Meta co-founder Mark Zuckerberg has advocated for involving independent experts to evaluate the safety of AI models instead of slowing down the entire industry. He wrote this on X.
"Every developer is responsible for working at the pace required to safely train models and has the ability to independently take the necessary measures to do so," he wrote. "Involving independent experts for assessment [of risks] and consultants is a best practice in the industry. Meta Superintelligence Labs (MSL) already applies this approach in several areas, as it allows for better results in work. Other developers can do the same."
Zuckerberg noted that Meta had previously delayed the launch of its AI agent Muse for several months due to safety concerns and did not make it a condition for finalizing its product to require other companies to delay the release of their new agents. "We just did it and treated it as part of our daily work responsibilities," the entrepreneur said.
According to the Meta co-founder, AI companies will be forced to pay due attention to ensuring the safety and controllability of AI, as these indicators "are rapidly becoming the most important characteristics distinguishing AI agents and AI models." Developers who do not prioritize this will fall behind, Zuckerberg believes.
On September 12, the director of the American company Anthropic, Dario Amodei, published an article calling for slowing down the development of advanced AI models. He noted that through a coordinated slowdown in work on advanced AI technologies, AI companies could "gain one or two years of additional time before models reach critical levels" of development and avoid serious miscalculations. This forecast was agreed upon by the founder of SpaceX, the wealthiest person on the planet, Elon Musk, as well as the director of the American technology firm OpenAI, Sam Altman.
