What's the strongest mechanistic case that next-token prediction plus RL yields AGI?
Repulsive_Twist_9221 · reddit · 2026-10-13
A Reddit discussion pushing back on the default assumption that AGI is on the way: the OP asks for the strongest mechanistic argument — grounded in causal mechanisms, computational learning theory, and information theory, not trend extrapolation — that next-token prediction plus reinforcement learning can produce open-ended general intelligence.
More from AGI Musings
- Mathematicians are reenacting the 1950s structural chemistry AI anxiety debate — layer07_yuxi · 2026-10-13
- Chemists in the 1950s were furious too: 'press a button and structures came out' — layer07_yuxi · 2026-10-13
- Skeptics push back on 'citizens building personal software with AI' — max_paperclips · 2026-10-13
- Are 3D game demos just stunts? They may signal the same creativity labs need — Angaisb_ · 2026-10-13
- From CSS Zen Garden to headless internet: HCI is becoming HHI — signulll · 2026-10-13
- Amazon now has over 1 million robot workers. What's next for human employees? — lulzxdxdxd · 2026-10-13