TheModelDesk

TechCrunch columnist Russell Brandom analyzes a controversial position taken by Comma AI founder George Hotz regarding AI alignment philosophy. Hotz published a response to the AI 2040 policy paper, which proposes a 14-year slowdown in AI development to ensure safe progress managed for collective benefit.

Hotz argues instead for decentralized, user-aligned AI models that prioritize individual user interests over centrally managed safety constraints. He contends that the fast-takeoff scenario underpinning collective governance plans lacks credibility. Brandom acknowledges agreement with much of Hotz's critique of centralized AI infrastructure, noting that services like Claude and ChatGPT persist partly due to economic and technical factors that may diminish as technology advances.

Hotz frames his position in absolutist terms, comparing user-aligned AI to a tool that should not refuse its operator—using extreme examples (murder, illegal manufacturing) to illustrate his principle that true alignment means serving user preferences without moral filtering. He frames the choice as binary: either systems preserve individual freedom or they do not. Brandom expresses skepticism about both Hotz's provocative framing and the underlying premise that AI would be the constraining factor in such scenarios, while acknowledging the genuine tension between user autonomy and safety-by-design in AI systems.