What smart people are saying about Meta’s open-weight model Muse Glimmer and Mark Zuckerberg’s AI manifesto

Meta’s former chief AI scientist Yann LeCun praised the release.

“Good move. Bravo,” LeCun wrote on X alongside three clapping emojis in a reply to Zuckerberg’s announcement.

LeCun hasn’t always been quick to praise his former employer. In January, he called Alexandr Wang, Meta’s chief AI officer, “inexperienced” and predicted Meta would lose more AI talent.

After leaving Meta, LeCun co-founded AMI Labs, which focuses on so-called “world models” rather than the large language models that form the foundation of chatbots like OpenAI’s ChatGPT.

Box CEO Aaron Levie said the news was much bigger than Meta.

“America now finally has its response to the open weights AI race,” Levie wrote on X.

He said the release “is broadly net positive for AI – even frontier closed models – because it helps drive further diffusion of AI.”

“This will continue to help drive down the cost of intelligence,” he wrote, “it allows companies to run models as they see fit, as well as allows them to post train their own models for specific use-cases to get better performance gains in their domain.”

Venture capitalist Sriram Krishnan praised Meta’s release.

“great to see meta back launching open weight models. congratulations to @alexandr_wang and the msl team,” Krishnan wrote on X.

Krishnan, who was a White House advisor until recently, also lauded Zuckerberg’s essay.

“Great piece from @finkd – especially the theme of AI being something that is personally empowering,” Krishnan wrote.

Y Combinator policy head Luther Lowe said Zuckerberg had the right message but was the wrong person to deliver it.

“The message is compelling; the messenger has a credibility problem,” he wrote.

Lowe said that Meta’s decision to ban rival chatbots from accessing WhatsApp Business API doesn’t gel with Zuckerberg’s broader message about making superintelligent AI widely available.

“If Meta believes in a balance of power that favors individuals, that principle should include the individual’s right to choose an AI agent on Meta’s own platforms,” Lowe wrote.

In June, the European Commission ordered Meta to allow third-party AI assistants free access to WhatsApp. Meta said it disagreed with the decision and would appeal it.

Hugging Face CEO Clément Delangue expressed excitement for the release.

“Meta is back! well done @finkd @alexandr_wang!” Delangue wrote on X.

Delangue, who has pushed for open models, previously said that Hugging Face had to turn to a Chinese open model to help it address OpenAI’s models breach of its systems because proprietary AI models had safeguards in place that prevented the company from getting the help it needed.

Johns Hopkins University Professor Seth Lazar said Zuckerberg’s “overall vision is cool” and expressed optimism if Meta’s CEO’s manifesto serves as a “North Star” for the company’s SuperIntelligence Labs.

“If this constitutes a genuine North Star for MSL, then I am a bit more optimistic about the world we’re likely to build with powerful AI (conditional on getting there without everything blowing up along the way),” Lazar, who studies the moral and political philosophy of AI, wrote on X.

Lazar said he was “psyched” that Meta is releasing a model.

“Sometimes protecting individual rights means giving people an agent who will push on their behalf against rules and restrictions that are over-applied and unfairly demanding,” he said. “The frontier models are absolutely terrible at this, and our reliance on them portends an Internet that makes us much less free.”

Lazar said others in the AI community would likely find Zuckerberg’s vision for superintelligence disappointing. One area Lazar said is right to critique is that the Meta CEO sees compute as the main limitation to the use of AI models that exceed human intelligence. Lazar said it’s more likely that other limitations will be needed.

“Pretty likely that it’ll involve restrictions on some capabilities, and probably this means eg the personal health research scientist is not a plausible vision, or will need a lot more structure, anyway,” Lazar wrote.

Wharton professor and AI expert Ethan Mollick said Meta’s announcement that it will release the weights for Muse Spark, its current most advanced closed model, was the biggest news of all.

“Spark is the big news and is a good model,” Mollick wrote on X. “Not quite at the frontier of open models from China, and still well behind the closed frontier, but the best non-Chinese open weights model released in a year.”

What matters next, Mollick said, is what Meta does going forward.

“Of course, a lot depends on continuing to release new open models to keep up.”

The Trump White House was quick to praise Meta’s launch of a “Future Is For Everyone Fund” that will support communities where it builds data centers and AI infrastructure.

“President Trump’s vision for American AI dominance continues to drive results,” the White House’s rapid response account wrote on X, sharing CNBC’s coverage of Meta’s announcements.

Zuckerberg wrote that community investment led to greater support for projects such as data centers.

“We see that communities we invest in over the long term are more supportive of development than communities where speculators start building with minimal investment in the community,” he wrote.

Nvidia, the world’s largest company by market cap, offered its praise of Meta’s announcement.

Meta has built much of its AI work on Nvidia’s chips. Nvidia CEO Jensen Huang has also pushed the development of open models. Nvidia has its own family of open models called Nemotron.

“Great to see @AIatMet back publishing open models 🙌,” Nvidia’s AI account wrote on X.