iamtrask: AI attribution is closer to solved than you think

iamtrask argues AI attribution is closer to solved than commonly believed, proposing Shapley values for inference, LoRA adapters for post-training attribution, and hierarchical approaches for pretraining—coining 'RGI' over AGI.

2026-09-19 ~ 2026-09-19 · 3 related posts

1 near-duplicate retellings: iamtrask