Transformers may be doing something more interesting than predicting the next word
soleio · x · 2026-08-25
Soleio quotes an essay arguing that "it's just a statistical machine predicting the next token" is one of the most misleading descriptions of transformers.
The core argument: an objective tells you what a system is optimized to do, but not what internal computation the system discovers to do it. When researchers look inside transformers — and, independently, when neuroscientists study populations of neurons — an interesting commonality emerges, suggesting richer internal structure than mere word-level statistics.
More from AGI Musings
- Post-work world: identity shift and the need for self-confrontation — davidpattersonx · 2026-08-25
- The Economist: Humans face moral reckoning if AI becomes conscious — AnnaCiaunica · 2026-08-25
- Frontier AI companies will trigger unprecedented wealth and power concentration — LuizaJarovsky · 2026-08-25
- Researcher abandons second brain: 90% of personal wikis fail in 3 months — solyarisoftware · 2026-08-25
- Neuroscientist: Consciousness is likely a property of life, not computation — anilkseth · 2026-08-25
- Tech progress is a series of step functions, not a linear rise — chris_j_paxton · 2026-08-25