tszzl: 'Sparks' of ability in AI models mean superhuman performance is months-years away
tszzl · x · 2026-11-03
AI researcher tszzl argues that whenever you see 'sparks' of any ability in AI models, you're months to years away from it being superhuman at that task — grade school math today, Navier-Stokes tomorrow. In his view, these results are almost impossible to overhype: 'you are almost always underhyping no matter what.' A representative optimistic take on capability trajectories.
More from AGI Musings
- A model better at editing its own weights is a very tight RSI loop — rickasaurus · 2026-11-03
- After visiting Chinese AI labs: the gap with US frontier labs is widening — pstAsiatech · 2026-11-03
- The Blindsight 'Conspiracy': Are Blindsight Patients Just Pretending? — theshawwn · 2026-11-03
- Paul Graham: the scariest part is calling this a normal year — paulg · 2026-11-03
- Is 'Higher Intelligence' Just What Takes Humans Longer to Learn? — GregCook2011 · 2026-11-03
- kalomaze: In-context learning is an incidental byproduct of next-token prediction, not deliberate design — kalomaze · 2026-11-03