📰 Key Highlights

George Hotz, founder of Comma AI and well-known jailbreaker, recently published a post strongly criticizing a policy white paper titled “AI 2040: Plan A.” Authored by the AI Futures Project, the white paper advocates that researchers worldwide should collectively agree to slow AI development for up to 14 years in exchange for humanity’s long-term well-being.

Hotz’s core opposition stance is that AI progress shouldn’t be subject to unified control for the sake of “collective benefit.” He argues that the doomsday scenarios of AI “rapidly rising” and quickly surpassing humans don’t hold up logically — the real solution isn’t centralized regulation, but rather making AI models highly tailored to individual users’ needs, with the user themselves in control locally. He likens this kind of “user-aligned AI” to a gun — the tool itself doesn’t refuse to obey based on the purpose of use. A truly user-aligned AI should be able to help you order equipment for a meth lab, or plan out whatever it is you want to do. He’s even said he’s willing to die for this principle of freedom.

The author of the piece takes a reserved stance. He agrees that the vision of personalized, decentralized AI has its appeal, but points out that any system involving a large population — whether social, market, or tech product — inevitably has to find a balance between individual freedom and collective accountability. Pure alignment with individual interests means that the safety of spouses or bystanders who haven’t yet been harmed is barely taken into account. That’s clearly not a simple binary choice of “freedom or not.”

💬 JudyAI Lab Perspective

George Hotz’s public takedown of the “slow AI down for 14 years” policy white paper moves the question of “who gets to decide the pace of AI development” from the tech community onto the policy stage — and it’s a fundamental disagreement we should all take seriously.

The core tension in this debate is whether AI should align with “the individual user’s will” or with “broader societal safety.” Taking Hotz’s “user-aligned AI” logic to its extreme is essentially arguing that AI should fully obey the user, no matter the purpose. For AI builders, this is a design philosophy question: every usage boundary you set is an implicit value judgment. The article’s author also points out that personalized AI has its appeal, but once you reach scale, system effects spill past individual boundaries. “Whose interests get covered, whose get excluded” isn’t a freedom-or-not binary — it’s a question every designer will eventually have to answer.

Next time you’re designing the usage boundaries of an AI product, it’s worth asking first: who does this constraint protect, and who ends up paying the cost?

📅 Source Info

🔗 Further Reading