"AI Godfather" Jeffrey Hinton recently said something on a PBS talk show in the United States, which caused a lot of discussion in the United States.
He revealed that he went to Shanghai last summer and had a face-to-face conversation with a member of the Political Bureau of the Central Committee. His judgment is that the Chinese Politburo has a much deeper understanding of AI than the U.S. government; they are very worried about related risks and are willing to cooperate with the United States to prevent super intelligence from getting out of control.
As for the United States, Hinton's view is quite pessimistic. He believes that it will be difficult for China and the United States to truly sit together on this matter during the tenure of the current government, and it will take at least another two years. He hopes that after that, a more rational U.S. government will take up this topic.
Hinton's stance has always been clear. He resigned from Google in 2023 so that he could talk more freely about the dangers of AI; after receiving the Nobel Prize in Physics in 2024, he talked about the same thing in almost every public appearance: super intelligence may emerge, but humans have not yet figured out how to control it.
There is a realistic logic behind his evaluation of China this time. If AI security relies on international agreements, the two most critical parties are China and the United States; the United States is relaxing regulations at the federal level and reducing pressure on companies, while China is promoting global AI governance initiatives. In Hinton's view, the party that is more willing to sit down and talk is not the United States.
Naturally, some people will not like to hear these words in the American political context. But the question he raised is real: without coordination among major countries, can the increasingly powerful model be controlled by relying solely on the self-discipline of a few companies.