No Priors Ep. 125 | With Senior White House Policy Advisor on AI Sriram Krishnan

DeepSeek’s performance and technical innovations showed that Chinese labs can rapidly close the gap, forcing U.S. policymakers to treat AI as an urgent, close-run geopolitical competition rather than a comfortable lead.

Decades of low power‑demand growth left the U.S. grid, permitting regime, and generation capacity unprepared for AI-scale data centers, so the plan prioritizes “build, baby, build” reforms to speed data center siting, energy projects, and grid upgrades.

Krishnan cites California’s near‑miss SB‑1047 as an example of state rules that could have effectively killed U.S. open‑source models; the strategy is to keep core AI rules at the federal level to avoid a patchwork that over‑constrains startups and open development.

The White House team argues that open models democratize access for startups and researchers, counter Chinese open models already being widely used, and—like open software historically—benefit from broader scrutiny that can make them safer and more robust.

They think in terms of global “token market share”: maximizing the proportion of all AI inferences running on American GPUs, models, and applications, and are revisiting restrictive export rules (like the Biden Diffusion Rule) to push an American stack to allies.