New White House Funding Blueprint: a Focus on AI

This episode examines a new Trump administration report outlining its vision for federally funded research and development (R&D), with particular attention to AI. BRSL Faculty Dir. Andrew Reddie argues that the report represents a significant shift away from the post-World War II model established by Vannevar Bush, under which the federal government funded basic research at universities through agencies such as the National Science Foundation (NSF), National Institutes of Health (NIH), and DARPA. Instead, the administration appears to favor a more venture capital-like (“VCification”) approach that prioritizes commercialization and later-stage technology development over early-stage scientific research.

Transcript

SPEAKER_00 0:06

Welcome back to the Berkeley Risk and Security Labs news podcast, the BRSL Weekly Brief, where we bring you the latest information on current events from our lab experts. I'm the lab's communications manager, Vivian Basou Skinner, and I'm here with BRSL faculty director, Professor Andrew Reddy, to discuss a new report from the Trump administration on federally funded research and what it all means for research labs like ours and specifically AI research across the US. So, first, Andrew, can you tell us a little bit about this new announcement and what kind of it all means?

SPEAKER_01 0:40

Sure. And per great to see you and fun to see you in the summer. Yeah, so this document is potentially quite a quite a big deal in terms of it spelling out what the Trump administration's disposition is towards the research and development landscape and offer something of a change to the way that we think about the ways in which you would traditionally pursue innovation, research, and development for both societal purposes and also military purposes, which is where our labs tends to focus. And so really what this document is doing is carrying on the last two years of the Trump administration's policies towards thinking about research and development, at least in terms of my read, um, as kind of like a vCation of the research and development funding landscape. Um and so, you know, this has strengths and weaknesses that I'm sure we'll we'll kind of get into, um, but is certainly an important document and really spells out where this Office of Science and Technology policy within the White House uh wants to move. And of course, comes following a series of disagreements between the administration and universities that have been all over the news for the last two years as well. Um and really, I think if there's kind of one headline from this report, it's that universities are um, at least by the letter of the report, likely to lose out in terms of the overall um research and development uh pipeline. Um and so to some extent that's also not unexpected given the relatively um new policy of the US government taking positions in the US um in US enterprise. Um so Intel is a good example of that as well. And so there's lots lots to unpick in this document. Um and uh you know it's something that I I care quite a bit about, particularly in the security context, because obviously the way in which a country can leverage its RD and human capital has a lot to do with kind of how it ends up being able to project power on the global stage.

SPEAKER_00 2:51

Yeah. So how does that happen? Um can you talk a little bit about that and how it's already been happening and what you expect now?

SPEAKER_01 2:59

Sure. Well, so I think it just backing up um a long way, actually. Um so following uh the Second World War, um and actually during the uh second world war, um, there was this big question about how um the US could leverage its human capital and resources in order to uh make sure that it was on the forefront of science, engineering, um, and what have you. And of course, to some extent, that was driven by security considerations. Um and so Benevar Bush writes the Endless Frontier uh report that really outlines um the way in which the US has pursued research and development funding over the course of, gosh, what is that, seven, eight decades now? Um and so that's the formation of the National Science Foundation, the National Institutes of Health. Um, of course, on the defense side, you've got the Advanced Research Projects Agency, eventually becomes Defense Advanced Research Project Agency. And really the theory of victory was that you use these government institutions to fund with relatively small dollar amounts, principal investigators or PIs across a wide variety of R1 research universities. And then from those R1 universities and the ideas that members of the faculty are coming up with, you then think about okay, how do we move that from technology readiness level kind of one through three to you know four through six and get ready to kind of prototype and then eventually bring to market and commercialize the innovations that are coming out of the research enterprise? Um, and of course, the national labs play a role in that ecosystem. Um, the military research labs like AFRL, um, Naval Research Laboratory also play a role, and uh as do other what are called federally funded research and development centers or FFRDCs, with apologies. This is an acronym SUP because it is the government, um, but I'll try my best to spell them all out. Um and so um the FFRDCs are also kind of supporting in that commercialization endeavor between very early innovation and development and then two the kind of go-to-market. Um and so that's the way that things really operated up until about two years ago, uh, where the consensus about how to do research and development funding is is was really kind of breaking. Um and that's meant all sorts of downstream consequences for all of the institutions that I just mentioned, as well as ARPA H, ARPA E, that are focused on energy um uh and healthcare respectively. And so um, you know, it's it's it's been a pretty significant shift um in in terms of um what it's meant for those institutions and also um and also universities. Um and you know, I've got thoughts.

SPEAKER_00 5:46

Well, hopefully we'll get to unpack a lot of them here. Um when you're talking about kind of the trajectory for um the partnerships between like research universities like ours and government agencies, uh what are the strengths and weaknesses of venture capital funding? Um you're talking a little bit about it earlier, but how do those partnerships kind of benefit from venture to capital versus um federal funding? And where do you see the future of that balance now?

SPEAKER_01 6:18

Yeah, so so I've written a little bit a little bit about this. Um so um I used an acronym, and I don't think I did spell it out, so I'll still do it now. Technology readiness level, TRL. Um so TRLs go from one through nine. It was uh a scale that was developed by NASA um in, I think it was the 60s. Um and really kind of basically the shorthand is TRL one through three, early, right? These are ideas in people's heads that are relatively untested, and that's where you get very basic experimentation. TRL four through six is in the middle. This is sometimes called the Valley of Death. This is where you might make a prototype, but you certainly don't have scale. And then you've got TRL seven through nine, that's when you're actually able to scale, bring it to market, drive costs down, etc. And all products, whether they are rocket motors or they are new vaccines, run through these technology readiness levels. Um, in in my view, um venture capital tends to focus more on towards the end, right? So it's TRL 6-7 transition, it's not TRL 3-4 transition. Um, and so in a lot of ways, venture capital and other forms of capital really benefit from this consensus that Vannevar Bush kind of put forward about the government's role in driving basic research, right? Basic science, basic engineering, um, and what have you. And so the notion that you could think only in terms of the private market and private uh funding for those lower TRLs is a very untested idea. Um, and indeed, I think one of the concerns that some of us have is that you're basically th you're you're strangling the pipeline that VC ultimately gets to leverage later on. And of course, there's broader conversations about whether universities um and other federally funded research and development centers have been adequately compensated for the basic research that they've undertaken. Uh, I mean, obviously, I mean, we can't do this podcast without talking about the internet, right? With thank you, with thanks to the American taxpayer for this massive engine of economic growth. Um, but there's lots of other examples besides that have come out of any number um of innovations that ultimately the government has paid for. Um and, you know, I there's also something else that's important to kind of unpack from your question as well, right? Like what's the relationship between the various different government agencies and those entities that are receiving basic engineering, basic science, um, basic research dollars? And the answer is there's not really a direct tie. Um there, you know, National Science Foundation, National Institute of Health, um, DARPA, they're often focused on really funding technology and innovation that's so early that it's very unclear what the use case is likely to be. And so it's not kind of being underwritten by, you know, the Air Force or the CDC, right? Um that the relationship isn't that direct. Now, of course, you know, there's various different um entities inside of these institutions that have become wise to research and development funding. Um, you know, here in the Bay Area, we work closely with a defense innovation unit, for example. There's also military services that have their own kind of RD shops. And so it's not as if they're entirely unaware of what's coming down the pipe from you know basic science and engineering. Um, you know, one could look at developments in quantum sensing technologies or various different artificial intelligence applications, and you can see that you know they kind of look where the science and technology is now and then try to extrapolate out, okay, how can it be useful to us? But the relationship's not that direct. Um now, of course, just because you've got you know members of the faculty taking National Science Foundation dollars for basic science work doesn't mean that they're also not receiving you know funding for more applied work from one of those agencies that I just mentioned. Um, and so that's kind of why it's really thorny to kind of unpick all of this. And so the way that I tend to want to organize things is in terms of okay, what's the contribution of these various institutions and actors to technology readiness level research? Um, and then think about the various ways in which the government can be a support or in in this case, potentially a hindrance um to that kind of scientific discovery.

SPEAKER_00 10:37

Yeah, so it sounds like there's kind of probably gonna be a gap in the future between um this very low-level stage of coming up with ideas and seeing if they might work. And we might not even know what that is yet for what do you think, like decades into the future?

SPEAKER_01 10:56

Um gosh, I don't I don't know if our we'll we'll certainly feel it at some point. Um I mean, in a lot of ways, the private VC market is cashing the checks of the taxpayer funded research and development from a decade ago. Um ironically, perhaps, the way that the government funding agencies and that National Science Foundation work is almost identical to the way that you would operate a VC firm. So effectively what you're doing is you're providing relatively small dollars to say 50 independent research units, of which half are going to fail, right? One is gonna succeed spectacularly and become your unicorn, right? So say CRISPR cast nine, right? To you know, shout out Berkeley. Um, and then you're right, a few others are gonna just about break even and become decent, right? And that's effectively what happens inside the VC market, right? So, you know, you might get your one company that ends up becoming alphabet or meta, right? And then a number of others are likely to fail, but the fact that the scale of the success from the unicorn underwrites everything else. Um, and so really it's kind of the same mechanism. It this difference is where the sources of for the funding come from. Um, and really the argument that you know, I I guess Venevar Bush kind of was the one um who kind of pushed it was look, there is a benefit to the country for having taxpayers actually fund some of that basic science and engineering because we're able to get all of the benefits in terms of you know, impact on the labor market, and then obviously, you know, broader um economic externalities by kind of getting growth associated with that research and development that otherwise wouldn't be funded. But in terms of like an overarching timeline, it's very, very difficult to say. Um, it's also worth pointing out that within the Republican rank and file, right, with midterms coming up and um potentially an you know an election cycle that's gonna start here pretty soon after that. Um, it could be that this innovation ecosystem is very likely to shift once again, and then who knows what the impact of having a two to four year gap was, right, for basic science. I will say that you know, one of the there's a couple of headlines that you'll have probably seen around this particular document being released um in places like the New York Times and others, where you've got former policymakers saying, look, this is a document that benefits China, right? Um I'm not sure I've that I would necessarily go that far, but I will definitely say that it makes for you know university professors, offers coming from Europe, from Canada, etc., far more attractive, all else equal. And you actually do see examples of countries around the world organizing their science and investment teams in order to try to go after right American scholars and attract their labs to move elsewhere. Um, and so that's something that you worry about having a long tail consequence, but again, like coming up with what that consequence is in a concrete way, very, very difficult to say. Although I'm sure that you know our colleagues 10, 15, 20 years hence will reflect on this moment and say, well, what was the actually the deleterious consequence?

SPEAKER_00 14:07

Yeah, yeah.

SPEAKER_01 14:08

And you know, maybe it sounds like maybe it'll kind of just the research will shift to other areas or other countries or and yeah, which is which I guess is fine, I suppose, if they're allies and partners, although of course our relationship with allies and partners is a bit constrained at the moment um compared to compared to normal times um in in scare quotes. Um I mean there there are, I mean, there there are to be fair uh to those that kind of point out that this document is really gonna benefit China. I mean, there are American researchers, right, um, who are leaving to China, um, which you know is is definitely interesting given the you know increasingly adversary relationship between the two countries in a security context, which is obviously our bread and butter here. Um so so yeah, you know, we'll we'll kind of see how it plays out.

SPEAKER_00 14:54

So this plan has a kind of intense focus on AI research. Um how do you think that they will decide who gets what funding, and how do you already see this reshaping the future of AI research?

SPEAKER_01 15:07

Yeah, so it really kind of plugs into something that I noted at the outset, which is that this administration's taken a little bit of a different view to actually funding companies that is relatively hands-on for a government, um, or sorry, for a US government. Uh traditionally, we've kind of let the market be the market. Um and so you've already got the US actually investing in companies across the AI stack, right? Or the AI supply chain. Um, it'll be very interesting to see where that um that investment goes. I think from my perspective, you're much more likely to see some of that investment going towards kind of the left-hand side of the supply chain. So that's things like semiconductor manufacturing, uh, potentially an additional um set of support for entities like TSMC that are building their factories out of Arizona, um, around uh around uh Phoenix. Um, and you'll probably see like support for that ecosystem. You also may have to have some of that money channeled towards energy. Um there are all sorts of accounts about how energy hungry um a lot of um the AI technologies have ended up being, with obviously consequences for um for consumers of electricity. And so that's another place where you might see some of these um AI AI dollars. And then of course, data center build-out um is another piece of it as well. So making sure that enough compute is there for um US companies, and that really does actually have a geopolitical competition component associated with it. So as we sit here in what is this, late July of 2026, China's tokens for AI applications are four times cheaper than those coming out of the US, such that companies even here in Silicon Valley are using the open source Chinese models, particularly as they get closer and closer to being as good as the latest offerings from anthropic, open AI, you know, etc. Um, so long as they're not getting a poor a performance um drawdown, the fact that they're four times cheaper, right, means that they're much more attractive. Um and you know, I've sat in multiple meetings where people have feared that that's you know an outcome that is, you know, something that we should worry about in the future, but no, it's here right now. Um and of course we see it in other contexts too, where Chinese models are used around the world to underpin all sorts of different applications. Um I think those are the places where you're gonna be the most likely to see some of that investment. Uh perhaps much less likely to see uh the US government like taking a position in Anthropic or in OpenAI or what have you, um far more likely to be a partner in terms of procurement, etc. Um, but you know, TBD. I think I think really what the US government needs to do in the artificial intelligence space is you know kind of what it's done in other kind of utilities-oriented spaces, which that you know, these are the preconditions for growth moving forward, right? So in the same way that you know the US government has a role in making sure that the energy grid is appropriate for all the things we want to do with it to run our market on top of, that's gonna be what needs to happen in the AI space too. Um of course, the degree to which that either partners with private capital, substitutes for private capital, um, you know, we'll see. Um I think there is some fear in any market that the government gets involved in that it kind of manufactures an artificiality to the price, um, drawing prices down for those that potentially stand to benefit the most. Um you can be sure that this particular line of um funding is not going to deal with things like labor displacement uh from AI technology, right? It's much more focused on kind of creating the preconditions for able to leverage these tools moving forward.

SPEAKER_00 18:53

Yeah. That was actually gonna be my next question is kind of what do you see being left out by this funding and how will that shape the future too?

SPEAKER_01 19:00

Yeah, I mean, I think the you know, there it's it won't be used for retraining, right? Which is something that we talk about a lot in the context of um, you know, the the coming AI revolution. Um it won't be used for um dealing with the displacement of labor capital. Um so you won't it won't be used for universal basic income, for example. Um and like I said, it won't be very it's it's unlikely in my view that they'll actually um be funneling money into the frontier model companies. Uh to be blunt, the frontier model companies don't need it. Um they've got plenty of private VC dollars. And also from their perspective, private VC dollars come with far fewer uh strictures on it than government money does. Um now it could be the case that some amount of these dollars could end up in kind of the defense technology ecosystem, um, very unlikely for it to be the defense innovation unit, but perhaps something like RE, where there's an emphasis on thinking about how startup innovations end up trickling into government services and or military agencies, etc. And so you could potentially see some of that um because some of those companies d are occasionally underserved by uh private investors, uh, particularly because for private investors, government applications don't scale nearly as well as consumer-facing products. Um, actually, one of the places where some of this money may end up going that's worth worth tracking is more on like the robotic side. Um so obviously the way that we think about AI in our heads sitting here in the summer of 2026 is you know, chat GPT windows and and and LLMs. Uh but of course there's lots of um there's a lot of emphasis amongst investors in the current moment, really thinking hard about what some of those advances might mean for physical applications of AI. Um and obviously, you know, here here in the Bay and increasingly around the country, Waymo is a good example on the driving use case. Um, but you can you can imagine any number of others, right? Tesla's famous for trying to give us humanoid robots. And so, you know, I think there might be some support there on the public side for that.

SPEAKER_00 21:15

Interesting. Well, it sounds like there's a lot of directions this can go, and I'm sure we'll follow up as things happen more in the space that we're that's relevant to the lab.

SPEAKER_01 21:25

Yeah, you bet.

SPEAKER_00 21:25

Um is there anything else that you wanted to mention?

SPEAKER_01 21:28

Uh only only that uh only a shameless plug. Um, I mean, much of what I've said today was kind of reflected in a piece that we wrote last summer. Uh so myself and a colleague from the Haas Business School, John Metzler, um, called the Double Power Law. And really, that's where we were kind of pointing out that the way that the government has traditionally treated low TRL work is very much akin to how VCs have treated later TRL work. Um, and of course, you know, talking a little bit about the pros and cons of of those models. Um and so yeah, if you if you want to read more, check out War on the Rocks, Double Power Law.