Meta Expands AI Openness, Zuckerberg Pledges More Power to Individuals

Meta CEO Mark Zuckerberg is championing a new strategy focused on broadening the accessibility of artificial intelligence (AI) and making substantial investments in communities that host the company's data centers. This approach aims to ensure that the wealth and opportunities generated by AI are shared not only by large corporations but also by individuals and local communities.

“

AI models should be deployed as broadly as possible, and many decisions about AI's application and value should be left to the end-users.

In a detailed essay, Zuckerberg argued that the most effective path forward for AI involves deploying AI models as broadly as possible and entrusting end-users with many decisions regarding AI's application and value. He expressed concern that if AI development remains concentrated within large institutions like corporations and governments, it could lead to an excessive concentration of power in the hands of these entities, rather than individuals.

To counter this, Meta is releasing "Muse Glimmer," a new open-weight model that allows users to download and adjust the numerical values governing the AI's operations. The company also plans to release an open-weight version of its advanced "Muse Spark 1.2" model in the coming weeks. This move is significant amidst ongoing debates about the safety of developing advanced AI through open-source or open-weight methods, with some researchers advocating for strict control over core technologies due to risks of misuse.

“

If AI development is concentrated within large institutions like corporations and governments, it could lead to an excessive concentration of power in the hands of these entities, rather than individuals.

Zuckerberg also proposed AI safety measures, suggesting federal government collaboration with companies to verify the safety of new AI models and enable developers to analyze each other's models. Meta's board of directors will play a larger role in future model safety standards. The strategy also includes a $1 billion fund for communities with data centers, aiming to share economic benefits and mitigate local opposition to expanding AI infrastructure. This investment is seen as crucial for securing social support for AI industry growth, especially as the U.S. faces challenges in AI development infrastructure compared to China.

“

We are going to release an open-weight version of our most advanced AI model, Muse Spark 1.2, in the coming weeks.