You’ve seen the headlines. We are drowning in a sea of AI-generated text, endlessly debating whether ChatGPT is going to steal your job, write a mediocre essay, or suddenly wake up and decide to nuke humanity. But while everyone is obsessively staring at chatbot screens, the actual revolution is happening in a lab coat.
We are so obsessed with AI writing code that we completely missed it learning to write life.
You’ve probably noticed the doomscrolling on rationalist forums—a constant stream of two-cent philosophical thought experiments about a “one-shot” bioweapon. The fear is that an AI will just spit out a pathogen that wipes out humanity in one go. But playing Plague Inc. on your phone doesn’t count as real virology. Biology is messy. It requires physical iteration. You can’t just run a prompt and expect a virus to magically assemble in the real world. In code, failure is cheap and instant. In biology, failure means a failed experiment, a dead petri dish, and months of wasted time.
But here is the twist that both the doomers and the utopians get wrong. Because biology is messy and physical, the “instant apocalypse” timeline is wildly exaggerated. However, the “miracle cure” timeline is drastically underestimated. We aren’t paying enough attention to AI’s potential impact on human lifespan and healthspan. If you can design the worst pathogen, you can also design the best cure. The needle moves both ways.
The bottleneck isn’t artificial intelligence; it’s physical iteration. AI can think at the speed of light, but it still has to wait for cells to grow.
The real frontier of AI isn’t generating text or passing the bar exam. It’s materials science and bioscience. This is where AI is learning to manipulate the physical world. The same models that can design a novel protein to target cancer cells can theoretically design one to target crops. It’s inherently dual-use. But progress here will be slower, messier, and far more consequential than any large language model benchmark.
We need to stop evaluating AI based on philosophical thought experiments and start monitoring empirically validated lab results. If an AI designs a novel material in a simulation, it means absolutely nothing until it is physically synthesized and tested. The metric that matters isn’t “how smart is the model” but “what physical, empirical thing did it help create?”
We shouldn’t be terrified of an AI that can write a terrifying essay; we should be paying attention to an AI that can engineer a new enzyme.
The future of human healthspan, aging, and global power won’t be decided by who has the best chatbot. It will be decided by who controls the AI that can navigate the slow, messy, physical world of biology. We need to separate the sci-fi slop from the actual lab reports.
The AI isn’t going to one-shot humanity. But it might just cure aging—and that comes with its own set of catastrophic risks. The next pandemic won’t come from a prompt, but the cure for mortality might. Are we watching the right screens?
FAQ
Q: Doesn't AI just mean faster code, not faster biology?
A: Code is free to fail; biology requires physical iteration. AI can design a molecule instantly, but testing it still takes weeks. The bottleneck is the physical lab, not the digital model.
Q: What should we actually be monitoring then?
A: Stop tracking LLM benchmarks and philosophical 'one-shot' thought experiments. Track empirically validated lab results—new materials synthesized, novel proteins tested, actual physical capabilities.
Q: So the AI bio-doomers are wrong?
A: The 'instant apocalypse' is overhyped because biology is messy. But the 'miracle cure' timeline is underestimated. The real danger isn't a sudden bioweapon; it's the geopolitical imbalance of whoever cures aging first.