Meta Platforms CEO Mark Zuckerberg has urged for reduced barriers on open-source AI in the United States to enhance competition against Chinese counterparts. This call came as the social media giant unveiled a more compact and customizable model, named Muse Glimmer, aimed at running tasks on standard personal computers using a single graphics card. Zuckerberg hinted at the imminent release of larger AI models in the future.
The move towards open-weight AI, where core components are shared for user customization, has gained momentum as concerns over escalating AI costs and cybersecurity breaches involving closed models grow. Notable Chinese open-weight models like Kimi K3 and recent incidents involving companies such as Anthropic, OpenAI, and Meta have put a spotlight on the advantages of open-weight models.
While U.S. developers like OpenAI and Alphabet’s Google predominantly use closed weight models, Chinese startups have taken the lead in developing competitive open-weight models. Meta’s stock, which has experienced a 10% decline this year, saw a nearly 3% increase in premarket trading following these developments.
Meta also intends to release Muse Spark 1.2, its cutting-edge model, developed by a specialized team formed last year to regain a foothold in the AI landscape. Mark Zuckerberg emphasized the importance of embracing open-weight models and announced a $1 billion US fund to address concerns regarding the impact of data center expansions on communities.
Zuckerberg highlighted the need for revised policies in the U.S. to support domestic leadership in open-weight models, including data use regulations and distillation techniques. Meta plans to implement a governance structure empowering independent directors to approve model releases based on safety criteria.
The U.S. government’s stance on voluntary safety tests for open-weight AI models was disclosed earlier this month, signaling a shift in approach. Zuckerberg stressed that limiting access to foreign open-source models was not a viable solution to promote American leadership in the field of AI.