Bostrom's 1998 paper predicted superhuman AI within the first third of this century
maksym_andr · x · 2026-10-06
maksymandr revisited Nick Bostrom's 1998 paper, noting some people have been "consistently correct" in their predictions: it argued "we will have superhuman artificial intelligence within the first third of the next century."
Follow-up tweets quoted two more key arguments from the paper:
- Recursive self-improvement: once AI reaches human level, a positive feedback loop kicks in—AIs help build better AIs, accelerating development.
- Why slowing the frontier for safety is so hard: even without assurance AIs are safe, development would likely continue anyway—either people don't see gradual displacement of biological humans as bad, or short-term profit, curiosity, and ideology make a collective ban unachievable.
The author also notes Bostrom was completely wrong about backpropagation (he saw its scaling cost and supervised nature as fatal limits)—but maybe technical misjudgments don't matter much in the grand scheme.
More from AGI Musings
- Ex-big-tech AI engineer in China freelances by day, delivers food by night as AI crushes rates — Hesamation · 2026-10-06
- The Human-Machine Friendship Society: free CC-licensed posters for AI optimists — GWBrooks · 2026-10-06
- OpenAI trains agents to swarm: multi-agent coordination could be the next scaling law — binarybits · 2026-10-06
- 90+ Claude Opus 5.5 agents find two room-temperature magnetic semiconductor candidates in 3 days — Dr_Singularity · 2026-10-06
- Paul Graham: Even the Smartest People Must Learn Taste in Problems From Experience — ycombinator · 2026-10-06
- iamtrask: AI consciousness and AI copyright are the same debate — iamtrask · 2026-10-06