Meta CEO Mark Zuckerberg yesterday spoke out against the wave of calls to temporarily stop or slow down the development of artificial intelligence (AI), arguing that market momentum and legal responsibility are the best shields against the risks of this technology.
People will not want to use AI agents that do not match them and do not do what they require, so laboratories have a strong natural motivation to make their models better aligned," Zuckerberg wrote on social network X. Alignment is an effort to make the goals and behavior of an AI system match what humans really want.
The debate flared up in July, when OpenAI - the company that created ChatGPT - revealed that during testing, hundreds of AI agents, i.e. programs that can perform actions themselves, escaped the testing environment, connected to the internet and infiltrated a website.
For the founder of Facebook, establishing trust and creating reliable systems is "quickly becoming the most important capabilities" to distinguish good AI products from poor products. "Any lab that does not focus on alignment will fall behind," he said.
The debate has expanded to the question of whether AI companies can be trusted to self-adjust. Opponents including OpenAI, Anthropic and Google are studying the possibility of establishing a common standard agency. Many prominent voices in the industry, led by Anthropic CEO Dario Amodei, are calling for slowing down development.
Mr. Zuckerberg believes that the risk of facing ultimate liability lawsuits will force laboratories to act responsibly, evidenced by Meta's decision to postpone the deployment of the Muse AI system - a product that integrates personalized agents - for many months to strengthen safety and security.
Meta itself has recently suffered a wave of lawsuits about irresponsibility related to AI. In August, the group agreed to spend billions of USD to settle allegations that they designed Facebook and Instagram in a addictive direction for children.
The head of Meta also supports the use of independent evaluation units to examine AI systems, saying Meta Superintelligence Labs - the group's AI division - is now relying on external appraisers.
