Zuckerberg warns centralised AI risks 'abandoning our values'

Meta CEO Mark Zuckerberg has criticised the approach of leading AI firms OpenAI and Anthropic, warning that tightly controlled AI development could centralise power and stifle innovation. In an interview with The New York Times published on July 28, and an opinion piece in The Wall Street Journal, Zuckerberg argued for open AI models that empower individuals rather than concentrating power in a few hands.

Without naming specific companies, Zuckerberg said that if AI research labs continue developing the technology in a controlled manner, it would be akin to 'abandoning our values' and would narrow people's access to AI. He emphasised 'individual empowerment, invention and balance of power' over centralisation.

Zuckerberg's remarks come amid a broader divide in Silicon Valley over AI openness. While Anthropic CEO Dario Amodei and others argue that some advanced AI models are too dangerous to share without tight controls, tech giants like Microsoft, Nvidia, Google, and Meta advocate for open development to drive progress and business opportunities.

The split was highlighted when Nvidia CEO Jensen Huang penned a letter urging US policymakers to avoid broad restrictions on open-weight AI models. The letter has been signed by over two dozen companies, including Meta, Microsoft, OpenAI, Google, and SpaceX. Anthropic is the only major frontier AI lab that has not signed.

Meta's own AI philosophy has shifted. Initially championing open-source AI with its Llama models, Meta later invested $14.3 billion in ScaleAI and hired its founder to revamp its AI division, now called Meta Superintelligence Labs (MSL). Earlier this year, MSL released its first in-house AI model, Muse Spark, which was closed-source. However, Meta has said it will continue developing open-weight models.

Zuckerberg criticised the 'doom-laden' discourse from other labs, saying, 'There needs to be a voice or several voices that are bringing realism to this debate.' He expressed scepticism about a single 'benevolent' superintelligence aligned with all humanity, calling it 'literally impossible' because people have different values. Instead, he advocated for 'personalised superintelligence' for everyone, configured to suit individual needs.

'I don't know if the goal is to have everyone agree on everything,' Zuckerberg said. 'I believe the general arc of the industry has been toward more openness and putting the power of tech into more people's hands, not fewer.'