DeepMind Researcher Quips: RL 'Race to the Top' Boosts Capabilities, Not Safety

tomekkorbak · x · 2026-11-22

DeepMind researcher tomekkorbak quipped at Anthropic co-founder Eric Mitchell that the "race to the top" in reinforcement learning was supposed to lift safety standards across labs — instead, everyone seems to be racing to improve capabilities. A pointed jab at the safety-vs-capabilities tension in current RL competition.

Original post →

More from Fun

Fun channel →