US-China AI competition could reshape innovation as we know it
In July, two tech leaders, Sam Altman (OpenAI) and Marc Zuckerberg (Meta), outlined opposing visions for how the United States can secure its artificial intelligence (AI) leadership amid strategic competition with China. Their op-ed exchange showed how closely our global technology future is tied to an increasingly zero-sum China debate in the United States. In the background is a new draft bill, the ENFORCE Act, that would take AI export controls to a new level, controlling not just hardware exports, but also access to the models themselves.
Policymakers in Washington increasingly worry that foreign adversaries like China’s government could exploit powerful machine learning models for dangerous purposes, such as developing weapons of mass destruction or launching cyber-attacks. Congress is now working through the details of the ENFORCE Act, which would allow the Commerce Department to impose export controls on AI models with capabilities that could threaten US national security. So far, US export controls targeting China have focused on the chips needed to train AI. Open-source models would likely be exempted, yet the Biden administration is reportedly thinking through the risks of keeping certain frontier models fully open.
Likely in preparation for more AI restrictions, Altman’s OpenAI in June blocked Chinese developers entirely from using its ChatGPT, following earlier actions against malicious uses of its services by state-affiliated actors. In his op-ed, Altman said that some security restrictions are warranted to ensure the US stays ahead. He envisages a future where a US-led club of democracies keeps control of the most advanced AI capabilities.
No comments:
Post a Comment