posted in Technology
AI’s recursive self-improvement might not come so quickly after all
www.technologyreview.com/2026/08/18/1142188/ai-recursive-self-improvement/
Replying to @inari@piefed.zip
A year ago they were similarly bad at writing code, often created unit tests that tested nothing, etc.
If the models are trained in what they’re doing wrong, that can accelerate their progress toward doing it right.
Replying to @MangoCats@feddit.it
They would need to be trained for open ended creative tasks, which is just hard in the current reinforcement learning paradigm.