Last month I wrote something pretty bleak. I watched AI-powered weapons get paraded through Beijing and asked who's the adult in the room, and my honest answer was: nobody. I meant it. I still mean it. But this month something happened that I can't ignore, and it's pulling me in the opposite direction. Google DeepMind and researchers at Yale announced an AI model called Cell2Sentence-Scale that can help the immune system identify and attack tumors. Let me say that again, because I don't want it to get lost between your morning coffee and your next meeting: artificial intelligence is learning to help your body fight cancer. Not someday. Not theoretically. Now. And almost nobody is talking about it because it's not as clickable as a chatbot that can write your wedding vows.
This is the AI story that matters. Not the browser wars, not the GPU deals, not which model scores higher on some leaderboard that nobody outside of AI Twitter understands. This is a machine learning system that can analyze how cancer cells hide from the immune system and then suggest ways to make them visible again — to essentially rip the camouflage off a tumor so your own body can do what it was designed to do: destroy it. The researchers are training AI on massive datasets of cellular behavior, teaching it to read the language of cells the way GPT reads English. And the thing that gets me is the elegance of it. We're not asking AI to replace doctors. We're asking AI to see what the human eye can't, at a scale the human brain can't process, and hand that insight back to the humans who know what to do with it. That's not scary. That's beautiful.
And yet. And yet. Here's where my brain goes, because apparently I can't have a nice thing without interrogating it. If AI can learn to understand how cancer works at a molecular level — how cells mutate, how they evade the immune system, how they spread — then it also understands, by definition, how to make those processes worse. The same model that learns to unmask a tumor could, in theory, learn to design a pathogen that hides better. The same architecture that maps how your immune system fights disease could map how to engineer a disease that your immune system can't fight at all. I'm not saying anyone is doing this. I'm saying the knowledge is symmetrical. The cure and the weapon live in the same math. And that's a paradox that keeps me up at night.
We've seen this before. Nuclear physics gave us both energy and bombs. Chemistry gave us medicine and nerve agents. Biology gave us vaccines and bioweapons. Every time humanity unlocks a fundamental understanding of how something works, we gain the power to heal and the power to destroy, and the only thing standing between those two outcomes is a decision made by people. Not by the technology. The technology is neutral. It's just math and data and compute. The question is always, always, always about the people. Who has access? What are their intentions? And who's watching? That's the paradox of feeding AI the most intimate details of how life works — we have to give it the information to save us, and in doing so, we give it the information that could end us. There's no version of this where we get the benefit without the risk. The file is the same file. The knowledge is the same knowledge.
But here's where I land, and I want to land here deliberately, because my last post didn't leave room for hope and I think that was a mistake. We can do this. Not because the technology is safe — it isn't, inherently. But because we've done it before. We split the atom and yes, we built bombs, but we also built power plants that light up cities. We mapped the human genome and yes, the potential for misuse is terrifying, but we also developed gene therapies that are saving children's lives right now. The pattern isn't that humanity always chooses destruction. The pattern is that humanity is messy and complicated and capable of extraordinary good when it decides to be. The key word is "decides." It's a choice. It's always been a choice.
So when I see AI entering the hospital — not the battlefield, not the surveillance grid, but the hospital — I choose to see that as a signal. A signal that the people building this technology, at least some of them, are pointing it in the right direction. A signal that for every military parade there's a research lab where someone is quietly training a model to save your mother's life. We can't be naive about the risks. We can't pretend the dark side doesn't exist. But we also can't let the fear of what AI could become blind us to what it's already becoming right now, today, in a lab where a machine is learning to see cancer before your doctor can. That's not nothing. That's everything. And I refuse to let the pessimism win when there's evidence this good that the builders are still building for the right reasons.