Dev: drop the AGI chatter until models get continual learning — frozen weights will never be general enough
cgarciae88 · x · 2026-09-28
Developer cgarciae88 argues that talk of AGI should be shelved until continual learning arrives: models with frozen weights can never become general enough. The claim targets the core limitation of current LLMs — training and deployment are decoupled, so models can't continuously absorb knowledge from interaction and only compensate with static snapshots bolted onto retrieval or external memory.
More from AGI Musings
- Flying supercar, persistent agent, 750 TPS: a wild week of AI launches ahead — ChrisGPT · 2026-09-28
- Fermi estimate: brain may pack 500k-5M molecular switching units per 'parameter' — JosephJacks_ · 2026-09-28
- Pundit mocks 'a million digital employees' fantasy: you don't manage what outsmarts you — danfaggella · 2026-09-28
- Gen-Z is starting to admit they enjoy AI-generated content — LaillaElgohary · 2026-09-28
- Austin car theft rose 24% vs trends after switching off Flock license plate cameras — garrytan · 2026-09-28
- "Alignment is easy, just look at Opus 3"? Researcher pushes back: capable systems will break it — JacquesThibs · 2026-09-28