Why Will Governments Regulate AI Compute?

David Friedberg argues that governments will regulate AI compute and assign who gets access to how much of it. Laws would target the compute layer after the chips are produced, so officials hand out capacity instead of controlling the chips themselves. Governments on the right and the left, in his view, will fight to subsume AI or keep control of it.

Humans form tribes, communities, and governments to build power for their members. In a democracy, government is a coalition aimed at what most voters want. Nonprofits, companies, and governments then compete for limited resources, because more resources mean more power for the people they represent. AI threatens government organizations through media, capital accumulation, the power to change minds, and higher productivity. AI is already drawing more interest from capital than governments, and governments will not let that continue.

In the near term, more capital is going to AI organizations. A handful of companies are competing with governments for capital on one narrow, one-way bet that attracts money otherwise headed to a two, three, four, five, or 10-year bond auction. He expects governments to try to stop data centers so that capital does not keep shifting to AI organizations.