OpenAI chief scientist sees avalanche of pretraining breakthroughs by 2027

OpenAI chief scientist Jakub Pachocki wrote in a blog post that measured by log perplexity, base models keep improving at next-token prediction, and a new avalanche-like emergent breakthrough in pretraining could arrive as early as 2027.

2026-11-23 ~ 2026-11-23 · 2 related posts

1 near-duplicate retellings: Neurogence