Schmidhuber says neural-network distillation was published in 1991
SchmidhuberAI · x · 2026-07-23
Jürgen Schmidhuber argues that neural-network distillation was already published in 1991, and that later work merely repackaged the idea.
He frames the discussion by pointing to a 2025 DeepSeek-related debate about who invented distillation, then says the original paper described a general teacher-student compression principle: a student network can learn by imitating a teacher’s behavior, thereby compressing knowledge into a smaller model.
More from Research
- Open-source decision models sold their own base in Red Alert 2 — then learned until no human could beat them — heathercmiller · 2026-10-27
- DeepMind paper: LLMs can't sample randomly on their own — pass random numbers from code instead — rohanpaul_ai · 2026-10-27
- World Model Assigns Weight and Friction to Objects, Then Simulates with a Physics Engine — jonstephens85 · 2026-10-27
- Token Collapse: Study Finds Graph Soft Prompting Tokens in LLMs Fail to Encode Structure — Sapienza · 2026-10-27
- Mathathon 2.0 launches with Terence Tao's SAIR, $1.5M in compute for teams — _sathvikr · 2026-10-27
- The 'sweet lesson' is real: two kinds of ML problems need opposite strategies — mike64_t · 2026-10-27