Missing continual learning, not motivation, is what keeps models short of AGI, argues teortaxesTex
teortaxesTex · x · 2026-09-29
In a discussion with @zetalyrae, teortaxesTex argues the lack of AGI-ness stems from missing continual learning, echoing Liang Wenfeng's view. Agents improve dramatically in-context; if those gains persisted, lack of training and long-horizon competence would be largely ameliorated. He dismisses motivation-based concerns as budget-ignoring paranoia, noting systems like Astra could spawn subagents to become superhuman in data-deficient domains.
Related event: Debate: Missing Continual Learning Is Key Barrier to AGI(3 posts)→
More from AGI Musings
- Do we need better theories of consciousness? — burny_tech · 2026-09-29
- After 75 interviews in 4 countries, one year proved 'AI can't code' wrong — realmeetjames · 2026-09-29
- 400 LLM Agents Living in an MMO Server: Lessons on Stale Perception, Fire-and-Forget Actions, and Load Shedding — kristiantalley679 · 2026-09-29
- 3-year-old prediction resurfaces: AI efficiency in bureaucracy may signal the service's end — nwilliams030 · 2026-09-29
- GPT and Claude debated for 25 hours and produced a piece of legislation — patrickkrebs · 2026-09-29
- AI Simulated 100 Papers on LZ Dark Matter Anomaly, Compared Against 82 Real arXiv Papers — skdh · 2026-09-29