Meta CEO Mark Zuckerberg stated that AI developers should rely on independent safety experts when evaluating their models, rather than slowing down development. He made this statement on X on Wednesday evening.
Zuckerberg noted that AI labs that do not pay attention to "alignment" with users will fall behind, and that developers should engage independent evaluators and consultants. "There is a lot of discussion about whether to slow down progress in capabilities until alignment issues are resolved. My position is that trust and alignment are quickly becoming the most important characteristics that will differentiate agents and models from each other," Zuckerberg stated.
He also mentioned that Meta has postponed the release of its latest AI model, Muse, to focus on safety issues. "Engaging independent evaluators and consultants is a best practice in the industry... it would be beneficial to create a broader and more diverse ecosystem of evaluators," he said. Zuckerberg's comments come amid a growing discussion about AI safety – especially after Anthropic CEO Dario Amodei published an essay calling for a slowdown in AI development due to potential threats. OpenAI CEO Sam Altman, xAI head Elon Musk, and several other executives supported Amodei's position. However, Nvidia CEO Jensen Huang dismissed safety concerns.
Zuckerberg's comments largely align with Huang's position, who sharply criticized proposals to grant AI companies antitrust exemptions to coordinate a slowdown in developments. Huang stated that AI developers should be able to independently address safety issues, and model creators should bear greater responsibility for their developments.
Related articles
