AIFIRST "Next-token predictor" is the wrong mental model for LLMs
The brief
A blog post argues that calling LLMs "next-token predictors" is a misleading mental model that fails to account for the planning and reasoning behaviors observed in modern systems.
Key points
- The author proposes a more accurate conceptual framing to better explain what large language models actually do.
Sources
- HNgmcgoldr.github.io