Skip to content

Avoiding Negative Side Effects: Concrete Problems in AI Safety part 1

160,945 views

About this video

Hand-picked by a BrighterTube curator · found via HackerTube (Karpathy's dataset of videos Hacker News recommends

Robert Miles opens his series on 'Concrete Problems in AI Safety' with negative side effects: why an agent rewarded for one goal may wreck everything else, and proposed fixes like impact penalties; a careful explainer of the reasoning.

Robert Miles explains the 'negative side effects' problem from the Concrete Problems in AI Safety paper, with examples and proposed approaches.

Video by Robert Miles AI Safety , played through YouTube's embedded player.

Part of Concrete Problems in AI Safety, Technical