Bounding AGI timelines: Josh Purtell on Carlsmith-style Fermi estimates
JoshPurtell · x · 2026-10-04
In a discussion with AI safety community figures including Yudkowsky and Rob Bensinger, Josh Purtell argues two lines of reasoning are being conflated:
- One asks how much compute it takes to do math/write code — the Carlsmith Fermi-estimate line mainly tries to find bounds for this
- The other uses what the brain is actually capable of to bound timelines for money-making applied AI outcomes
He finds the Carlsmith et al. approach fruitful and suggestive despite not having read every word.
More from AGI Musings
- LLMs aren't conscious yet but artificial consciousness is buildable, argues anti-pain-vector stance — basedjensen · 2026-10-04
- Ex-Goldman exec Raoul Pal: agentic AI will eat traditional software and SaaS — rohanpaul_ai · 2026-10-04
- Heidel's 'Claude is a sentient god' quip ignites fight over the industry's consciousness assumption — RileyRalmuto · 2026-10-04
- Why cybersec folks bristle at 'AI hacking' probes: turf, expertise and being sidelined — austinc3301 · 2026-10-04
- Four roles for humans in an AI world: from human premium to augmented craft — msharmas · 2026-10-04
- Datacenter buildout would have been less aggressive in a world where everyone gets AGI at once — willcb · 2026-10-04