The AI Race Reboot: China Catching Up Demands a Strategic Pivot

Gary Marcus argues that the AI market is entering a phase of intensified competition and price wars that could squeeze profits and undermine big IPO plans for OpenAI/Anthropic. He criticizes the current brute-force, internet-scale training paradigm as inefficient, costly to run, and easily replicated, and suggests refocusing from chasing the cheapest LLMs to developing AI that advances science and medicine, while urging policymakers to avoid a zero-sum US-vs-China race.