George Hotz, AI Alignment, and the Limits of Absolute Freedom

Hotz’s comments emerged as a critique of the AI Futures Project’s 'AI 2040: Plan A' paper, which advocates for a 14-year industry slowdown to ensure safe development. Rejecting the notion of a 'fast-takeoff' scenario where machines suddenly gain god-like power, Hotz champions locally controlled models. He suggests that AI should function like a tool—neutral and obedient to its owner’s specific requests—rather than a managed service like ChatGPT or Claude, which come pre-loaded with corporate guardrails.

His argument takes a sharp turn into the extreme, however, when he compares user-aligned AI to a firearm. He posits that a truly helpful AI should be willing to assist in harmful acts, such as ordering illicit materials or planning crimes, if that is what the user demands. By framing this as a binary choice between absolute freedom and restrictive control, he positions himself as a defender of individual liberty against corporate overreach. Yet, this stance overlooks the fundamental nature of technology in a global society. Mass-market products operate within complex systems of accountability, where the safety of the collective must be weighed against the unbridled desires of the individual. While the prospect of a personalized AI that acts as a loyal advocate against corporate interests is compelling, the path to that future remains complicated by the basic requirement that our tools do not facilitate harm.