Brief notice.

I wrote this almost in one sitting, so there are going to be grammar mistakes, bad wording, and stuff like that. I just wanted to get this posted; I’ll work on editing this post later.

So, this is based off of all the stuff I’ve heard from the AIP’s FYI science funding briefs. Before anything else, I would first like to give a shout-out to all of the staff writers for FYP—they all have, more than likely, had to sift through mountains of horseshit to actually deliver information to us. In particular, Claire Zhang and Lindsay McKenzie wrote a huge portion of the FYP articles I’m citing here. I can’t thank them enough for the tireless, demoralizing work that they have put into making sure that physicists—and hopefully scientists at large—can keep track of the Trump administration’s science “policy.”

This entire post will be focusing primarily on physics, because that’s what I know about. Physics, thankfully for me, has not been hit quite as hard as other research fields, such as biology and climate science. So anybody in, say, oncology will have a much bleaker outlook on the Trump administration’s science priorities than me—one of my friends, who works in ecology, had their entire research project canned. That said, physics is still not looking great; the field, and the institutions representing it, are not looking great.

For instance, the Trump administration has been attempting to give NASA a “national security” mission (which, seems partially a way to get rid of unions at NASA); the current nominee for NASA director, Jared Isaacman (a billionaire), is mainly planning on just privatizing NASA’s research and making NASA, somehow, generate a profit. Essentially, Jared Isaacman wants to make NASA get everything from contractors rather than doing anything in house—this is counterproductive for research (section 2.4.4).¹

The same goes for the CHIPS act, which current secretary Howard Lutnick (a billionaire) has seen fit to gut, saying that it will be taking a “venture-capital approach” for. Instead of going for a research center such as IMEC—one of the premier semiconductor research laboratories in the world, with 75% of its funding coming from private industry, after being funded solely by the Belgian government in 1986—like in the original plan, the government now wants to make a profit off of the CHIPS R&D budget, and also to take a cut from patents filed by universities from federal grants. The R&D centers initially planned by the CHIPS act are dead. The CHIPS and Science acts were already a mess of compromises, sacrificing on numerous things like union labor—which would have been critical for maintaining a healthy semiconductor workforce considering the amount of horrible chemicals (i.e. hydrofluoric acid) used in semiconductor foundries—it did at least have a purpose, and it had some ambition behind it.

Really, the Trump administration seems to be completely ambivalent towards microelectronics, going so far as to take it off the list of military R&D priorities. The new military priorities are seemingly written by tech industry weenies, rather than career military weenies, judging by the use of tech-buzzwords, like “scaled”—this makes sense considering the fact that the military commissioned tech CEOs to be officers, specifically from Palantir (shocking), OpenAI, and Facebook. The Biden-era approach, while no less horrifying in intent, at least had a series of research areas which followed logically from one-another, meanwhile the Trump administration is focusing on six general areas—AI, biomanufacturing (to produce critical minerals at scale?), logistics, quantum information/computer networks, directed energy (lasers), and hypersonics (the DoW used way more vacuous buzzwords than I did)—that are all disconnected from each other. It’s worth noting that DoW secretary for research and engineering, Emil Michael, is not an engineer or scientist—or anyone with any real experience managing R&D—but a businessman who ran Uber and some dumb startup called Klout that no one has ever heard of; he wants to push basic research off the table for the military, instead giving it to universities and DARPA—of course, short-termism is another factor hurting research productivitiy, so that means that critical research for building new bombs and satellites to watch people jack off will be reduced, while cool new toys are being built.

While the idea of reducing military R&D productivity delights me to no end, it is clear to me that the Trump administration is not prioritizing, say, high-risk/high-reward research that would take a while, but whatever just sounds cool. See, the vast majority of things that the US government did R&D on are, to put it bluntly, boring. Science is boring. Most of what we do is not going to be exciting to the average person, because those things are either simply not feasible or because they are overrepresented in the media. Take semiconductors for example: Most people would find research on semiconductors very boring, in particular things like transition metal dichalcogenides or vacancy defects, and yet these are extremely active areas of research in physics; meanwhile engineers are working on ways to make more efficient devices, and chemists/materials scientists are working on finding new types of semiconductors. You can’t just focus on a single part—say, making qubits from silicon vacancy centers in silicon carbide—because they all work together.²

Really, almost none of the Trump R&D executives have any kind of know-how when it comes to the long, complex process of science and engineering. This probably happened during the Bush years, too, but I’m not sure if it was ever on such a scale. Judging by the fact that the DoE’s under secretary of science, Dario Gill, was the director of IBM research. Gil’s whole schtick is that he loves quantum computing and AI, as evidenced by some lecture he gave at his alma mater—he’s not a scientist, but a career businessman. To quote Gil Lindsay McKenzie’s article on his senate hearing and the profile on him from Stevens:

It is time to build a quantum-centric supercomputer. It is time to elevate the capability of AI to advance scientific discovery and do problem-solving in ways we couldn’t do. It is time now in fusion to shave off decades of development time cycles…. If we do not succeed on those… China [will] get ahead of us, [and] the consequences will be terrible. It is time to mobilize the nation’s best teams to achieve the final frontier of building an error-corrected quantum supercomputer before the end of this decade.

Gil’s vocabulary is approximately 80% buzzwords by volume—and what’s worse is that he’s the best one. Gil might be a businessman, but he at least seems to care about science—in particular, he seems to be very interested in reforming American STEM education which, if done with care,³ is laudable. Compare this to literally any of Trump’s other guys, say Emil Michael, and it’s night and day—Michael is a tech-industry dumbass who failed upwards, much like most of the other R&D executives. In his op-ed on the “Genesis Mission,” which I’ll talk about in a second, he said that “researchers, institutions,… and funding agencies [must use] open-source models, standardized tools, and ready-to-use data” for AI to be useful for science; compare this to what Emil Michael was talking about, and you can see there is a clear-cut difference. The fact that Dario Gil is the best of all the R&D executives is unacceptable.

The Trump administration has two major R&D projects to work on, a sequel to SDI and the other a directionless AI program.

The Golden Dome is originally from an executive order, wherein Trump is basically just calling for boost-phase missile defense systems, most notably orbital interceptors along with sensing systems. While not the same as the SDI, which called for orbital directed energy weapons, instead calls for kinetic kill vehicles with “non-kinetic augments,” similar to the Bush-era missile defense initiative. As such, the recommendations from the APS’s 2004 boost-phase study group report are still relevant for today, and are, in fact, almost more relevant since hypersonic missiles are on the table. Essentially, it was almost impossible to make a missile defense system then, and it’s somewhere on the order of 99.9999% probable that it’s impossible now, too. The actual science or engineering required to build a massive missile defense system are, of course, not relevant to the dumbasses in power, who seem to believe that American ingenuity—and probably AI—will be able to overcome any challenge.

As an ansatz, let’s say that the probability of hitting a missile is at least proportional to the area that an interceptor will be hitting from ($A$), and that the interceptor has some cross-sectional area that arises from errors in missile tracking ($\sigma$). Barring all other constraints, this probability will then be $$P \propto \frac{\sigma}{A} \,.$$ Hence, somewhat obviously, the maximum value of $\sigma$ will be $$\sigma \leq A \, P,.$$ For a high hit probability, we can just say that $P = 1$. The US long-range hypersonic weapons have a radius of somewhere on the order of $0.44 \ \text{meters}$, and hence $$\sigma \leq \pi \, R^2 \implies \sigma \leq 0.608 \ \text{m}^2 \,.$$ This estimate is only for intercepting American weapons—hypersonics from other countries would have different sizes, and therefore different cross sections. That said, this really basic analysis doesn’t factor in countermeasures, or countries just, y’know, launching dummy missiles to distract from the super expensive satellites.

Oh yeah, also, that thing about the Golden Dome mostly eschewing lasers? I lied—somewhat. The Trump administration has not released any details, other than saying that other countries have big, scary missiles, and that the Golden Dome should have boost-phase interceptors. You would think that around forty years of criticism, including three technical reports⁴ would in some capacity deter the government from trying it again—you’d think that they’d just give up the ghost—but no, they’re still trying for it. It’ll work this time, for real!

The other major R&D project is the “Genesis Mission,” which seeks to double America’s scientific productivity through AI, and it’ll be accomplished by building a big, federal AI platform. How will scientific productivity and novelty be measured, what problems this supposed platform will be made to tackle, and how this will be good in any way are left to the indeterminate future. The DoE is now tasked with developing a plan for this big, beautiful system, and we’re supposed to have an answer in 120 days or so, just like how the Golden Dome plan was supposed to be ready in three months. To support the DoE’s AI mission, a whopping nine new supercomputers, with several being tailor-made for—what else?—AI.

The Genesis Mission is a really strange bit of legislation to me, mostly because, unlike prior major R&D pushes, it doesn’t really have anything specific. Other major projects—such as mRNA vaccine development, the Manhattan Project, and the Apollo Program—all had a specific thing to be built: a rocket to send an American to the moon, a lifesaving vaccine that will later be vilified for no good reason, and weapons which have left literal and figurative stains upon the earth. With the Genesis Mission, it just sounds like something that some guy thought of while he was bored—there’s a really lofty goal there, but there’s no real good way to show progress on it.⁵

This entire administration’s approach to R&D is almost exactly that of a middle schooler—get rid of the yucky and/or boring stuff (like the shrimp on a treadmill, or anything to do with trans people, or semiconductor research)—and replace it with Cool Toys. That’s what the administration wants: Cool Toys. Perhaps it is quite fitting that so many of the R&D executives are from the tech industry, because most of them are only used to building useless shit under incredibly short time frames. What we are seeing is not merely the ambitions of a deeply stupid president, but the result of Silicon Valley’s combined isolation—both geographically and intellectually—and undue influence over the average American. The software industry has created a generation of businessmen who think that every single problem is as simple as a few lines of code.

I wrote about the Silicon Valley approach to R&D almost exactly two years ago—I’m horrified to see that the problem has only gotten worse. I can only hope that four years of federally funded mismanagement, empty promises, and unvaccinated children will see to some sort of corrective mechanism. On a lighter note, it seems that other nations are noticing America’s year of humiliation in R&D, and are seeking to get American researchers—Europe and China specifically seem to be keenly interested in attracting American talent, both into industry, government, and academia. Chances are, I’ll be looking to do graduate school in Europe, since I’ll at least be able to rely on funding there.