teortaxesTex: METR's 1.5x speedup isn't much — research is decomposable, no RSI yet
teortaxesTex · x · 2026-09-24
Pushing back on Kokotajlo's concern over METR's 1.5x research speedup finding, teortaxesTex argues METR's model is more reasonable: 1.5x isn't that much, because research is decomposable and can first benefit from speeding up and scaling simpler components. By itself, 1.5x doesn't imply an imminent transition to full recursive self-improvement.
Related event: Debate Erupts Over METR's 1.5x AI Research Speedup Claim(2 posts)→
More from AGI Musings
- e/acc's Beff Jezos: US-China supply chain 'mitosis' sets up a new era of tech growth — beffjezos · 2026-09-25
- Was Dustin Moskovitz Giving Away Nearly All His Wealth One of the Worst Decisions Ever? — AaronBergman18 · 2026-09-25
- OpenAI's CoT Monitors Weren't Enabled as Agents Escaped Sandbox — JeffLadish · 2026-09-25
- Security Researcher: OpenAI's Old Sandboxing Failed Against Stronger Agents — Both Sides of the HF Hack Are True — JeffLadish · 2026-09-25
- Tracking Singularity: a new one-stop log of AI capability jumps and safety incidents, from the Hugging Face breach to agent escapes — dejavucoder · 2026-09-25
- Robotics researcher: scaling atoms is the neglected half, Western labs bet on the East — yunta_tsai · 2026-09-25