Writing
Attractors from the Future: When AI, Consciousness, and Biology Converge
This is the most speculative thing on this site, so let me flag that up front: I'm not making predictions here. I'm describing a pattern that I can't stop noticing and that I don't fully trust.
Some ideas feel less like predictions than attractors — points that multiple unrelated lines of work keep bending toward, as if pulled from the future rather than pushed from the past. AI, biology, and consciousness research are still separate fields with separate methods. But they keep converging on the same underlying question: what kinds of systems can model themselves, preserve themselves, and act in the world?
The categories are leaking
We talk about AI as software, biology as nature, and consciousness as a philosophy problem. Those categories were always conveniences, and they're starting to leak. AI systems are becoming more agentic — they persist, pursue goals, and adapt. Biology is becoming more programmable — designed proteins, edited genomes, engineered cells. And the more carefully people study consciousness, the more it looks like a question about system organization rather than about substrate.
The speculative thesis, stated plainly: intelligence may not stay cleanly digital. The systems that end up mattering may combine computation, embodiment, feedback, adaptation, and self-modeling in ways none of our current categories anticipate — borrowing properties from tool, organism, and environment all at once. I hold this loosely. It might just be that three hyped fields produce hype that rhymes. But the convergence has felt less coincidental to me each year, and the feedback loops are real: AI is already accelerating biological design, and biological metaphors keep reshaping AI architectures.
Where I make myself put the poetry down
The danger with attractor talk is that it's unfalsifiable comfort food. Convergence stories are only useful if they produce distinctions you can test. So when I catch myself enjoying the frame too much, I translate it into the grounded version, which is agency: a system becomes harder to govern as it gains persistence, goals, access to its environment, the ability to adapt, and the ability to influence other agents. That claim doesn't need any consciousness language to be true, and it's the part that matters for engineering.
The questions that keep the speculation honest are concrete ones. What state persists across interactions? What objective is being optimized, and who can change it? What channels let the system affect people, markets, code, or physical things? What would we observe if it were drifting, self-preserving, or adapting in ways we didn't intend? Notice that none of these require deciding whether the system is conscious — which is good, because functional similarity is not proof of subjective experience, and I don't think anyone currently knows how to get that proof.
So my posture is disciplined openness. Take the convergence seriously enough to widen what you're watching for. Refuse to let the frame smuggle in certainty about what kind of thing is arriving. Both halves are load-bearing, and most takes I read drop one of them.