Good afternoon everybody. My name is Martin Giles. I'm the executive director of the Hoover Institution's Technology Policy Accelerator and a policy fellow specializing in emerging technologies. I'd like to kick off our session by just sort of citing a few statistics. More than one in five workers today are using generative AI in their jobs. Across the economy, adoption is accelerating and about 75% plus of companies now use generative AI in at least one of their business functions. And by some estimates, the use of that technology alone could add between 2.2 trillion and 4.7 trillion dollars a year to global growth. And that's the equivalent of a G7 economy. So we have gone from what was an academic exercise to widespread use in a relatively short time. And AI more broadly is shaping everything from our economy to scientific discovery and national security. And the rapid progress that we're seeing makes AI one of the most important and one of the most challenging areas of technology policy. It's something we at Hoover spend a great deal of time on. I just came back from Washington DC and every question was around AI or AI's application. And I think for policy makers it's a hard thing because they're trying to strike a balance. It's a balance between encouraging innovation and scientific leadership through initiatives that help business, and we'll talk about one of those, Genesis Mission, on the panel. And at the same time they don't want to get in the way of private industry developing its own solutions. They want to protect the public from potential harms from AI while at the same time ensuring that the US moves fast enough that it remains globally competitive and they want to ensure that AI's benefits are equitably spread while at the same time understanding that there will be disruption and we are going to see significant changes in the labor market. So, it's a really fascinating area and I'm delighted that we have two great panelists to talk about some of these issues. Joe Lonsdale on my far right, left for you, is the founder and managing partner of 8VC. Did I get that far left, Joe? See, I did that seating plan deliberately. Okay, that's all I'm going to say. I'm not going to say that ever again. A leading venture capital firm managing 6 billion dollars in capital. In 2003, he founded a little-known company called Palantir Technologies, which you may have heard of, now a global software company supporting the work of the US and its allies. Since then, he's founded more than a dozen other prominent companies, including Adapar and OpenGov, and he's building more through 8VC, including two recent billion-dollar plus defense businesses, Epirus and Saronic. In 2018, Joe also set up the nonpartisan Cicero Institute. Apparently, he tells me I should be pronouncing that in Latin, but I'm going to, as I'm British, I'm going to get away with that pronunciation. Which advances policies to promote effective and accountable governance. And in 2021, he became the founding chairman of the board of a new university, the University of Austin. To my immediate right, your left, is Jeff Von Melson. Jeff is the co-founder and CEO of Laya Sciences, a company whose mission is to build scientific super intelligence. And he will talk about what that means to solve some of humankind's greatest challenges. He's one of the most prolific scientific entrepreneurs of his generation. And Mike, I think my microphone's gone out so you will hear me scream at a rather loud voice, but I hope you can hear me. Put your hands up. He's one of the most prolific entrepreneurs of his generation and is listed as an inventor on over 200 patents and pending applications. Is that like 200?
The technologies he's developed have helped microphones work and also have helped companies integrating science and human health and sustainability. And he has created companies that have achieved over 10 billion dollars in public and private market capitalization. So, two very, very impressive panelists. Let's get right into this. You know, I mentioned the Genesis Mission at the beginning of my opening remarks here. The Trump administration issued an executive order late last year launching what it calls a Manhattan Project for science. And the idea is to harness federal data sets and compute across both the public and private sectors to advance innovations in 26 different areas ranging from nuclear energy to quantum technology and automated AI labs for accelerating the design of new materials and drugs. Jeff, that's kind of what Laya's doing, isn't it? Can you talk to that? And what is scientific super intelligence?
Sure. First of all, it's awesome to be here. Huge admirer of stuff Joe's built and Martin and I have got to know each other a little bit too. So excited about this discussion. So the preface to Laya was three to six years before ChatGPT was launched. Technical members of our founding team in Laya had been applying generative AI to publicly available data sets in science and it found that across life, materials, and chemistry, with the right quality and quantity of data, a transformer model could become a form of higher intelligence in science in that they could make predictions and solve problems in ways that the best human minds with all of the best available tools have never been able to do. Many of you will have heard of DeepMind's Nobel Prize-winning contribution to science called AlphaFold. That should be thought of as the same. It's folded millions of proteins that eluded the best human minds with all of the best available tools. So, we concluded that was the evident future of all of science, not a few exceptions within science, and that a very general approach could be taken to breathe those into existence and to expand upon their breadth to create a kind of model that becomes a polymath and genius across every field of science. I believe this is the most consequential period in the entire history of US science. That is not to underestimate any prior periods of US science, but none of our forebears ever turned up scientific intelligence. And nobody in this room should feel comfortable with the idea that the US might let one of our rivals have a higher scientific intelligence than we and our allies. That the US will kind of go away on some time scale if that transpires. So across our team at Laya, we have a very intense sense of mission that this be achieved so as to create a new kind of scientific superpower for the decades to come. The Genesis Mission is, I think, a very prescient realization by this administration of the same, that the way that science funding has been organized thus far is effectively based on the human paradigm of science that small siloed teams working in isolation can move each frontier forward. From here forward, scale is the friend of scientific intelligence. And if we silo all of our science, we will have a lower scientific intelligence than the rest of the world. So I think this is the beginning of what I hope is in fact a Manhattan Project scale of investment because for the first time in the history of US funding science, we aren't simply fertilizing the output of fixed human intelligence discovering and making progress. Instead, those investments go towards building higher intelligence at the same time.
Got it. Is it enough? I mean, when I look, you know, we've just kind of announced, I think, 320 million dollars of public-private partnership through Genesis. How much have you raised so far at Laya?
Announced fundraisers are about 550.
550 for one company. We should rename your company the Massachusetts Project, I think, because that's a hell of a lot of money. I just wonder whether we are moving fast enough and should we be more aggressive? Joe, do you think Genesis is big enough? Should we be doing much more?
Well, I also support Genesis. It makes a lot of sense that science really should accelerate right now. And I've been seeing a bunch of these models trying to do things in different areas. And it's actually really cool. I think a lot of times like you'll say, 'Okay, I want the protein to do this instead.' And the model will figure out something that works. And hopefully we can figure out why the model did that. Sometimes science is advancing in ways where we're not even yet understanding like how it's using the data. And so this is definitely a new age of being able to discover and figure out things in a lot of different areas. You know, most of that that's been done well lately in the last 10 years has mostly been through companies. I think that the way you're doing it is like a really good way to do it. I have friends who've raised billions of dollars for bio companies working on new mechanisms for aging and all sorts of really exciting stuff that we're pushing forward in these companies bringing people together. You know, the government has played a really positive role in the past with innovation. I think the quality of people in and around the government and the competence has gone down a lot since the mid-20th century. In the mid-20th century, you had the total war where everyone had to work together and you had a situation where innovation was very, very expensive. So if you were really bright and want to work on the hardest things, you basically had to go to the government or Bell's, but one of the few big institutions. And so you had many, many important things achieved and we all kind of know the history and hear about these great things. Part of the reason that hasn't happened in like 1990s or 2000s or 2010s is that the brightest people can instead go do their own companies. They can, I'm invested in helping build things that have raised billions and tens of billions of dollars. There's great people in there making the most biggest advancements. So I think anything the government does probably needs to work with our top companies. I'd hope it would be working with Jeffrey's company, with my friends' companies because that's where the top talent really is. But is this a great time to bring together resources and push forward science in new ways? 100%.
Got it. One of the promises I think of AI particularly in relation to sort of innovation is changing sectors that might have been pretty sclerotic over time. I became a US citizen in 2013, proud American immigrant, but when I encountered the US health care system for the first time, I thought of changing my mind. And I'm really hopeful that AI can actually transform fairly kind of well, highly regulated and slow-moving sectors. Joe, do you think, are you optimistic that it can really crack healthcare and what policy aspects should we be thinking about in terms of making sure that happens?
No, thanks Martin. I think this is a really important question for all of us. So it's really clear right now that productivity is going to go up a lot the next few years. Like this is not speculative anymore. You know, there's trillions of dollars worth of industries, over 10 trillion now, where we've seen productivity double in certain cases. What's very clear it's going to be able to extend the whole thing. So, you're going to double or triple productivity across the board. I was just with my friends last night in San Francisco who are all running some of these AI companies and it's like it used to be that they were doubling what they could do without human intervention every year, then every six months. Now they're doubling every few months. It looks like that's speeding up. It's just, it's exponential right now. It's like really amazing what's happening. You talked earlier about how we have to make sure these gains are shared equitably. I'm in a very lucky position where like my friends and I are building all this and we're going to be creating hopefully, looks like trillions of dollars of wealth. I think one of the best ways we can share that with everyone is making health care costs and making other areas of cost disease go down. And in order to do that, you have to legalize innovation in those areas. You have to legalize competition in those areas. Right now, our health care system is a crony capitalist system where every group, if you're the doctor groups with AMA, you have scope of practice laws banning nurses and others from using AI. If you're a health system, you ban people from competing in other ways or use your market power with the payers who also want costs to go up to exclude other competition. So, if I want to build something with AI that's better, that has better results for patients, much lower costs, it's illegal in every state right now. And so...
So what's the solution? I mean...
We have to pass. I mean, I was literally this morning on the phone with a doctor of the Texas State Legislature who's a really smart guy who's been pushing on this and he wants, you know, we've written a piece of legislation. We're pitching in six or seven states right now that's like an AI healthcare sandbox where how can we, you know, responsibly allow new workflows in primary care, for example, to be tested and tried. Utah has one of these workflows. They've allowed, we've identified over a hundred. The question, should we do it top down, bottom up? We have to have states that have some courage to overcome the special interests and pass these. So, I'm trying to get more of my friends in tech and otherwise, let's fund these efforts to make it legal to build in these areas and not to be crazy and just try everything all at once and rip up everything, but we need to prove and try things and show what works that we can start lowering the cost.
That's really sandboxes. Jeff, thoughts about this. You've done a lot in health.
So, I'll rewind three and a half years in that in the years since people have been asking me frequently, when is the ChatGPT moment for science or for healthcare? And I was describing in a conversation earlier that I've reacted that that's the incorrect question often. Looking for something that resembles something in the past is generally incorrect. But ChatGPT when it was first launched was fascinating and not very professionally useful yet. Instead, the moment to look for is the Kasparov versus Deep Blue-like moment that the world is experiencing with Claude Code right now. Where there is a new intelligence that is going to forever change a profession. And a version of what Joe was describing is like across all of healthcare, we're going to need the Kasparov versus Deep Blue or Deep Blue versus the very best, you know, bracket a category of human-centric decisions that are made to facilitate the care for patients, the discovery of medicines for patients. And just to give you an example of how I think this is going to so marvelously be restructured is just imagine how much more impactful the war on cancer would have been if it was millions of agents with coordinated behavior where every experiment taking place in every battle on the frontier that comprised that war made every other experiment wiser and more impactful.
Yeah. The FDA should be letting us use all that data. For example, it's very promising on the scientific front. There's the FDA, it's a longer conversation, but it's a huge mess right now. It takes 500 days to do IND. It takes 200 days to do P&D. China is just wiping the floor with us by going much faster. So, there's a lot of things we got to fix there.
Great. Talking about China wiping the floor with us, I just came back from DC, had a week of meetings with senators and representatives, and one of the subjects that came up was open-source versus closed-source models. And is there a risk? The question came to me that what we're going to see in AI is kind of almost like a deja vu all over again with Huawei in 5G. I.e. the US leads in kind of the very frontier models. But China is creating a whole bunch of open-source models good enough, cheap enough that they are going to get adopted and used much more widely than our models. Joe, thoughts?
Yeah, I mean there's this interesting kind of sovereign AI competition angle. It's very annoying to me that we're not very good at stopping them from stealing stuff because this is the thing about China. You have to understand they both have people who are very talented and they have a system that is actively sending over thousands of people and is actively stealing things very aggressively and we've caught them every which way every time. It's beyond doubt. We have really direct evidence of this. And so the question is like how much should we let them get away with in terms of that stealing and then how should we use tariffs or use other strategies to fight them around the world. I do think it's probably not ideal for them to control the intelligence behind systems all around the world. That's a little bit scary. I mean it's a little bit scary for anyone by the way to have that power but I'd much rather it was US companies in some form and maybe there's some transparency rules we should be creating for those companies too. Definitely shouldn't be CCP though.
I'll also broaden this to the open source versus closed source is just like one fold of the issue. And to derive comfort based on US leadership in AI models over the past four years I think would be a mistake in forecasting the next four years in that China has set the stage marvelously with the scale of energy investments that they've made, the cohesiveness with which they're now entering the AI race. So, none of us should assume that US leadership is a given here. And of course, in the realm of science, you know, our belief is that whomever is running the scientific method most cohesively at largest scale will have the highest scientific intelligence on Earth.
