Free idea for labs: train on synthetic data from Tatsu's DL alchemy quizzes for RSI
jasondeanlee · x · 2026-11-22
A free idea for AI labs: feed the quizzes and homework from Tatsu's deep learning alchemy class into a synthetic data generator, then hill climb. The author argues this should at least buy some "RSI" or good-science-of-DL abilities for pretraining.
He notes he briefly deleted the post fearing it contained alpha, then reconsidered — the idea is probably obvious enough that labs already do a version of it.
Related event: Stanford's Deep Learning Alchemy course sparks synthetic data ideas(3 posts)→
More from Research
- ReSFlow rectifies categorical flows on statistical manifolds for one-step generation — jiqizhixin · 2026-11-22
- Ben Goertzel on the AI Riemann proof: 'a new era of math'—and math glory now belongs to those with tokens — bengoertzel · 2026-11-22
- LessWrong: looping boosts no-CoT reasoning far more than general skills in NVIDIA's Dory 4B — CFGeek · 2026-11-22
- "Hilbert's program vindicated": Lean brings us near a practical decision machine — teortaxesTex · 2026-11-22
- Simulation scores, not vibes: when can you claim you've explained a neuron? — burny_tech · 2026-11-22
- Model harnesses a year on: filesystem maximalism wins, long-lived named agents rise — voooooogel · 2026-11-22