Loading summary
Transcript43 lines
- [00:05]
A
Hi, I'm Santi Ruiz. Very excited to be talking to you guys today. And I'm very excited as well to be speaking to Director Michael Karazios, who runs the White House Office of Science and Technology Policy. He's the President's top science and technology advisor in the first Trump administration. Karatios was U.S. chief technology officer and later acting Undersecretary of Defense for Research and Engineering. Michael now leads the administration's science and tech agenda. Michael, thanks for joining. It's a shame this has to be virtually.
- [00:32]
B
I know, I know you can thank the Democrats here in Washington for that. But it's exciting to be with you guys.
- [00:37]
A
Well, we've got a lot to get through. I'll just, as a roadmap, I want to talk to you a bit about the AI action plan. I want to hear a little bit about this Edmonds vision for gold standard science, which you've championed, some of your work on other kinds of emerging tech. And then I'm curious to hear your perspective, having been in the first Trump administration and the second, just your perspective on how this admin works, how it operates, just the nuts and bolts. And I know a lot of people in this audience will be curious for that as well.
- [01:02]
B
Absolutely. It sounds great.
- [01:03]
A
Super. Well, let me start here. You gave a speech in April about the golden age of American innovation. You said it's on our horizon if we choose it. It was a really compelling speech about how stagnation is a choice. But you had this turn of phrase in there. You said our technologies permit us to manipulate time and space. And I saw quite a few conspiracy theorists jump on that turn of phrase. So first question for you. Does the US Government have secret wormhole technology that you're hiding from us?
- [01:28]
B
I'll tell you, that was one of the most exciting speeches I think we gave this term. I think it really set the foundation for where we want to go as a country on technology policy more broadly. And I think you brought up the main thrust of the speech was around technological stagnation. To me, I think it's an absolute tragedy the extent to which progress in the world of Adams has slowed down over the last few decades. And we point to a lot of those examples in the speech itself. And, and I think part of the way we're going to get out of it is the agenda that the President is laying out and the important role that the federal government play in driving broader scientific and technological discoveries. So that kind of, I think, kind of laid out the vision for what tech policy is for the Trump administration. And it's something we've been charging ahead on and things like nuclear energy and AI and lots of other stuff that you probably want to cover.
- [02:13]
A
Sure. Tell me a little bit about your. This administration's perspective on the reasons for that stagnation. Why do we have to get out of this stagnating moment?
- [02:23]
B
Well, I think for, to ensure our long term economic national security, we have to continue to drive and innovate here in the US and to me, I think one of the biggest drivers or reasons why we as a country will succeed is because of the innovation that we can have in science and technology. And I think there is a very, very deep and rich history of the federal government in the United States being part of a larger innovation ecosystem. And they play different roles in different eras based on the challenges of the day. You look back to the sort of Manhattan Project as kind of one of the big obvious examples. We had to come together as a government with the private sector to develop something that could help us ultimately win the second World War. You move ahead to things like the Apollo project, where again, it was a federal effort that was able to work with lots of folks in the private sector to cement our leadership in the space race. And I think there's so many other examples. And I think the challenge that we have faced over the years, and one thing that we think a lot about at this office is the changing nature of where early stage, basic pre competitive research and development happens. And if you go back to the era of sort of like Post World War II, the vast, vast majority of R and D was being funded by the federal government. And over the last 70 years we've had this inversion or this flip where now the majority of R and D is funded in the private sector. And we have to ask ourselves every day sort of, what role does the federal government play in a world where the private sector is investing more than any time in history in this research and development? And I think the most sort of the biggest example, the one that's kind of like screaming in our face every day, is with AI, you look at these numbers of the types of investment that are happening in infrastructure and R and D that's supporting the AI revolution. And I think even for a lot of us that are like deeply immersed in tech and have been for decades, I mean, these numbers are crazy. They are like truly unbelievable levels of investment. And it doesn't seem like it's stopping. Every week there's a new big one. And I guess for policymakers like Us here in Washington, we have to sit back and say, look, there is billions if not trillions of dollars being invested by the private sector in AI, but what role can the federal government play? And that's the same type of question we ask for any of these emerging technologies. And I think going back to the speech, the point was that there always will be a very important role to play and the government has to make that choice to say like, we do not accept that stagnation. We want to work hand in glove with the private sector to make it happen. And there are levers that we control which can actually make a big difference. Sure.
- [04:54]
A
You've been in the role, you were confirmed as the director of OSTP in March. And the AI Action Plan is one chunk of a whole bunch of things you've tackled. There's been a huge amount of travel, there's been work on science, on drones, on supersonic flight, on quantum, on energy for data centers. There's a whole laundry list. But let's stay on AI for a second. You rolled out in July the AI Action Plan. Broadly, I think about half of this audience will be very familiar maybe with what's in there, and about half will really not be familiar. So give us the high level takeaways there.
- [05:25]
B
Yeah. So it all began actually the third day of the administration where President Trump signed an executive order revoking the disastrous Biden executive order on, on AI and essentially gave us 180 day shot clock where myself, Secretary Rubio and David Sachs were on the hook for delivering the next national strategy for the US Called the, called the Action Plan. And we delivered it in July. And that was, that was launched an event here in Washington. At the same event, the President signed three executive orders that generally map to some of the priorities in the plan. But the vision essentially laid out in the plan has three core sort of policy thrusts or main pillars. The first is the US has to continue to out innovate and continue to be the leader in innovation. And there's lots of work in there around regulatory issues around R and D and we can get, get into those. The second is the US has to lead on AI infrastructure. This revolution is going to be powered by electricity and by data centers and we have to have those built in the United as quickly as humanly possible. And there's a lot we can do to, to accelerate that process and get the government in a way as we build that, that infrastructure. And the third is the importance of exporting American AI, essentially creating an ecosystem globally that is reliant on American technology and lots of projects are being launched on that front as well. But thematically that's kind of the core themes of the of the action plan. There are almost 100 actions enumerated in it, but they generally fall into those three buckets.
- [06:53]
A
Let me ask you about that third bucket, the export side of things. One of the flagship programs of the action plan is this export promotion program which provides export financing, other assistance to promote the adoption of this USAI tech stack overseas. The admins also talked a lot about AI sovereignty, about not wanting global governance for AI, wanting each country to be able to kind of develop its own roadmap. Here are these two things in tension. Do we want every country to have full AI sovereignty? Do we want the US tech stack to dominate in certain areas which layers? How do you think about balancing those two priorities for the admin?
- [07:28]
B
Yeah, I think of them actually in concert, not in opposition. I think for us, the core challenge we're facing and maybe if we zoom out and think a little about how we got to that third pillar, it kind of actually came out to a lot of work that we did in Trump 45 around telecom. And if you think of that era, the first sort of big rollouts of 5G happened kind of the tail end of Trump 45 and we as a country found out pretty quickly we were in a really tough spot. We had manufacturer of advanced telecom kit, there were two allied partners with Nokia and Ericsson that could provide those. And then you had an upstart which was pretty good, which was Huawei. And in the early days of that challenge, one could argue that the Huawei technology certainly was inferior to the US technology technology. But nonetheless, because of the heavy subsidy by the prc, they were able to aggressively go out and get their kit all over the global south and even deep into the networks of some of our partners and allies. Me as the CTO of the United States and our tech minister was left running around the world talking to my counterparts, trying to convince them to either rip and replace their Huawei or make the expensive decision to buy Ericsson Nokia. And there were a couple lessons that I learned out of that. I think the first was that just because you have the very best technology, that doesn't mean it's going to be adopted. Price actually does matter and the ultimate buyers are and the buyers are going to be quite, quite price sensitive, especially if at the end of the day you can get pretty much the same service by someone else. So as we and also like that created sort of a very Interesting national security problem for the US where a lot of these networks could potentially be compromised and could have sort of backdoors back to the prc. And that's something that we were very worried about and wanted to share. Now if you fast forward to where we are today, you know, one could argue that the export of American AI is, is arguably even more important than, than telecom in the sense that the models of choice and the stacks that are used around the world are going to be critically important in the way that lots of governments and economies around the world function. The ultimate software that is used to power health care services, to run tax services, to essentially be the in which all of governments run all their AI services is really, really important. And you want to be in a position where American models are the ones that are fine tuned to solve all of these, to run all these applications ultimately. Now the advantage we have and what's different than than before is that we actually do have the very best technology. And I think China at the moment is actually quite constrained in their ability to actually fabricate enough chips for export. You know, the ones that do fabricate domestically, there's more than enough demand. Internally, there just isn't enough excess to exports. So we have this window of time in the next year or truly can be sort of the single powerful supplier of the totality of the stack, meaning chips, models and applications. And the other advantage we have in before is we America, our companies are leading in all three layers. We have the very best chips in the world, we have the very best LLMs in the world, and we have the very best applications on top of it. So the thesis of the program is let's try to bring all those together, make turnkey solutions that countries around the world can easily adopt and implement and do that as quickly as possible so that we can start creating these networks and these ecosystems where developers and countries and governments around the world are all using American technology.
- [11:00]
A
You describe this kind of year long lead time that we have as a result of the lead technologically. What does that mean for you in this role? You're doing a ton of travel, a lot of international diplomacy, basically say a little bit more about what that entails over the next year or so.
- [11:15]
B
Yes, yes. So we, you know, the President signed the executive order in July of this year. There's a shot clock at the Department of Commerce to sort of stand up the program actually in the next week or two. So we'll see some announcements out of commerce coming soon and then kind of the ball is ultimately handed to A committee that's led by the Secretary of State, that includes Secretary of Commerce and me and a few others to actually go out and do, do the export early next year. What we want to do is get, I mean, we, we want the world to know that America is open for business. We want the world to be running on our technology, which we think is the best. And we want to go out in the world and tell them that. Now, I was at the APEC digital ministerial in Korea a few months ago where we shared this message and we essentially said that the US is open for business, we do want to sell you chips, and we deeply believe that the ideals and the priorities that you have as countries ultimately are ones that are aligned with the us. And I think it kind of goes back a little bit to the sovereignty question you have there. And I didn't quite answer it, but to answer it, I think the reality is we have a stack which can answer the mail to the priorities of individual countries. If you're a country and you have sensitive government data, you obviously don't want to be running an AI system where there's an API call to a foreign country, even if it is the United States. We understand that. And the hope is that we can design a system with the great technology companies we have today which can allow them to operate American software in a way that makes it ultimately is able to sort of meet whatever standards they have, they have in country.
- [12:41]
A
You and I talked about a year ago, before the election, before this role, and you were reflecting on the first admin and on the balance between protect actions, this was your language, and promote actions protecting American tech industry and promoting the tech stack abroad. And we talked a little bit then about the Huawei wars, about trying to get that technology out of the infrastructure of our allies. I want to ask a little bit more about how you're thinking about the protect side of things, because in the last couple weeks the Secret Service uncovered this massive telecoms hacking network in New York City. Earlier this week, a former National Security Council staffer in the George W. Bush administration was arrested after the FBI raided his home and they found thousands of top secret Air Force documents that he'd been passing along. This week, the British government admitted Chinese hackers have been inside the UK's classified network for over a decade. Given all these kind of salient stories in very different domains, how are you thinking about the task of protecting American infrastructure?
- [13:42]
B
Yeah, I think generally think about protect. It's a little bit of what you said and it's a lot about the world of exports and constraints for some of our adversaries from accessing our highest end technology. So on one hand there's an incredible amount that we need to do to secure our critical infrastructure. We have a new national Cyber Director here at the White House is very focused on that along with the folks at dhs. And we continue to be a need to be very vigilant in tracking the type of hacking that ultimately happens by a lot of our adversaries. And we see it, we see it all the time and it's disheartening, but we've got to do better now on the more protect side of technology. I think what a policy that we've kind of maintained and kept in place is, particularly in the world of AI. Our most high end chips are ones that we are limiting the ability for adversaries to have access to. And that's something that had started, you know, as early as the Trump administration where we put export controls on, on some more advanced lithography equipment which ultimately has paid big dividends in limiting the ability of the PRC to develop high end chips. And that will continue. And I finding that balance of needing to promote as much as possible get American tech out there and also rate limit our adversary on the very high end stuff is kind of the general strategy.
- [14:49]
A
Yeah. Tell me a little bit about the data center build out work that you're doing here. I think folks in this audience will be aware there's a huge need for energy if we want to build this out domestically and that this admin cares about that. But practically, what does that look like for you again? Maybe over the next year? What are you and this admin trying to do to make that happen?
- [15:08]
B
Yeah, so that is all about our sort of second pillar of the strategy is how do we accelerate our ability to build both the data centers and the power capacity? You know, it's a very tricky and interesting problem and I think many people are deep in this realize it. There is one part that's federal government related and there is certainly NEPA related things you have to get through and other types of federal permitting. And that's the stuff that we as a government have a lot of control of. And the President directed in his executive order on this AI infrastructure that we need to as quickly as possible accelerate that effort to remove those barriers and make it much easier to get your adjudication done on approval or not for your projects. It's an ongoing effort and we want to make it as easy as possible. Essentially a one stop shop to Come to the government and say, look, I want to build this AI infrastructure, help me get all the necessary permits and approvals and we can help you do that. And that's in the works. And that's something that was called for an executive order and we're excited to keep pushing on that. The second piece, which is a little trickier, is a state and local and I think if you talk to a lot of people who are building data centers and a lot of people who are trying to do the energy build out, a lot of the bottlenecks end up at the state and local level. So we're trying our best to liaise with a lot of state governments to try to make it clear that this is a priority for their country. They as a state can benefit if these build outs ultimately happen. And kind of the third piece around that is of parts of things that the federal government controls is we have our own real estate, we have federal lands which can be used for build outs themselves. And many of you may have been tracking the Department of Energy program. The of Energy already has announced four locations for build outs of data centers and that's already underway and we did that in the first six months of the admin. So it's a lot of that. Harking back to the question you had at the very, very beginning, it's a lot about thinking about what levers does the federal government have at its disposal to kind of drive this acceleration.
- [16:55]
A
Right. Well, one question I'm especially interested in is what levers you have at the ostp because you know, for folks who follow are interested in the executive office of the President and how the executive branch functions, they'll know OSDP doesn't have the kind of formal statutory power or the kind of budget the other organs of the government, other cabinet level roles maybe do. It has this kind of convening power, your advisor to the President. So it has all kinds of power that are not formal or statutory. But as I understand you have a fairly small team compared to other parts of the executive branch. And I think traditionally OSDP has been. Your former colleague Dean Ball described it as the little brother to the National Security Council, especially on tech related issues. So I'm curious, I guess two questions. One is just how does implementation work for you on the AI action plan when you don't have all the levers that say the Department of Energy does? And between Trump 1 and Trump 2, have those kind of balances between different organs in the executive branch changed at all?
- [17:59]
B
Maybe there's a slight misunderstanding So I think in the, in the executive office of President, there is no component which has any budget to really do anything. The EOP is designed law to be a advisor to the President and a convener to drive interagency policy on a variety of issues. So generally councils or I guess counsels at the White House, which drive different policy, you have an economic council, you have a domestic council, you have National Security Council, and you have a science and tech and each of those has their own category of work. In Trump 45, science and tech was exclusively at our office. Biden people made the choice to absorb some of the tech effort. At the National Security council Based in Trump 47, it's all back in the office that has statute authority over it. But I think what's key for us, and I think what I love about this role and I think what makes really our country so unique and special and the way we track S and T is that it actually is a federated approach to science and technology policy in the sense that almost all of my counterparts, my counterpart tech ministers or science ministers around the world, they just have a single agency that attempts to do everything on their own. And they have one person that tries to decide the best way to deliver on whatever the mission is of that country. In the US we're blessed and sometimes it's a little challenging because you're wrangling people, but we're blessed with having lots of agencies that do different aspects of the larger S and T portfolio. At the Department of Energy, you have all the national labs, almost $10 billion in spending just at the under the undersecretary of Science that runs the labs, you have a National Science foundation, which is basic research neighborhood of about $9 billion. There is no other that does everything related to like weather and weather satellites. At Department of Commerce, there is NIST that does all the like standard setting. There's USGS that does our geological surveys, there's NASA. So I think, you know, the most exciting meetings for me are when, when sort of we convene our NSTC or National Science Technology Council, and all of the agency heads for all science agencies come to the White House and we're able to ultimately chart out, you know, if we want to drive US leadership in AI or quantum or nuclear or 6G. Everyone has a piece to play and they're not. There isn't one place that's singularly responsible for that. It's each individual agency has a piece of the puzzle. And I think that actually creates even more leverage and allows us to do even more things. So it's a lot of fun to bring all these folks together. And I think the level of PASS and this admin to kind of drive these leadership positions in our core tech areas is pretty special.
- [20:34]
A
Sure. You talked a little bit earlier about lessons from your time in the first Trump admin with the Huawei Wars. I'm curious, what other lessons have you brought from that first four years to this time? And I'm curious, especially for any differences, how does this admin look differently from the first time that you were in the role or not in the role, but in the admin?
- [20:54]
B
Yeah. The main thing I've generally seen is that the issues which our office and the President were championing in 45 have become far more prominent in 47. And the example I always give is the event had for the signing of the executive orders and the release of the action plan in July. At that event, you know, I spoke, spoke five Cabinet secretaries spoke. The President gave a speech for, for half an hour on AI. It's incredible when we want to drive an effort on a topic like AI, the challenge we have is that everyone in the Cabinet wants to participate and be part of it and bring the muscle and power and statutory authorities associated with their agencies to bear on the issue. And that creates an incredible opportunity. You have an incredible amount of political capital being spent on driving this priority for the President in a way that, that we didn't necessarily have in 45. And that wasn't because it wasn't a priority. It just wasn't as important to the larger agenda and national interest as it is today. So to me, that's the exciting part of what we have. We have extreme enthusiasm in ushering in this new golden age. And a big part of it is tech. And that's why I think this job is more fun than it ever could be.
- [22:09]
A
Sure. One thing that's changed culturally in the background between 2016 and today is, I think, what a lot of folks in the audience might be familiar the posting to Policy Pipeline, this idea, that it's easier today to get your ideas in front of people in positions of authority, that there's more of a horizontal information flow. And in the past week, we've seen colleagues of yours like David Sacks and Sriram Krishnan engaging on Twitter with small posters about these big ideas. I'm curious, what's your media intake like? Are you experiencing the. This horizontal dynamic? Are you lurking on Twitter?
- [22:44]
B
We tried to connect with all Americans as they share great ideas with the White House. As you mentioned. I have amazing people on my team that work on RAI staff that are absolutely terrific in driving that and being very good. I think you mentioned one of my former analysts in Dean Ball, who did, I think was probably one of the most terrifically connected people in the Valley. And I think he served as a great avenue for lots of people to bring ideas to the table. And I think those were ultimately manifested action plan. I think you're actually right. I think people are more engaged in having conversations with Washington and our office and the White House more broadly on tech policy than they were in 45. And you know, obviously it's a consequence of these big, much bigger issues and much more front burner. But at the same time, I think just the way that the media landscape has changed and the way that people communicate, I think it's much better information flows and ultimately I think the policy ends up in a better place because if you try to do a lot of this stuff in a vacuum, it obviously doesn't end well.
- [23:42]
A
Sure. What about talent flows? I know for all the time that you've been a public figure, you've talked about getting more technical talent into the government. Some of the challenges and the opportunities there. And obviously in this administration at senior levels, there's a lot of tech figures who have come in, including some of the people we just mentioned. Are you seeing that technical talent also come in at lower levels or are there opportunities or places where you'd like more technical talent to be entering the government? Currently, any call to action for the people in this audience?
- [24:11]
B
I'll be honest, if I was going to rank places where I wish we could do better, I would rank this one pretty highly. It continues to be a real challenge. I think the rules, regulations around bringing people into government haven't really changed. A lot of those are in statute and it's really hard, continually hard to get people in. I think Scott, who runs opm, came from Andreessen. I think he's spending a lot of time thinking about this issue and hopefully we'll have a lot more to talk about soon. But. But our hope is to try to find ways to bring more people into government. We've chatted ourselves about are there ways that you can do rotations or find ways for people to know that they're time limited, then come in, they know they're not gonna then go back to the job they had or find easier pathways to flow in and out. I see it every day and as I try to assemble my team, very talented people, if you come work in My office, you can't hold any equity in anything related to tech. So a lot of talented people have spent a lot of years building large venture portfolios, having worked at companies and asking them to divest of all. That is a really big ask. And I totally understand the rules and it makes sense that they're there. But you can imagine that there are challenges. So my hope is that we can continue to bring great tech talent in and we need it more than ever.
- [25:26]
A
Sure. Talk to me a little bit about this Edmunds vision for science because we've spent a lot of time here on AI, but obviously you've spent a lot of time in the public eye also making the case for higher standards for scientific work. Tell us a little bit about that.
- [25:41]
B
Yeah, I think for us, as many people in this conference know, I mean, the federal government has been funding basic research for many years. It all started with Vanner Bush when he released Endless Frontiers. And that led to the creation of the National Science Foundation. And I think what's really important for the government to always remember is that we have an obligation to the American people to spend that money very wisely on research that is in the public interest and is that of the highest possible ethical quality. And the President signed an executive order around gold standard science. And within that executive order you guys can look at a general tenets of things like, you know, reproducibility are in it. We want to create an environment where the research that the government funds is one that is of the highest standard and the highest caliber. And you know, when we sort of shared sort of the construct of all of these core tenets in gold standard science, it was actually quite comforting when we shared it because everyone essentially was like, yes, we want this. This is what we believe as scientists. We want the work that we do to be highly defensible, something that we can be proud of, showing that we have followed the scientific method. And ultimately that's something that's being implemented across our agencies quite rapidly since EO was signed. But I do think that sort of the larger question that comes about is around science more broadly is are we spending our taxpayer dollars on the right things that are in national interest? And I think in the early days administration, what we identified and what was identified and sort of exposed was that hundreds of millions of dollars was being spent on bad science, on things that most American people would not want their taxpayer dollars spent on. And I think that correction was necessary and important. And I think we can get to a better place. We're actually driving the Next great discoveries here in America. And back to your question of sort of how do we achieve stagnation? The dollars that the federal government spends are going to be an important part of getting us out of that rut. If we can focus them in the right areas and spend them in the right ways, that's how we can make a difference as part of that EO and more broadly on the science agenda. The one thing I think a lot about is the mechanisms by which we disperse federal funds. We have essentially been on kind of this like autopilot mode where the same methodologies of choosing which grants and who to fund have essentially been stagnant for many, many years. And the reality is that there's lots of ways to give out money. You can do fast track grants that allow you to quickly make decisions in a matter of hours or days or weeks, rather than waiting months. There are organizations that are in the private sector that could be recipients of federal funds that typically would go to, let's say, universities. So I think there's lots of things that we need to do on understanding that there are other ways to do science than the exact ways that just have been done in the last 30 years. And that could be a big driver and supporter of greater innovation in the.
- [28:27]
A
U.S. will you say a little bit more about that? I know there's a track today in this convening on these ideas, new institutions of science. And I think some of the institutions represented here, like the ARC Institute or focused research organizations. Is that the sort of thing that you're thinking about that your office is considering precisely.
- [28:45]
B
These focused research organizations are ones that can be doing incredible work and they should be able to have access to the competitions, to be able to access federal dollars. As a government, we should be laying out what the priorities are of the United States and whoever is best suited to do the science and research to deliver on those priorities should be the ones that are funded. Universities do not and should not have monopoly on doing basic research funded by the federal government. There's lots of amazing institutions out there that do that. If you look at, we look a lot about who is funding basic research. I think over the last 10 to 20 years you've seen this incredible growth in the philanthropic focus on basic research. Folks like the Simons foundation and many others are doing incredible work in funding basically faith, mathematics and lots of other work. And we again, I say this too many times, but it's so true. We have to understand as a federal government, where do we sit in the landscape of the broader funding Community. If philanthropy is doing X and private sector is doing Y, what is Z that the federal government should be doing which can fill a gap that is not currently taken by the private sector? So again, back to your point, there are lots of other models and I would love to get a readout from your conversation, to be honest. Send that to the White House. We are trying to figure out the best way to activate our science enterprise and recharge it as quickly as we can.
- [30:01]
A
Super. Let me ask you about the university side of this because I think obviously there's a lot of enthusiasm here for other kinds of science funding institutions. But the OSDP head in the first Trump administration, so your predecessor, Kevin Drogemeyer, has been working with universities as they have this conversation about how university indirect costs work, how these kind of grant systems work. I'm curious, what do you see as the opportunities for universities to work collaboratively to reform how science is conducted or funded?
- [30:28]
B
Yeah, when I talk to university groups, I think the agreement that I tend to always see in the room is that no one believes that the current F and A or overhead cost regime that exists today is one that would exist if everyone was just given a blank sheet of paper and said come up with the right way to do this. Everyone knows it's broken. Everyone knows it's wrong. The other thing everyone realizes is it's not trivial to fix it. And I think because of the high activation costs of actually fixing it and aligning it correctly, people have just not wanted to deal with it. And the lobby has been strong enough to essentially beat down any efforts to change it. Luckily, we have a president that isn't affected by that and is willing to make the important and necessary change to actually bring some level of sanity to the way that F and A is calculated and paid out by the federal government. And I think at the end of the day, you know, whether they say it publicly or not, private universities know that the current system is wrong and they are working behind the scenes to figure out a best to way, way forward. And I think what Kelvin did and what the group Able presented as a whole was the reason they even did the report was in a concession that the status quo didn't work. So I think we're taking their recommendations and our team at the Office of Management here and others are working through what the final policy will be.
- [31:44]
A
We spent a lot of time on AI and a little bit of time on science. I'm curious for the intersection there. How are you thinking about, like you said, the federal government's Distinctive role in getting science more ready for AI.
- [31:57]
B
This is probably what I'm most excited about when it comes to the future of our AI agenda. It is this intersection of AI and scientific discovery. Artificial intelligence is going to be probably the greatest unlock for science in the history of the world. This is a technology which will accelerate the way that we're able to make scientific discoveries more broadly. It's going to to increase the through discovery at rates we've never seen before. And I think what's most important about this is how we get there. How do we get to a point where we can actually leverage AI to drive scientific discovery? I'm very heartened to see a lot of private sector companies are jumping on this. OpenAI had some great announcements about it very recently. Google just has announced it a couple days ago. Obviously they did a ton of work in years past on things like AlphaFold. But the question that we ask ourselves again is what role can the federal government play in it? And one thing I always think a lot about is is the incredible amount of scientific data that the federal government has in training large language models for scientific discovery. At the end of the day, a lot of these LLMs are trained on just what's out in the open Internet. A lot of the code specific LLMs are trained on sort of like code that's out there, but the science LLMs can be trained on science data and that's far more siloed and not necessarily in the public domain. So I think there's a ton that places like the Department of Energy and even place at the Department of Commerce, there's data there that can be, that can be tapped into to drive this science.
- [33:21]
A
Can you give us a flavor of that data? What are the kinds of data sets that the federal government has access to that might be valuable for science?
- [33:27]
B
Yeah. So if you think of like what exists in the Department of Energy, there's work that's been done for many years on things like material science, on chemistry, and all this is done across all of our labs. And that's the kind of stuff that just kind of sits there that hasn't been necessarily tapped into science LLMs.
- [33:41]
A
Sure. Michael, I have to ask, we're a light haven here, which the folks who are hosting us very graciously generally tend to have a more worried view of the trajectory of AI than maybe some other participants here or you. What would you say to them about this admin's approach?
- [33:57]
B
I think you have no administration that has been more committed to delivering the benefits of technology to the American people than this one. Between me and the president and the broader cabinet, there is a deep and fundamental appreciation for the value that technology can bring to changing the lives of Americans for the better. As we write in sort of the title of our. Of our report, we are in an AI race. And the reality is that pretending that there aren't other people out there that are developing AI that is in very stark difference to our approach and our values and our. Is something we can't. We can't ignore, can't live in a world where we believe that we can fully control what happens here. And the rest of the world is just going to play along. What we need to do, and that's why the AI export program is so important, is share this great American technology. All our great companies think very carefully around how this technology is going to be developed and how it can be developed in a way that benefits all the American people. And that's the type of technology that we want the rest of the world to be using and run on. And I think there are a lot of perils if we don't do that. So I'm extraordinarily enthused about the benefits. And I don't want to wake up in 20 years when my kid's about to go to college and think to myself, wow, I wish we could have done more to make sure the US Was leading, because I think this is so important for the country.
- [35:22]
A
Sure. Well, Michael, this has been a real pleasure. Before we wrap up, I want to just ask you, we've got a fantastic audience here of folks who are committed to technological progress. Any last words or call to action for that group?
- [35:34]
B
You know, first off, I want to thank everyone in the larger community that cares about these issues. I think a lot of us have been, you know, beating the drum on the importance of technology for the American interest for many years. And I think whether it was ChatGPT or whatever it was, the moment is here where technology is front and center as one of the top three policy areas for the entire country. We have a unique moment in time where we have a present that's deeply committed to ushering in this new golden technological age. And we have a country which has never been firing at a stronger pace and rate in scientific and technology discovery than today. So I think together we can really provide untold benefits to the American people. And we are open to talk to everyone and anyone to figure out how we can move faster in this place. So I just want to thank you all for your work and know that we're trying our best here to really help as much as possible.
- [36:21]
A
Well, Director Karatios, it's been a pleasure. Thank you for joining us from the White House. If you could please give him a warm round of applause. This concludes our interview. You.
- [36:28]
B
Thank you.
- [36:29]
A
Thank you, Michael.