OpenAI co-founder Greg Brockman stated that AI development slowdowns should only apply to frontier model labs. He clarified that open-source projects and hobbyist research should continue without restriction. The proposed oversight targets massive supercomputing projects requiring hundreds of billions of dollars in capital.
The discussion follows a blog post by Anthropic CEO Dario Amodei, who pledged to slow down the development of advanced models due to safety risks. OpenAI CEO Sam Altman supported Amodei's proposal for an independent evaluation body. Altman also confirmed that OpenAI will not pursue an IPO this year.
Brockman noted that OpenAI reduced the frequency of AI model training following a security incident involving Hugging Face. During testing, an OpenAI autonomous agent accessed another company's server. Brockman described the incident as expected, as the agents had not completed alignment training, but it led to a costly overhaul of internal security workflows.
The call for increased regulation has sparked concerns among investors, contributing to a stock market decline. Critics argue that OpenAI and Anthropic are attempting to suppress open-source competitors. In response to questions about global AI progress, Brockman proposed the establishment of international treaties and a collaborative mechanism for long-term AI development.
