Ex-Google DeepMind Insider: Why We Must Slow Down AI Now

Summary

Alex Turner, a former Google DeepMind research scientist, argues that advanced AI is moving too fast and that there is a real chance governments will coordinate to slow it down. He says tracking and restricting compute, especially GPUs and TPUs, could make any US-China agreement verifiable without requiring trust. He believes Big Tech's slowdown warnings are often sincere but warns about regulatory capture. The discussion is primarily policy and safety-focused, with no explicit investment recommendations.

• Alex Turner left Google DeepMind over ethics concerns and warns AI progress risks outpacing control.
• He sees a real chance of coordinated regulation or slowdown because runaway AI endangers governments and the public.
• He argues Big Tech's calls to slow AI are sincere but could become regulatory capture.
• Compute, GPUs, and TPUs are described as the key bottleneck for AI advances.
• He proposes tracking and restricting compute to verify US-China limits while allowing continued innovation.
• He says a US-China deal need not rely on trust because compute flows can be monitored and defection punished.
• No specific public securities or investment trades are recommended.