Meta Platforms CEO Mark Zuckerberg has urged for reduced U.S. barriers on open-source AI to enhance competitiveness against Chinese rivals. The company introduced a more compact and customizable model named Muse Glimmer, tailored for running tasks on individual devices with a single graphics card. Zuckerberg hinted at the upcoming release of larger AI models in the near future, emphasizing the importance of democratizing AI access rather than concentrating power in a few hands.
This shift towards open-weight AI aligns with a broader industry trend as businesses seek cost-effective and flexible AI solutions following security breaches involving closed models. Open-weight models, such as the Kimi K3 from China, offer customizable components, enabling users to adapt them for specific needs and are cost-efficient compared to closed models.
Meta, alongside other tech companies, backed open-weight AI through a joint letter supporting its adoption. While Chinese startups have excelled in open-weight model development, U.S. firms like Meta are now focusing on advancing their models to compete effectively. Meta’s shares rose in premarket trading following the announcement of Muse Spark 1.2, their advanced model developed by a specialized team.
Zuckerberg highlighted the need for the U.S. to rethink policies around data use and model distillation to lead in open-weight AI innovation. He announced a governance structure to empower independent directors to approve safety standards for model releases. Meanwhile, the U.S. administration’s decision not to impose voluntary safety tests on open-weight AI models could impact the industry’s future development.