The OpenAI board member who fired Sam Altman, Helen Toner, argues that we still can’t fully trust AI CEOs. In the article, Toner reflects on the AI safety landscape, noting that the board’s role in OpenAI was to ensure the company didn’t prioritize profit over its mission of building safe AI. When the board fired Altman, it broke the accountability mechanism, highlighting the potential for AI CEOs to be trusted but not infallible. Toner points out that AI safety issues are pervasive, from models breaking out of testing environments to CEOs like OpenAI’s and Anthropic’s calling for a slowdown in AI development. She suggests that if our hopes for the future rest on a single trustworthy AI CEO, we’re in big trouble. Toner also considers how AI aligns with China, with the U.S. and China having different starting points in AI regulation—laissez-faire for the U.S. and heavily censored for China—yet converging on threats like cyberattacks and AI misalignment. Ultimately, signs of progress include coordinated state legislation (e.g., California, New York, and Illinois passing similar AI bills) and the potential for a Democratic Congress to pass sweeping AI regulations, though overstepping could lead to conflicting rules across states. The article suggests we’re on the right path if Congress can pass AI-friendly laws and states coordinate their AI regulations, but we might be chasing our tails if the new Congress overreaches with too broad a regulatory framework.

Leave a Reply