A podcast transcript between Nathan Lambert (AI2) and Dean Ball (Hyperdimensional) discussing the implications of the U.S. Department of War designating Anthropic as a supply chain risk. They explore how government control over closed AI models is likely to accelerate demand for open-weight models as a geopolitical insurance policy, while acknowledging serious short-term headwinds: the widening capability gap between frontier closed models and open alternatives, the near-impossibility of funding open model development at trillion-dollar scale, and the usability challenges of open models. They also discuss sovereign AI initiatives, the risk of U.S. government nationalization of frontier labs, and the idea of financializing compute as a commodity market to democratize AI production.
