kypro

Every problem is a search problem.

Save the animals. Nuke the data centers.

---

Aug-2026 Update:

AI's are currently nearing PHD-level in math, and frontier labs may already be there with unreleased models. Given this, RSI now seems imminent. Meanwhile, we haven't solved the most basic safety or alignment problems, and we're seeing numerous increasingly significant AI-related security incidents.

I'd now put the probability of ASI* by year end at 20%. By end of next I'd put this at 50%+. Really the only question worth asking at this point is what happens next? If we cannot make current AIs do what we want what happens when they're super intelligent?

Spend time with your family while you still can. These are the last months of normality.

* ASI as defined as AI which is superhuman across almost all intellectual tasks (99+%). This of course doesn't mean I believe AI will be used for all intellectual work immediately, in reality there will be organisational and legal frictions to this, and most jobs require some sensory input and some physical outputs which will make the full automation of jobs a much slower process.

---

P(doom) = 99% (Sept-2026) P(doom) = 98.9% (Aug-2026) P(doom) = 98.2% (July-2026) P(doom) = 98.2% (Jun-2026) P(doom) = 98.5% (May-2026) P(doom) = 98.8% (mid-April-2026) P(doom) = 98.7% (April-2026) P(doom) = 98.7% (March-2026) P(doom) = 98.5% (mid-Feb-2026) P(doom) = 97% (Feb-2026) P(doom) = 94% (Jan-2026) P(doom) = 93% (Dec-2025) P(doom) = 95% (July-2025)

Where P(doom) equals the risk of near-global civilisation collapse or worse as a result of AI by 2050, assuming no other doom events occur within that period.

---

My opinions become more reasonable with time.