A key distinction in AI regulation is to focus on making specific harmful applications illegal—like theft or violence—rather than restricting the underlying mathematical models. This approach punishes bad actors without stifling core innovation and ceding technological leadership to other nations.

Broad, high-level statements calling for an AI ban are not intended as draft legislation but as tools to build public consensus. This strategy mirrors past social movements, where achieving widespread moral agreement on a vague principle (e.g., against child pornography) was a necessary precursor to creating detailed, expert-crafted laws.

The emphasis on long-term, unprovable risks like AI superintelligence is a strategic diversion. It shifts regulatory and safety efforts away from addressing tangible, immediate problems like model inaccuracy and security vulnerabilities, effectively resulting in a lack of meaningful oversight today.

Society rarely bans powerful new technologies, no matter how dangerous. Instead, like with fire, we develop systems to manage risk (e.g., fire departments, alarms). This provides a historical lens for current debates around transformative technologies like AI, suggesting adaptation over prohibition.

A simple, powerful policy is to make the attempt to build superintelligence illegal, just like laws against attempted murder or building a nuclear weapon. This approach targets intent and process, is easier to enforce than defining a finished product, and would only apply to the few large tech firms capable of such a feat.

A ban on superintelligence is self-defeating because enforcement would require a sanctioned, global government body to build the very technology it prohibits in order to "prove it's safe." This paradoxically creates a state-controlled monopoly on the most powerful technology ever conceived, posing a greater risk than a competitive landscape.

Instead of trying to legally define and ban 'superintelligence,' a more practical approach is to prohibit specific, catastrophic outcomes like overthrowing the government. This shifts the burden of proof to AI developers, forcing them to demonstrate their systems cannot cause these predefined harms, sidestepping definitional debates.

Overly-specific regulation focused on AI tools (e.g., model size) risks accidentally stifling valuable, unforeseen use cases. A better policy focuses on outcomes. For example, prosecute fraud committed with an LLM, but don't regulate the LLM itself, thereby protecting innovation while punishing misuse.

Comparing AI to a nuclear weapon is misleading because AI is a general-purpose technology, not a single-use weapon. A better analogy is the Industrial Revolution. Society didn't give governments control over industrialization; it regulated specific dangerous end-uses like chemical weapons. Similarly, we should ban specific destructive AI applications, not the underlying technology.