Calls to regulate AI are premature and, in several cases, self-serving. ChatGPT launched 9 months ago. GPT-4 launched 6 months ago. The harms cited to justify regulation, including bioweapon synthesis, cyberattacks, and mass misinformation, remain theoretical. No coherent logic chain explains how existing laws fail to address them. The loudest voices calling for regulation are incumbents like OpenAI and Google, who would benefit directly from rules that freeze the competitive landscape in their favor.
The cost of getting this wrong is not abstract. Google's Med-PaLM2 already outperforms physician experts on medical benchmarks, with results suggesting that adding more physician input to training actually degrades performance. AI tools at this capability level, deployed without regulatory friction, could extend expert-quality healthcare and education to populations that have never had access to either. Premature regulation does not pause harm. It pauses that.
The original piece is worth reading for its breakdown of how regulation structurally shifts industries from user-centric to government-centric, and for the narrow carve-outs where regulation is actually warranted, specifically advanced chip exports to foreign adversaries. The argument is not that AI needs no guardrails ever. It is that the technology needs room to mature before the rules calcify around today's incumbents.
[READ ORIGINAL →]