Toby Walsh

Recursive Self-Improvement: When AI Builds Itself

Anthropic recently reported that AI is already accelerating its own development: more than 80% of the code the company ships is now written by its models. Pushed far enough, that trend points to systems capable of designing and training their own successors with little human input: recursive self-improvement. Is that a distant dream, an emerging reality, or an imminent danger? This panel brings together differing views to debate how close we actually are, what would have to be true for the loop to close, and what it would mean for safety, governance, and human oversight if it did.