate what's ahead. Yeah, I I think the confluence of the, the issues of cyber security that we have all wrestled with for a long time now continue to work on and now comes A I so many different implications because of course, there are many ways in which A I technologies can help us with maintaining and boosting cyber security. The use of A I can be very powerful but can also expose, you know, can increase the attack surface for Attackers to come in. And of course, in addition to all the conventional ways that people have gone after attacking software systems, ai offers new kinds of attack surfaces because of the way that training data can undermine the performance, for example of a system. So I think there are a lot of uh to deal with here. Uh you know, a very particular area of concern is the vulnerability of critical infrastructure and the power of A I to increase those vulnerabilities but also to be harnessed to protect them. So I think this is an area where um we we will continue to see a lot of innovation, a lot of concerns and and we've got a lot of work to do now. Uh So I would say the government has a reputation for wanting to do a lot of good, but also sometimes being a little slow. And this is the fastest moving technology we have ever seen. I think when it comes to government regulation and social media, it wasn't fast enough. Was it iterative enough? Even with the internet? Not fast enough, not iterative enough. How are you thinking about this wave? How do we not make those same mistakes? How can you remove barriers to move as fast as you're going to need to move to put what we need in place to be protected? Yeah, you know, Alison, I've spent half of my professional life in the public sector and half in the private sector and in my private sector life, it was, of course, we need to move quickly when technology is moving. Um But the thing that is most, when I'm in a company, I definitely want want regulatory clarity, but I actually don't want regulations that change and Twitch as fast as the technology moves. And so I think it's really important to get clear regulations um in place in ways that can be robust enough, but also adaptive enough as the technology changes. And I'll come back to where I started because I think being clear about the values that we're driving to is essential uh precisely because the the fast pace of change makes for some very choppy waters right now. Uh And when you're in choppy waters, that's when you have to be really clear about what it is you're steering to. So staying focused on the idea of uh systems that have to be safe systems that have to prove that they are effective and accurate and that they work in the ways that that that was intended, being clear, that embedding bias has it cannot happen that we can't use systems until those issues have been run out. I think those are going to be the guideposts so that as the technology changes, we continue to steer to those core values.