0:08
Michael, welcome to Y Combinator uh Startup School.
Search this video
0:10
Uh this is uh we've got an amazing uh crowd here.
0:13
It's been an amazing day and um really excited to talk about uh AI policy and your role at the White House with President Trump.
0:24
Um but I wanted to tell the audience a little bit about your background.
0:28
Um uh first of all, Michael Kratsios is director of the White House's Office of Science and Technology Policy.
0:35
He's the president's top advisor on uh science and technology and one of the key architects of America's national AI strategy.
0:46
He previously served as the country's chief technology officer in the first term uh for President Trump, where he led the early federal AI initiatives.
0:54
Uh he was the chief operating officer of Scale AI in between government tours.
0:59
So, he's seen the frontier from both inside a hypergrowth startup and inside the White House.
1:04
And so, we're going to talk about how Washington actually thinks about AI.
1:08
Um so uh I want to get in you know, you you have gone from the government to the private sector and you went back into the government, which is actually not a common thing and I think it speaks uh uh to your character, Michael, because the um you know, public service is not easy. And it is a sacrifice.
1:31
You know, you could be out making a lot more money uh doing you know, God knows what.
1:35
Your your choice of roles, given the level of connections you have, the uh your background.
1:44
Uh why do you choose to do this work?
1:47
>> You know, to to me I I fundamentally believe that uh American leadership in these emerging technologies is one of the most critical questions of our time.
1:57
For the American people to have all these benefits that AI is going to offer, we have to make sure that we have a regulatory environment that allows that that succeed.
2:04
And to me, I think even a little bit somewhat selfishly, I think there is there is um no place where you can work on bigger problems than the US government.
2:13
Even at the the biggest tech companies in the world, you're never going to be dealing with problems of this of this scale.
2:18
So, um being able to work on that is is something that that I find very rewarding and very fulfilling.
2:23
Um and I just deeply believe that we we have to find a way to keep winning.
2:26
And uh and the government, you know, can can either help or they can unfortunately kind of screw things up.
2:33
So, so being there to to try to put in the right direction is something that I enjoy doing every day.
2:38
>> Let's go rewind the clock.
2:38
How did you even find yourself in this place?
2:41
I'm really actually curious, you know, at what point did you in your life did you know that you were going to be passionate about science, technology, this this whole field?
2:50
Like, I don't know, like 14-year-old or even earlier Michael Kratsios and up until sort of the age of uh the audience, like kind of early 20s, uh even late teens.
3:03
What What You know, tell me about that part of your life and kind of how you got gravitated toward this type of work?
3:10
>> Yeah, I mean, I I I'd always been sort of obsessed or interested in in technology.
3:12
I would like follow all the Steve Jobs uh you know, keynotes every year obsessively.
3:17
I I still remember in college when the first iPhone came out.
3:21
I'm dating myself here, but it was uh it was a this like amazing moment.
3:23
We're like running around and talking to our friends about it.
3:26
Um to me, I was I was not an engineer in college.
3:29
Um so, to me, I I always wasn't quite sure how to sort of like manifest my sort of extreme excitement for technology into something I could do as a career, but um ultimately ended up in San Francisco and and worked for for Peter Thiel for for almost 7 years.
3:40
And I'm working with him and and working um with a lot of the companies that he invested in, that's when it all kind of came together where um you know, over the course of my time there between 2010 and 2017, you know, as we were looking at more and more companies in the portfolio, what kept coming up was this question about regulation.
3:57
So, whether you were thinking about um Lyft and the challenges that they were having at a sort of like state and local level, SpaceX with the challenges they were having for sort of launch permits and things with the FAA and and and Department of Commerce were doing, no matter what kind of industry you're looking at, there was this government angle where the government's policy actions could be ones that could actually unlock technologies.
4:18
So, when the president won and I had the opportunity to join the administration, it you know, my first sort of instinct was like, "How do we like look across all the rules that we have and make it easier for innovators to build?
4:28
How do we get drones flying for commercial drone operations?
4:32
How do we get autonomous vehicles on the road?
4:33
How do we get drone deliveries to happen?"
4:34
And those are things that that the government can unlock.
4:38
>> What's something surprising about how technology policy actually gets uh made and that would surprise everybody in this arena?
4:47
>> Yeah, I I think um you know, typically what people who haven't really worked in this space, what they don't realize is that these decisions are extraordinarily federated.
4:55
There isn't just like one or two people in the White House who decide something.
5:00
Um kind of a a a a blessing, I think, of the US system is that there isn't one agency that does technology.
5:06
You know, we have a health agency, you know, we have a defense department, but there is no like technology department.
5:11
Um and what that means is a lot of these these tech sort of issues are spread out across multiple agencies.
5:17
So, you have equities from national security questions you may have to um sort of commercial-oriented questions to like core science and technology research questions.
5:26
And to make good policy, what the White House has to do is bring all of these agencies together.
5:30
So, whenever like take for example drones.
5:33
If you want to make sure that the rule is right to allow for commercial drone operations, you have to make sure that the FAA has the right rules in place, but also that the people who oversee our nuclear weapons are happy so that drones aren't aren't flying over nuclear sites.
5:49
Um Help take us into like the last week or so.
5:53
Really this week felt like a noid kind of a noisy week in AI policy.
5:55
You saw the the letter, I guess, that dropped uh was that Friday morning from a lot of the uh larger companies.
6:02
Y Combinator had signed it, but also Y Combinator helped organize the letter uh that you were one of the recipients on actually on Wednesday uh evening that was uh fighting for really advocating for uh the government to not clamp down on uh open weight models.
6:19
And the the uh and the impetus for that, the energy behind it was sort of this buzz and rumor lots of kind of speculative uh uh worry that the White House was on the verge of uh doing an EO that would have kind of restricted or clamped down on on open source.
6:37
Can you talk a little bit about that?
6:39
Like how uh how connected to reality is some of the uh the stuff that you've seen on Twitter in the last uh week or so.
6:47
And then is the um you know what what is what is the White House's policy?
6:54
I I I actually spoke with uh the Secretary Lutnick last night at the Correspondents Dinner, and he said that uh you know, we White House strongly supports uh open source.
7:03
And so this is a great audience to kind of clarify what what is the White House's position on open source, open weights.
7:09
Uh can you talk a little >> so Luther over here has turned into a little mini reporter.
7:13
He's got uh he's got lots of people following his Twitter, but I I I will say I think what the secretary said yesterday is the same policy that we had on page one of our AI action plan that was released last July.
7:25
And for those of you who aren't necessarily tracking this, the the US strategy for artificial intelligence was released in in July of last year.
7:29
It was something that I co-authored with uh with David Sacks and uh and Secretary Rubio.
7:35
And one of the the number one thing, the first thing that we talk about in chapter one is a commitment to open source.
7:40
And this idea that if the US wants to lead in artificial intelligence, we have to have a vibrant closed and open source ecosystem, and that's the only way they can all work together.
7:49
So, to me, I think I think that's number one.
7:52
But, I think what what what this week really showed and what excites me about the job that I'm doing is we have to have these conversations.
7:58
And if if DC is in a vacuum isn't hearing anything from the startup ecosystem or even from big tech ecosystem or even from financial services or all these different industries, we can't make the best decisions.
8:10
So, to me, I find it inspiring and I find extraordinarily like tactically like practically helpful for weeks like this to happen because it sort of forces a lot of the community to come up and say like what do we as Americans believe in?
8:21
And we as as as policy makers can can internalize that and make sure that we're we're setting the right the right policy going forward.
8:28
>> That's the democratic process in action, I guess.
8:30
It So, every founder here is uh building with AI.
8:33
And where you sit in Washington, what do you think is the single biggest gap between how Washington sees AI AI and how this room sees it?
8:43
>> I think the um in in in Washington when people think about AI, uh it is a a a sweeping technology that covers everything from the perception of Middle America, their perception on things like data centers, um to questions about AI and healthcare,
9:01
uh to questions about job loss and how it's going to impact the labor market, um and then also there's a conversation about how it's going to impact um new company growth and and productivity improvements across across our tech ecosystem. So, I think what what
9:12
So, I think what what typically is different in Washington is you can't you you can't sort of separate those conversations.
9:19
It's very hard to like not think about, you know, the the the the labor implications of of AI boom or or not be worrying about how, you know, generally Americans don't really love data centers.
9:31
They pull horribly, they don't want them in their backyard.
9:34
But, everyone in this room knows very very well that we need as much computers we possibly can spread across the country.
9:39
So, for us I think we end up having to balance a lot of this stuff, which is is kind of is kind of our job and and but we can't, you know, we need the input from the startup community to know at least that element of the conversation and we make sure that gets done right.
9:53
>> And governments are usually regulating industries that have settled.
9:55
AI seems to be kind of reinventing itself every 6 months.
10:01
How do you write the rules for something that is moving so quickly?
10:07
>> Yeah, I mean the answer to that really is you don't want to set very firm red lines in the sand because they ultimately don't work.
10:16
And I think the best example of that in the world right now is the EU AI Act.
10:20
So, when we were in the first Trump administration, the EU went through a lot of fanfare to and spent many years putting together this EU AI Act.
10:31
And yet you this this rule essentially like regulation for Europe was passed and finalized before ChatGPT was either was ever invented.
10:40
So now going forward, all of these large language models and everything that's happened since November of '21 when when ChatGPT came out, you know, have to abide by this rule that was written before LLMs were even a thing.
10:51
And that shows how challenging sort of the the the situation is if you try to set the line too firm to begin with.
10:57
And I think our own administration has challenged with this.
11:01
The Biden administration set a specific hard cap compute threshold for if you're if you do anything above a certain threshold, then you have to do a bunch of disclosures to the government.
11:11
You know, over time having these sort of like firm red line thresholds, I mean they do not they do not, you know, stand the test of time and a lot of what we do is to make sure that that any type of of action is is can move along with the frontier.
11:24
And and what's been proven very much so by the government is that once it sets it's it's very hard to reset it.
11:30
So, we are very cautious in trying to and try to set these hard thresholds.
11:37
>> It seems like a lot of the policy conversation around AI has to do with safety and risk.
11:41
What do you think is an AI risk that builders underrate and that you think is probably overblown?
11:52
>> Um I think, you know, I think that the the risk of the day, at least today, there's there's two main ones.
11:56
I think the first one is obviously the the cyber risk presented by by the Mythos moment.
12:02
Um and ultimately the analysis that the labs had to do when they were putting out Fable on what kind of guardrails are put in place to make sure that sort of the the the the the more exquisite, powerful, you know, quote-unquote dangerous capabilities of of Mythos were were appropriately limited for for the for the release.
12:18
Um but again, the challenge with a lot of these cyber capable models is the same model that is able to sort of do something nefarious is the same model that can be very valuable in hardening an existing system.
12:31
Um so, there's these these inherent tradeoffs that happen.
12:32
Um and I think that the the second risk that always kind of comes up that, you know, you know, we think is is sort of coming over the horizon is this is this sort of the biological risk question.
12:41
I mean, I think I think my sense is I think it at the moment and it has been for many years a bit overblown.
12:47
I think people were sort of like shouting about the bio problem back in '21, '22.
12:49
Hasn't been an issue for 3 years.
12:53
Um but I think we definitely definitely need to build the right infrastructure to run the right test and eval processes on models as they as they creep, you know, past the frontier. >> Got it.
13:04
Um a lot of tech policy conversations end up dominated by the five biggest uh tech companies.
13:10
And I should say this is one of the reasons that uh Y Combinator with a handful of other companies a couple weeks ago helped spin up a a trade association called the Little Tech Association. >> Mhm.
13:21
>> Um because it seems like sometimes, you know, I live in Washington D. C.
13:24
you know, that there's you do as well obviously that they we're actually on the uh the flight this morning uh uh at at 8:00 together.
13:36
Um I think that uh it feels like kind of uh Google, Apple, Facebook, Amazon Truman Show sometimes in Washington that like there's every you know every position paper you read or speech you hear is like you you think gosh that that has some some big tech influence behind it.
13:53
How does the two uh two-person company even have a fighting chance if if those large companies are writing the rules?
13:59
How does the White House kind of make sure that the kind of company that's sitting in this room uh get's heard?
14:07
>> Well, the most important thing that I do is talk to you as often as I possibly can.
14:11
But, I think I I think the reality is there there's there's a great number of institutions that that help support and stand up for for for little tech.
14:19
And I think you know, I think generally government is quite is quite cognizant of of of sort of the influence of a lot of sort of like large players in in every industry.
14:28
And I think this isn't a problem or this isn't a situation that is uh is is only occurring in tech.
14:32
It happens in everything from, you know, energy to health care to whatever it may be.
14:40
And I think what we try to do very hard is there's lots of mechanisms where you can get input um from a wide variety of folks.
14:46
So, the traditional sort of request for proposals or you know, request for information process where you can get anyone to kind of submit comments um And and for us I think I I I think it it's actually working.
14:56
I mean, if you if you think about one of the one of the major priorities that you know, the president him spells himself spoke about in the speech that he gave when our strategy was released in July of last year was this question about um about uh preemption of state laws around AI.
15:13
And and and and the fundamental um thesis there is that if you create a or allow for the creation of a patchwork of regulations, meaning there's one set of regulations for AI in California, another one in Maryland, another one in Texas.
15:27
You know, the big tech guys, they can deal with that.
15:29
I mean, Google can harm can hire an army of lawyers and they'll figure it out and they'll be fine.
15:34
Um but for all of you out here, that's just that's just not going to work.
15:37
And the president stood up and say like, "No, we have to have one national standard for AI so we can make it easy for anyone who wants to build an AI company to know what the one set of rules is and build their company like that way."
15:48
So I think that's an example of of us really taking that to heart.
15:53
>> Well, I will say I I uh commend the White House for kind of continue and really it started with uh the first Trump term that uh the the bipartisan antitrust project with respect to uh Big Tech.
16:04
US V Google was started under uh the first Trump administration.
16:07
Uh US V Apple is being continued.
16:09
Hopefully the the reports that that might uh settle in imminently are are incorrect and uh FTC versus Facebook, uh FTC versus Amazon, uh these are cases that uh that uh the the Trump uh DOJ and FTC have continued and I think it's uh that coupled with the president's own uh usage of the phrase Little Tech and championing Little Tech companies.
16:33
I I think the the thing I think yeah, when probably when we met years ago, I said we just have to make sure that Little Tech has a seat at the table.
16:40
And so I just uh as you go back to Washington, kind of you know, making sure that we're we're doing everything we can to to make sure that that Little Tech has a seat at the table. >> It's important to me.
16:51
I mean, I spent mo- most of my 20s here in San Francisco working at a venture firm.
16:55
I mean, I I understand how challenging it is to to build these companies and and I think, you know, what what makes the US so special is that there is no place in the world that is better to do a startup than America.
17:09
It is easier here and it um and the opportunities are just unbelievable.
17:15
So, as as a public servant, what I think about every day is how do we protect that ecosystem?
17:18
How do we make it even easier?
17:20
How do we create all the opportunities for all of you to create great companies and succeed?
17:23
Everyone in the world is clamoring to come to America to build companies.
17:26
There's something very special about here that we must preserve.
17:31
>> Is there a version of AI regulation that protects consumers, but it doesn't hand the incumbents a moat?
17:41
And and what does that look like concretely?
17:43
>> I I I think back to what I was saying with the sort of one what like one national framework.
17:46
I think it's all about having regulatory clarity and ease for all sizes of of companies.
17:54
Um and if we can work with Congress to do something like that, I think that would be that would be the biggest boon for for for startups.
18:01
>> The United States' advantage has always been that anybody can start something. >> Mhm. >> To your point.
18:08
I mean, we are uh unique in that respect.
18:09
This is the greatest place uh in the world to to build a company.
18:14
Uh what would you tell a founder here who is worrying about compliance costs and that locking them out before they even start?
18:22
>> You know, I would tell them that there's probably no administration history more committed to reducing regulations than the current one.
18:29
Um our sort of uh Office of Management and Budget, which sort of that runs kind of our regulatory process, the guy was running it is this guy called Russ Vought and he his life's mission has been to eliminate as many regulations as possible and that is what we think about every day.
18:42
To me, in the tech standpoint, you know, when we talk about regs, you know, the biggest question to me is how do we remove barriers to innovation?
18:51
I talked a little bit about this.
18:51
To me, there was a you know, there's kind of two ways to think about uh about the world of regs.
18:56
There and about technologies.
18:57
And sometimes a lot of people talk about this in Washington.
19:00
There's there's there's technologies that are either born free or born in captivity.
19:04
And for each of those categories of technology, you have a different set of regs to look at.
19:08
So, born free technologies are ones where there aren't regulations on the books.
19:14
Things of like the internet when it just started.
19:15
And those are the types of technologies you have to preserve.
19:17
You have to be very very very thoughtful about whether or not you're going to introduce new regulations into that domain because people can sort of build anything and strive in those areas.
19:26
So those born free technologies you want to preserve.
19:30
And then the born in captivity technologies are technologies where you're building something but you can't commercialize it or you can't take it out to market unless you get some sort of government approval.
19:39
So think about um commercial drone operations.
19:41
You guys could go build an amazing drone in your backyard.
19:44
You could build the most amazing software to connect sort of a uh a vendor to a to a customer and set it all set up.
19:50
But to actually close that transaction legally and have the uh and have the drone fly, you can't do that unless you get a waiver from the FAA.
19:56
So those are the types of technologies that we will relentlessly think about how to sort of remove those regulatory barriers or make it much easier to do so.
20:05
I think earlier some of you may have heard um from the founder of Boom Supersonic.
20:10
You know, I've been obsessed with supersonic flight for years.
20:13
I think it's the most obvious sort of like in your face example of technological stagnation.
20:16
We had the Concorde flying years ago.
20:19
We're flying slower than we were back then today. Absolute tragedy.
20:22
So in that situation, you know, he can't get his his supersonic plane to fly unless the rules are set such that there is a noise limit instead of a speed limit um over the United States for example.
20:36
So in those kind of born in captivity technologies that's what we relentlessly try to figure out how do you kind of remove those barriers and make it easier for these technologies to work.
20:45
>> I'm I want to go back to a question that I've I've I want I forgot to ask at the beginning which is like about your day in day out role.
20:52
Like you're not only a uh the director of the Office of Science Technology Policy and maybe you can talk a little bit about this when I when you answer but also a special advisor to the president.
21:04
So like what is I mean there's probably not a typical day, but what is like if you had to kind of average the days across, like what is a typical day look like?
21:11
What time do you show up to work?
21:13
There's a lot of people don't realize in the White House there's like the White House is a sprawling complex, so there's like where the you know, there's the Oval Office obviously, but then there's uh something called the EEOB, the Eisenhower Executive Office Building, like where how big is your team? Like >> Yeah.
21:30
>> What is it just kind of help us visualize what it's like working at the White House and what your job is actually like day-to-day and are those is the is the role of a uh director of the OSTP >> Mhm.
21:43
>> is that always a special advisor to the president?
21:44
I think that's like a kind of an extra additional role you take on.
21:47
So, talk about that a little bit.
21:49
>> Um so, I think maybe the most abstract way to think about it is and I mentioned this a little bit earlier, the federal government is made up of all these agencies.
21:56
You have HHS which does health care, you have Department of Defense which does defense, you have Department of Energy that runs kind of energy in our national labs, and um for any type of policy that you do, you essentially have to get concurrence or have some sort of conversation among all of these different agencies.
22:11
And there's only one building in the whole world that can get all these agencies to come together and have that conversation and bring some resolution, and that's the White House.
22:21
And generally there's there's sort of four categories of policy that get sorted out in the White House.
22:26
There's national security, then there's a council that runs that process, there's domestic policy that's sort of like health care and immigration type stuff, and there's another council that runs that.
22:36
There's economic policy that does sort of like tax and other like econ policy, and then there's us that does science and technology.
22:42
So, essentially there's sort of four policy councils and there's sort of four policy leads and each of us kind of run policy processes on the topics at hand.
22:49
So, in our portfolio we obviously have AI that we talked about a lot, but we do things like like quantum, like civil nuclear energy, biotech, space, you know, and for each of those portfolios on my team, I have one or two people that help run that portfolio.
23:05
So, I have a space team of about three guys, and they coordinate space policy across the government.
23:08
So, they bring NASA in, they bring the Department of War that has a bunch of satellites for national security purposes, and they all sit together and kind of sort out sort out the policy.
23:18
So, I would say on a typical day, you know, the the general things that you do are one or sort of stakeholder meetings.
23:23
So, Luther comes and says, "Oh, hey guys, you know, you got to look out for little tech, da da da da da." And I listen to him.
23:27
And then and then go >> Don't settle the Apple case.
23:31
>> [laughter] >> So, there's [clears throat] there's a kind of the stakeholder stuff.
23:33
And the second category of work is just the blocking and tackling of doing policy.
23:38
It's like work the the president has said, you know, like we, you know, have to you know, have to make sure that commercial drone flight is happening.
23:46
So, like let's figure out how to do it.
23:48
Let's push FAA to change this rule, you that kind of stuff.
23:50
Bring the agencies together and do and do that that kind of work.
23:54
>> [snorts] >> Um so, those are kind of the two big big buckets of stuff.
23:56
And then the third is doing things like this, like sharing the president's message and talking to to people around the country, understanding kind of what their challenges are and and trying to see how we can be how we can be helpful.
24:06
To your point on the director versus the the assistant to the president, um the uh with within the White House, there's um there are folks called commissioned officers, and they sort of have three three ranks.
24:17
There's an assistant to the president, which are sort of the most senior advisors to the president.
24:23
There's about 20 or so of those people, and and that's what that's the title that I have, but that's one that's sort of the the chief of staff has and others.
24:29
Uh and then there's there's deputies, and then there's specials.
24:33
There's kind of this like hierarchy where we all kind of flow up to support the president.
24:36
>> So, practically, what does that mean?
24:38
Can you just wander into the Oval Office whenever you want?
24:40
How often are you interacting with the president?
24:41
And and like how does it how does it function?
24:45
Because it it is like I think people, um you know, don't necessarily appreciate how many people work work at the White House.
24:52
>> so I think um not getting too much in the weeds and mechanics, but I as I mentioned, there's there's these policy processes where you bring all the agencies together and you start working on a problem.
25:00
And you can imagine that is kind of the bottom of the pyramid and they try to like sort out the problem.
25:06
And if they sort it out, then it's great.
25:08
Then policy is over and it gets executed.
25:10
If there is disagreement and like someone's like no, no, no, no, like this making this up like some guy at the Department of War is like no, I don't like this drone policy because I don't want drones anywhere near military sites.
25:19
That has to be way more stringent.
25:21
So it's like if they can't agree, then it like kicks up to the next level.
25:24
So then you can imagine like more senior people at all the agencies kind of chat.
25:27
And if they can't agree, then it gets up even higher.
25:29
And then if it if they can't agree, then it ultimately goes up to to the president for a decision.
25:34
So what you try to do is like limit the the decisions that get to the president because he has limited amount of time and he should be focused on the things that are most important to to the country and to and to the national priority.
25:44
So for us when there's certain high stakes, you know, AI or technology issues that that the president needs to weigh in on, um we bring them to him.
25:50
We have a conversation and he and he weighs in and makes the ultimate decision.
25:56
>> Well, part of the reason I I I ask this uh and I wanted to kind of unpack that.
26:01
I wanted people in this audience to appreciate how busy Michael is.
26:04
>> [laughter] >> Because I mean, truly we owe a debt of gratitude to just uh the amount of the amount of public service and the fact that we we got some time with him today uh because sometime in in in sort of managing all of that uh flurry of uh activity, you've just released a like 150-page report.
26:27
Uh I want to talk about uh science and new golden age.
26:33
Um this came out this week. >> Yep. >> So tell us about it.
26:38
>> Um Well, maybe we can start with just a little bit a little bit of history because I think that's what sort of inspired us to to write this report.
26:43
Um in in 1945, uh FDR wrote a letter to his science advisor.
26:51
Since the person who had my role was that called Vannevar Bush.
26:55
In that letter he essentially asked the science advisor what should we do and how do we how do we approach the science ecosystem after World War II?
27:02
If you can imagine during World War II a lot of the the energy that the federal government put into the science and tech ecosystem was around getting the nation ready for the war.
27:13
So a lot of money was spent in launching the Manhattan Project and all this other stuff.
27:17
And and Vannevar Bush replied back to to FDR with with a famous report called Science Endless Frontier where he essentially said the government has a very important role to play in funding science and technology in the national interest and particularly in funding early stage basic research.
27:34
And he made the point that you know there is stuff that only the government is going to do because the private sector isn't incentivized to do.
27:41
And essentially that report laid out how we've been doing science as a country for the last 70 years.
27:50
But what's so but times have changed pretty dramatically.
27:52
Back then around like 1950 almost 70% of all R&D was being funded by the federal government.
27:59
So they kind of had monopoly over where the money was going to go and only about 30 or less was done by the private sector.
28:05
Over time that has sort of inverted completely.
28:07
The private sector plus philanthropy now do about 70% of R&D and the and the government only does about 30.
28:13
And just if you think of like pure basic research, kind of stuff that you see at universities there's almost parity between the federal government and private sector now.
28:19
So the system has fundamentally changed and the actors within the system have changed dramatically. All of you guys exist.
28:25
You guys are doing incredible work in the as as as startup founders.
28:29
You have these organizations called FROs which are a focus research organizations that sit outside of universities that do their own research.
28:36
And because that system has changed, you know, we as an administration believe that we have to re-look at the way that science and technology is done this country to get the most out of it.
28:46
And when I uh was confirmed by the Senate for my role, President Trump wrote me a letter, very similar way that FDR did, and and asked me a couple of questions around how can we revitalize and re-energize the science ecosystem.
28:58
And we spent the last year writing a report um back to him on on on what we can do.
29:03
And there's a couple of of of main themes that we can get into uh around that, but the the core sort of thesis about it that I think applies to you, one of the one of the main pillars of it is around this fundamental belief that artificial intelligence is going to transform the way that scientific discovery is done in this country.
29:22
If you are working on material science, if you are working on pharmaceuticals, if you are working on chemistry, you in 2 3 years or even today are doing your role as a scientist dramatically different because of artificial intelligence.
29:34
And we as a government that spend almost 200 billion dollars a year on funding R&D need to be aware of that and prioritize that so we can make sure that our ecosystem is putting out the best possible research in the world.
29:49
>> Um so the last chapter of this report sketches this kind of like almost sci-fi picture.
29:56
Um and yeah, you're talking about AI agents uh posting bounties for experiments, contracting robotic cloud labs, settling results on a ledger, and then the budget memo gets really concrete.
30:09
Fast grants decided in under a month.
30:13
Prizes built for three to one private leverage.
30:15
Every major agency filing an an action plan within 90 days.
30:20
And you write in this report that the government's job is to shape the arena, not direct discovery. >> Yep.
30:25
>> So, for the thousands of people we have assembled here, what what's the piece of that vision you're hoping a couple of 20-year-olds build because the government cannot or should not?
30:38
>> To me, I think the the infrastructure that is going to a you know, support the scientific ecosystem in the United States is going to fundamentally change over the next few years.
30:51
The idea that you can have autonomous cloud labs running experiments on a loop without human intervention, testing hypotheses, running the experiments, seeing the results, creating a new hypothesis, testing it and running it, and ultimately getting to a conclusion, that is in our sights.
31:05
But for all of that to work, there's lots of things that still need to be done.
31:09
We have to get robotics perfected so that you can create these labs.
31:12
You have to create the right software ecosystem to be able to like think through the next hypothesis and create the next hypothesis.
31:18
So, to me, I think there's almost a sort of like infinite category of work that can be done to create this this world of autonomous experimentation that all of our scientists across so many domains are going to be leveraging to make these discoveries.
31:32
>> I mean, you alluded to some of the the things that I think you're looking out toward into the future, uh but beyond AI, what technology do you think that Washington will care enormously about in 5 years that it barely discusses today?
31:49
>> Well, we discuss it a lot, but I don't think it gets as much airtime as it should, and and that would be that would be quantum information science technology or quantum computing.
31:54
Um to me, I think, you know, back in in the first Trump administration, uh you know, in 2017, 2018 era, you know, I was as I was I was chief technology officer United States, and in that role I was I I remember kind of constantly running around the West Wing trying to convince people that artificial intelligence was an important thing.
32:14
And like, maybe once every couple months some journalist would like be nice enough to write an article about AI and what the government was doing.
32:23
>> This is like 2017, 2018. >> 2018.
32:26
And um and you know, I give President Trump an incredible amount of credit.
32:30
I mean, he stood up and said, "I'm going to sign the first executive order in the history of the United States prioritizing artificial intelligence uh in February of of 2019.
32:38
And that essentially set out the first national strategy in history on on AI.
32:43
And through that, we created the first kind of like regulatory thinking around how our agencies should be contemplating air power technologies.
32:51
Um but, you know, times times have really changed.
32:53
And I think the effort that we put in there to essentially double the amount of R&D that we were putting into AI in our in our budgets then, you know, and you fast forward through 4 years and and we have ChatGPT and everything's exploded now.
33:03
You know, obviously what we did in the administration wasn't necessarily the reason why ChatGPT happened or whatever, but I think we've spent a lot of work trying to prioritize and prime the the S&T ecosystem to be ready for the moment, and I think it was.
33:15
And if we paralyze parallelize that to today, I think that same moment is happening with with quantum.
33:23
There are truly some some basic fundamental scientific questions that still need to be answered on how quantum can can sort of um uh be applied to things like computing and to sensing, and that work goes on.
33:32
And you know, the president just signed an executive order on on on quantum information science, and we're going to do a ton of work over the next few years to prioritize it.
33:40
He set a pretty ambitious goal for the Department of Energy to build a scientifically relevant quantum computer by by 2028.
33:49
My sense is we're going to wake up in 3, 4, 5 years and and kind of see the progress we've made.
33:55
>> We've talked a lot about the uh you know, the executive branch and its authority, executive uh executive orders being used to um shape AI policy.
34:06
What do you think Congress could be doing better on uh these issues?
34:10
Cuz it seems like you know, the the mere fact that we uh we we read so often about executive orders in AI might imply that, you know, our legislative branch could could be doing a a better job.
34:22
>> Yeah, I think that the congressional stuff is a little bit a little bit tricky with AI.
34:25
I think that the challenge is that we spoke about a little earlier where you don't necessarily want to want to set rules of the road too early that that end up, you know, hurting the industry rather than supporting it.
34:39
The thing about executive orders, which is a little secret, is when administration changes, you can just revoke the executive order and and start from scratch again.
34:48
Um when something is a law, you you can't do that.
34:51
It's it's the law and that's that's the law of the land.
34:54
So, you know, we work very hard to to try to find places where we can collaborate with Congress to set rules that are actually pro-innovation and help the country.
35:02
And the White House put out a set of a set of sort of legislative proposals to to Congress earlier this year that walk through a couple areas where where we think, you know, a lot of benefit could happen if it was in statute.
35:15
But it's a couple times, I'll say it again, you know, we would urge Congress to to be able to pass some sort of some sort of law around around preemption so we don't have this crazy patchwork of all these different states doing all these different things these things on AI.
35:27
I think another area that a lot of Americans want to see Congress step up on is is how AI interacts with intellectual property and with and with sort of name, image, and likeness of certain Americans.
35:44
I think there's a lot of creators out there that are are worried about how AI is going to impact what they do.
35:49
And I think all of us, I mean, you know, we all have a craft of some kind.
35:52
You know, some people are singers, some people are writers, you know, so all of you are unbelievable coders.
35:58
All of you have sort of skills and talents and and having some protections and knowing around that is kind of is kind of important and something we've been talking about.
36:06
>> Yeah, where do you think what is the sort of the White House line on that?
36:07
If you had a magic wand and you could sort of solve the IP issue, how would it how would it be described?
36:14
Cuz I I feel like I've spoken with people from like the traditional kind of media industry that are just, you know, filing lawsuits against AI companies left and right.
36:23
And And on the other other end of the spectrum, I've talked to kind of pure maxis that say, you know, that that uh you know, you should not even be able to opt out of uh a being crawled for the purposes of training.
36:38
And is is could you talk a little bit about maybe unpack that issue a little bit?
36:42
Um >> Yeah, I I think the one area that I've been clear on is the is the model outputs.
36:47
And if you create a model that is is outputting uh Mickey Mouse, >> Yeah. >> you know, not cool. Can't can't do that. >> Wait. >> [laughter] >> What?
36:56
That's Walt Disney and and they shouldn't be allowing people to to output that.
36:58
So, I think I think kind of making that uh more clear and in in in in statute, I think is I think is very important.
37:04
I think there's also a lot of things that um that the industry can kind of like um self-coordinate on a little bit on finding interesting um uh kind of like re- revenue sharing models for for creators.
37:16
There's a lot of creators out there that would love to license their their own personal IP to to um to to companies that can do all sorts of stuff with it, whether they're musicians or anything else.
37:29
>> I actually just read this morning my old company Yelp did a deal with ChatGPT where I guess they're outputting some of the reviews for um so, yeah, you're starting to see a lot more of that. >> Ex- exactly. And look, it's early.
37:37
I think I But but but I think over time that that that market is going to is going to mature and I think that's maybe a category of places where you probably don't want to legislate too fast.
37:45
So, you got to kind of let the let the market kind of sort itself out.
37:47
Um but but I I do think it's important cuz Americans generally generally do care.
37:54
>> Well, to to close us out, I one final question, which is that you were one of the youngest people ever in your role.
38:01
To the 20-year-olds in this room who care about technology and how it shapes the country, should some of them work in government someday?
38:10
And what would they find there?
38:11
>> You know, I I would recommend to all of you if there's if there's a moment in your career or a moment in your life where um you can take a role in government, um I guarantee you they will you will never feel or have a moment where it's more fulfilling and more rewarding.
38:26
I think it can be a slog, it can be bureaucratic, it can be painful, but when the when the outcome actually happens and you deliver the result, there is no place where you can have a bigger impact on on this country.
38:38
And I think all of you are here, you're here to build companies, you're excited about building new things, about hiring more Americans, about building technologies and amazing things that will touch the lives of so many of our of our fellow citizens.
38:51
Um and there's some flavor of that in government.
38:54
In creating sort of the environment that can allow startups to thrive, that can allow new companies to be built, I believe there's something very fulfilling and rewarding about that.
39:04
And in being able to do that in the service of kind of your fellow Americans is is something I'd encourage all of you to do.
39:10
And if any of you want to work in the in the White House Science and Technology Office, you you just look us up because we're always looking for good people.
39:18
>> Uh well, Director Kratsios, I just have to say as uh Y Combinator's head of public policy based in Washington, um I really appreciate uh the accessibility of your office and uh the administration and the thoughtfulness on these issues.
39:34
And I also really appreciate you coming out and speaking to Y Combinator AI Startup School today.
39:38
Let's give it up for Director Kratsios. Thank you. >> Thank you.