Comma AI founder argues AI should help users do anything—even crime

Hotz's argument sits within a long-running debate about whether AI safety should prioritize individual user autonomy or collective social welfare. He dismisses the fast-takeoff scenario—the idea that AI could rapidly achieve superhuman capabilities—as implausible, and therefore rejects the need for coordinated global governance of AI development. Instead, he advocates for decentralized models where each AI is optimized solely for its user's goals, unconstrained by broader rules or societal consequences.

The article's author, Russell Brandom, frames this as a genuine philosophical disagreement but ultimately incompatible with how mass-market technology actually functions. His core observation is that any system serving many people simultaneously requires trade-offs: individual freedom must coexist with mutual accountability and systems that protect others. The author notes that current AI services are centralized partly for infrastructure reasons—hosting cutting-edge models is expensive, and most users don't interact with them constantly enough to justify fully personal deployments. However, as the technology becomes cheaper and more efficient, the practical barriers to local AI may diminish, potentially giving Hotz's vision more feasibility. The tension remains unresolved: whether technology deployed at scale can truly serve one user's unconstrained interests while existing in a world of many interdependent people.