I Demand My MAHA Genesis Mission Medbed

But I'll settle for vaccines and a functional health system since medbeds don't exist no matter how much AI you throw at it

My favorite movie is Aliens, partly due to the strong, technically competent female protagonist who likes to wear tank tops, smoke cigarettes, operate complex equipment, protect children and the hot military guy she works with, and quip sarcastically at stupid people. It is also partly because the larger Alien universe contains so much interesting biology and biotech (ask my long-suffering husband, who I’ve subjected to numerous lectures on xenomorph biology).

In particular, I’ve always been enamored with the cryo-sleep technology due to the fact that I’m an insomniac who frequently travels. I have devoted entirely too much thought to how much my health would improve if I could just cryo-sleep the entire flight to my next conference in Asia.

Medbeds aren’t cryo-sleep chambers, but they exist in the same realm of hypothetical future medical technology: a pod that conveniently solves a major biological problem without much explanation about how it does so. For cryo-sleep, it’s keeping people alive but not using resources like food or oxygen over months or years. For medbeds, it’s healing people almost instantly and permanently with a magical scan. We are nowhere near having the technological capability of building devices that can do either of these things. I often wonder if either of these technologies are even possible, since at least I am not innovative enough to imagine a realistic mechanism for developing either one.

I was curious enough to do some research about where US President Donald Trump got the idea when he posted an AI video to Truth Social a couple of months back promising medbeds for all by next year. It turns out that medbeds are a fantasy that sprang from the QAnon movement. Medbeds are what they sound like: high tech devices that look like a futuristic coffin that medically fix any ailment of anyone who climbs into one. It’s not clear how medbed interventions are carried out—presumably AI is involved and they have a bunch of robotics and lasers for surgical procedures. They are also powered by prayer.

Except they’re not here and they don’t exist. From the New York Times, 2024

Medbeds make me very sad, because a lot of sick people in chronic pain are being exploited by terrible people who sell them a promise of an imaginary instant healing machine that doesn’t exist. As the New York Times reported in a 2024 article about the medbed phenomenon, any skepticism is met with repetitive insistence that medbeds exist that includes equal parts pseudoscience and religion. The only reason they are not yet mainstream is that they are being kept a secret by the military at the behest of woke liberal billionaires who are keeping the technology for themselves. For the people who are suffering badly enough to seek knowledge about whether medbeds might help them, resistance to the idea is gradually replaced with hope through relentless medbed disinformation.

No better place for secret alien tech than secret military tunnels

So, in other words, medbeds are possibly alien in origin and can do things that seem completely outside the realm of our current medical technology. A holographic scan is just a projected image, so I don’t know how it could diagnose anything, much less repair an injury. I’m not sure what “personal DNA” is, but I do know that synthesizing DNA does not enable regrowth of missing limbs in minutes. And no matter what the longevity grifters tell you about epigenetic reprogramming, there is no way to induce “reverse aging” and erase the ravages of time. It is not biologically possible to physically rewind yourself to having a 25-year-old body and mind.

Medbeds make no sense whatsoever scientifically. They are imaginary creations used to advance the plot of science fiction stories and movies. So, naturally, this administration has turned them into a lucrative opportunity to profit from pain and suffering. Rather than actually improve access to health care or invest in biomedical research, they just point to the nebulous mythology of the medbed, which is advertised as being just barely out of reach. Trump is going to roll them out any day now.

Medbed foundation models

I call X/Twitter “the hellsite” for a variety of reasons that are obvious to most of its extant users, but one circle of the hellsite that I find particularly annoying is the one where the “AIxBio” cheerleaders dwell. To hear these entrepreneurs talk, their newest model is about to cure everything tomorrow because it met a bunch of artificial benchmarks they created for themselves.

Everything is a revolutionary new advance that is catapulting us forward light-years into a utopian future of perfect, bespoke health, yet most of the time they don’t bother with experimental validation to prove that it works. Don’t get me wrong, people have done some cool things with protein, drug, and antibody design. People are developing some interesting models for forecasting evolution, predicting binding and enzymatic interactions, and exploring different complex biological systems. There are real advances in biology and medicine that will come from AIxBio (and full disclosure, my lab uses machine learning models to predict disease outcome and we are developing an AI model of our own to study host responses to infection).

But in my opinion, at least 95% of AIxBio claims are grossly overhyped fantasies of tech bros who read a book or watched a YouTube lecture about biology one time. Outside the big foundation model developers (OpenAI, Anthropic, Google DeepMind, etc), most of these companies don’t even have a product apart from rinky-dink models that are more focused on meeting benchmarks for model performance with dubious relevance to actual scientific or medical problems.

So, as soon as I saw that AIxBio Twitter was salivating over Trump’s latest Executive Order (EO), Launching the Genesis Mission, I had a bad feeling about it. Then I read the EO and screamed, because this EO is a simultaneous path to a surveillance state and supposedly possible but not remotely realistic medical innovations like—you guessed it—medbeds!

Executives at government tech contractors rejoice! Trump administration just dropped a new angle to cash in on taxpayer-funded AI bullshit by making your “cloud lab” dreams come true

The “Genesis Mission” as laid out in the EO and accompanying Fact Sheet is “a new national effort to use artificial intelligence (AI) to transform how scientific research is conducted and accelerate the speed of scientific discovery.”

That sounds pretty good, right? Let’s do more science, better and faster! Priority areas include biotechnology, critical materials, nuclear fission and fusion energy, space exploration, quantum information science, and semiconductors and microelectronics. Unleash the raw scientific power of American AI! Also, there’s a lot of stuff about robotics being deployed to implement whatever the Genesis Mission platform ends up being, so that sounds cool and futuristic too.

It’s like going to space or building nukes, just with more fraud and graft

But how will the Genesis Mission actually use AI and robots to achieve scientific dominance? Whoever wrote this EO does seem to understand the most fundamental scientific problem with “AI-driven” approaches to accelerating scientific discovery and progress: the need for an adequate training set.

Large language models (LLMs) like ChatGPT, Claude, and Gemini are “generative” because they are trained on vast datasets across billions or trillions of parameters. They can extrapolate new (and not always correct) information based on that, but they can’t accurately generate any new insights about systems they haven’t been trained on. All AI models require training, including others besides LLMs like small language models, machine and deep learning models, diffusion models, generative adversarial models, vision learning models, and masked language models. They all depend entirely on the quality of the training set.

Biology is extremely complex. Despite centuries of biological and biomedical research, there is still a lot we don’t know. In virology, we lack a great deal of knowledge about even well-studied viruses like influenza. Tens of thousands of flu papers have been written and we have made huge leaps in our understanding of how influenza viruses evolve, infect their different hosts, replicate, and cause disease, and yet the virus continues to surprise us. There are huge gaps in the available training data for one of the best-studied viruses out there, and this is just one little virological corner of the vast biological world. Across all fields, there is no training set that contains enough data to cover absolutely everything.

The reality is that to understand most biological problems, you have to collect experimental data. If that data doesn’t exist because nobody has ever done the experiment, you can’t train an LLM or any other model on it. If you then ask the LLM to answer a biological question relevant to a gap in its training set, the answer it gives will be no more accurate than if you took your best guess based on vibes and limited prior knowledge.

This gets to the real purpose of the Genesis Mission: to leverage both computing capacity and proprietary datasets within the Department of Energy’s (DOE) network of national labs for developing expanded training sets. The national labs run by the DOE have been collecting huge amounts of non-public-facing data for years about everything from semiconductors to nuclear testing to the human blood proteome. These data would be exceptionally valuable for model developers.

Lest you think the government itself is getting into the AI business, this platform will be developed by staff at the national labs, many of whom are actually employees of companies that are contracted by the government to operate them, such as Battelle and Lockheed Martin, as well as partners in the private sector. Those partners include foundational model companies like Anthropic, chip makers like Nvidia, and hardware manufacturers like Dell. I strongly suspect this will also involve software companies like Palantir, which will feed these data into its own burgeoning surveillance state models, or Anduril Industries, a weapons manufacturer that builds things like autonomous AI-powered drones and surveillance technology.

While the Genesis Mission platform will surely improve some AIxBio models (depending on their purpose), it’s hard to imagine the current administration using it to accelerate vaccine or drug development, make fundamental discoveries about biology, or make health care more accessible and efficient for Americans. The current administration loves to talk about Gold Standard Science and ending the scourge of chronic disease, but not so much about developing innovations using the scientific merit-based approach that made America the world leader in biomedical and health research. They like to cut grants instead, especially when those grants are for vaccine development. They give grants with no scientific review at all to their cronies. They enforce EOs that shut down essential laboratory research. They are misrepresenting essential research and in some cases outright lying about it. They screw around with regulatory processes to make it more difficult to approve vaccines and drugs. They are making scientific research funding contingent on ideological alignment. They want to use AI to develop alternative treatments, for the purposes of selling said alternatives and getting the government to pay for them. They are more inclined to delete government datasets than to share them.

For these reasons, the Genesis Mission is doomed to fail at its biotechnology mission. The datasets at DOE are huge, but they will still be insufficient to fill knowledge gaps in the training set without investments in experimental biology and clinical trials.

This brings us back to medbeds

It’s true that medbeds aren’t mentioned anywhere in the Genesis Mission EO, but I can see their presence nonetheless. The EO sells a fantasy of all-powerful biomedical AI that will “usher in a new golden age of human flourishing” and can “unlock cures” for terrible diseases like pediatric cancer. As with most Trump scientific EOs, it’s light on details and heavy on grand promises that are unlikely to be kept. It fills in any blanks with grandiose claims about AI dominance, citing only “DOE training sets and high-performance computing resources” and “robotics” as the means to achieve it. The Fact Sheet brags about the Trump administration’s steps toward AI dominance so far, which is primarily issuing other EOs and Action Plans asserting that said dominance has already been achieved.

Of his many crimes against America, Biden’s biggest was his woke AI policies

Now that Trump has freed American AI from the shackles of wokeness and is blathering about the “export of full-stack American AI technologies,” the next logical step is medbed development. The Genesis Mission has all the ingredients: nonspecific promises of spectacular advancement, unlocking the secrets of deadly diseases that everyone hates, assurances of American dominance, lots of techno-babble, infrastructure building, and robots. This is the same sales pitch used by scammers to ensnare medbed enthusiasts already.

There is an entire scam economy built around selling medbed access, including $800 tickets to the secret military tunnels containing the government’s secret medbeds and companies that sell canisters of “life force energy” to fuel medbeds. According to the New York Times, a Florida mattress salesman named John Baxter started a medbed company that actually manufactures them. From the article’s description, it sounds like these are either hyperbaric chambers or vibrating medbeds that work when the subjects undergoing treatment are aligned with the right “frequencies.” This is a long way from AI-powered robotic medbeds, but there are already magic finger medbed spas that exist to take people’s money, create hype, and build demand.

The Genesis Mission sounds good as a national scientific initiative, because AI and its many applications (including biomedical research and health care) is indeed an area where the government should make strategic investments. However, a quick glance at the underlying context, like nearly all MAGA/MAHA scientific initiatives, reveals that it’s built on a foundation of bullshit and exists to advance the ongoing corruption of the US federal government. The Genesis Mission is a stepping stone to medbeds, which is to say it’s a stepping stone to the continued destruction of actual American scientific innovation, theft of taxpayer money by the autocrat-in-chief, and further eroding the democratic norms that truly did make America great.

We will never have medbeds. We will have authoritarians with even more technological power and tools to prevent serious dissent. We will have desperate people, praying for a solution to their ailments and receiving snake oil and broken promises in return. We will have scientific stagnation as our technological infrastructure is appropriated for dictatorial control of the citizenry and military applications and a technofascist surveillance state.