VraserX: AI slowdowns must bind internal labs too, or power concentrates
VraserX · x · 2026-09-14
VraserX argues that a coordinated AI slowdown is only acceptable if the rules apply inside the lab as well. He warns of a scenario where researchers use an internal GPT-7 to accelerate R&D while the public is stuck with GPT-6 forever for safety reasons — a dynamic he calls an enormous concentration of power.
More from AGI Musings
- P(doom) is like cocaine: mocking AI doomerism as a self-reinforcing social bubble — charles_irl · 2026-09-14
- MIT study finds AI explanations make people more confidently wrong, but doctors resist — alex_verem · 2026-09-14
- Joshua Saxe bets an out-of-control AI botnet appears within 18 months — joshua_saxe · 2026-09-14
- Bindu Reddy debunks AI doom arguments: no self-copying, and you can just unplug it — bindureddy · 2026-09-14
- Commenter argues Anthropic's book training garbles texts, 'losing' millions of books — StewartalsopIII · 2026-09-14
- Giulio Prisco on the AI Navier-Stokes solution: science will be a human-AI joint endeavor — Chris_Armstrong · 2026-09-14