Definition
Plain language
The idea that AI should keep learning from its own trial-and-error during use, rather than only from a frozen pile of training data.
As stated in the literature
A framing from Silver and Sutton arguing the next phase of AI progress comes from agents that learn at test time from self-generated experience; cited as motivation for note-taking and self-improving long-lifecycle agents.
Why it matters: It points toward AI that keeps improving from its own use rather than staying frozen, which could let systems adapt to new tasks without retraining from scratch.
For example, instead of relying only on what it learned before release, an assistant might keep notes on what worked while helping you and use them to do better next time.
Heard on the show
“There's this idea from Silver and Sutton last year — the "era of experience" — agents that learn from their own experience at test time rather than only from a frozen training set.”Episode 160 — Training an AI to Take Its Own Notes, So Its Future Self Works Better