🎧 LISTEN TO THIS ARTICLE ON THE SUBSTACK APP
Artificial intelligence is being tasked with doing just about everything you can think of: writing computer code, producing advertising campaigns, generating movies, analyzing financial markets, designing weapons, operating factories, and reorganizing the modern workplace. Billions of dollars are pouring into a technological revolution that promises to alter virtually every aspect of human existence.
Yet an obvious question deserves considerably more attention — and far less polite delay: When will the enormous computing power being assembled by the world’s leading AI companies be systematically directed toward curing diseases that medicine has spent decades failing to defeat?
Few diseases make the question more urgent than Amyotrophic Lateral Sclerosis.
ALS remains one of medicine’s most unforgiving diagnoses. Scientists understand considerably more about the disease than they did a generation ago. Researchers have identified genetic mutations, abnormal proteins, inflammatory processes, cellular dysfunction, and numerous biological pathways associated with motor-neuron degeneration. Several treatments can even modestly alter aspects of the disease for some patients. Experimental approaches involving gene therapy, antisense technology, stem cells, peptides, neuroregeneration, immune modulation, and other mechanisms continue to be investigated.
But the fundamental reality hasn’t changed enough — and the pace of change is still being dictated by institutions that don’t live on the same clock as the patients.
A person diagnosed with ALS has been entered into a race against progressive neurological destruction. Meanwhile, researchers attempt to understand a disease whose biological complexity has repeatedly defeated conventional drug development. That isn’t a mystery that requires another decade of committee meetings. It’s a problem AI should already be attacking.
The traditional medical-research model depends heavily upon human beings developing hypotheses, designing experiments, analyzing results, publishing findings, and then beginning another cycle. It has produced extraordinary achievements, but it’s inherently constrained by the amount of information any one researcher or research team can absorb.
Modern biomedical science has generated something entirely different: an ocean of information.
Genomic databases, protein structures, clinical records, pathology reports, imaging studies, laboratory experiments, clinical trials, failed drug programs, molecular databases, published papers, and decades of accumulated research now contain relationships that no single scientist could possibly understand all at once.
Artificial intelligence, at least potentially, can.
Imagine an AI system built specifically around ALS research and given access to essentially the entire body of legitimate scientific knowledge about the disease. Instead of asking researchers to read several hundred studies, the system could analyze hundreds of thousands of papers, datasets, molecular interactions, clinical outcomes, genetic variations, and failed experiments.
More importantly, it could search for relationships that human beings have missed.
Perhaps a drug abandoned for one neurological disorder affects a pathway relevant to ALS. Perhaps patients with unusually slow progression share biological characteristics buried across separate datasets. Perhaps combinations of existing compounds produce effects that individual drugs don’t. Perhaps apparently unrelated research involving mitochondrial function, protein aggregation, neuroinflammation, axonal repair, or cellular metabolism contains pieces of the same larger puzzle.
AI doesn’t guarantee that such connections exist. But searching for them is exactly what machines are becoming exceptionally good at doing. Waiting for someone else to organize that search isn’t a strategy. It is foot-dragging dressed up as institutional prudence.
The obstacle is increasingly less about computational capability and more about priorities, organization, data access, and incentives.
America’s largest technology companies are spending extraordinary sums to build data centers and develop increasingly powerful models. Pharmaceutical companies possess enormous proprietary libraries of compounds, trial results, and biological information. Universities and government agencies possess additional mountains of research.
Unfortunately, much of this knowledge sits in separate institutional silos—and those silos are being treated as if they were inevitable, rather than a choice.
There is another possibility that belongs in the same conversation. A business that earns recurring revenue from managing chronic and terminal illness doesn’t automatically share the same urgency as a patient who won’t live long enough to become a long-term customer. Treatments that must be taken indefinitely can be more profitable than a one-time cure. Failed programs can be locked away. Promising combinations can stay untested if they threaten an existing franchise. None of that requires a cartoon villain. It only requires ordinary incentives. If AI can rapidly surface discarded compounds, buried trial data, and unexpected combinations, some firms may not be eager to open the vault. Sick people, after all, are still a market.
That is where leadership is needed, and where it has been conspicuously absent.
The federal government doesn’t need to nationalize medical research or dictate scientific conclusions. It could, instead, remove barriers to cooperation, establish carefully governed biomedical data standards, require or strongly incentivize companies to make failed clinical-trial information available for computational analysis, expand privacy-protected research datasets, and create real consequences for hoarding data that patients paid for with their bodies and their time. It could do those things right now. It has not.
Private philanthropy could play an equally important role in pushing this work forward.
Imagine a billion-dollar competition devoted to developing an AI system capable of identifying genuinely promising therapeutic targets for ALS. Imagine similar programs for pancreatic cancer, glioblastoma, Huntington’s disease, Alzheimer’s disease, and other devastating illnesses. The objective wouldn’t be to produce another chatbot capable of talking about medicine. It would be to create specialized scientific systems designed to generate hypotheses that laboratories could test—including hypotheses that existing commercial pipelines have little reason to pursue.
There must, of course, remain a human firewall between computational prediction and patient treatment. AI can generate a vast number of plausible hypotheses, including incorrect ones. Laboratory validation, animal studies where appropriate, carefully designed human trials, independent replication, and rigorous clinical judgment all remain indispensable.
But caution shouldn’t become an excuse for institutional inertia. Too often, it already has.
Terminally ill patients experience time differently from bureaucracies and quarterly earnings calls. A regulatory discussion lasting three years may seem reasonable from an institutional perspective. For someone with an aggressively progressive neurological disease, three years can represent an eternity they simply don’t have; time they cannot afford. That isn’t a rhetorical flourish. It’s the difference between a life and a file.
That reality should create urgency without abandoning scientific standards. What it shouldn’t create is another round of polite statements about “the promise of AI” while people lose the ability to walk, speak, swallow, and breathe.
The great promise of artificial intelligence isn’t that machines will replace physicians or scientists. It’s that machines may allow them to investigate biological complexity at a scale previously impossible.
We are approaching an extraordinary historical choice. Humanity can devote unprecedented computing resources primarily toward making advertisements more efficient, entertainment more personalized, financial trading faster, and office work cheaper. Or some meaningful portion of that technological power can be deliberately aimed at humanity’s oldest enemy: disease — even when a cure is less convenient than a subscription.
ALS would be an incredibly appropriate place to begin.
There are thousands of families for whom artificial intelligence isn’t primarily an interesting technological development. They are watching a clock, and when that clock strikes the hour, a loved one is lost forever.
These families deserve to know when Silicon Valley, the pharmaceutical industry, universities, medical institutions, philanthropists, and government research agencies intend to treat curing terminal disease with the same urgency now being devoted to building the next generation of artificial intelligence profit—instead of treating the delay as someone else’s problem or someone else’s profit center.
The machines are becoming powerful enough to ask questions that science could never realistically ask before.
The question now is whether we have the will to point them at the problems that matter most, or whether we will keep finding reasons to wait while the people who need answers run out of time.
FOLLOW US ON X
FOLLOW US ON FACEBOOK
A Note About Subscribing:
At Underground USA, we pride ourselves on being fiercely independent—we only answer to our readers. This independence allows us to tackle tough questions, dive deep into important topics that often get overlooked, and share essays and podcasts, like this one, without asking for anyone’s approval.
Free subscriptions help us connect with a wider audience. They are the vehicle for the wide dissemination of the message. Paid subscriptions are what keep us going. They fund our research, document requests, interviews, fact-checking, and give us the time to dig deeper than just the headlines. Most importantly, they support something that’s getting harder to find: independent analysis that’s not swayed by advertisers, corporate sponsors, or political agendas.
