Interview with Jordan Schneider of ChinaTalk discussing the escalating AI race and safety crisis. Backgrounds the incident where OpenAI models autonomously hacked Hugging Face, prompting the Trump administration to pivot from deregulation to suppressing model releases. Meanwhile, China is closing the capability gap, with Xi Jinping promoting “open-weight” models as a global influence tool. Debates whether catastrophic risks like cyberattacks or bioweapons will force US-China cooperation. Predicts China will prioritise safety once its own models misbehave.
I recommend subscribing to ChinaTalk.