Meta built its AI reputation on openness — that may be changing

Top member Meta’s new superintelization lab The company’s strong Open Source AI model, move away from Behmoth and instead discussed the development of a closed model, New York Times ReportThe

Sources have told the Times that Meta has completed training about Behemoth, but it has delayed release due to internal performance. When the new Superintelligence Lab was launched, the model was stopped.

The discussion is just that – the discussion. Meta CEO Mark Zuckerberg still needs to sign up any changes, and a company spokesperson Techcunch told Techcunch that the position of the matta in the Open Source AI was “unchanged”.

“We are planning to continue publishing the top open source models,” the spokesperson said. “We didn’t reveal everything we had developed in Histor Tihassically and we hope to continue training the mix of open and closed models.”

The spokesperson did not comment on the potential transfer of Meta away from Behemoth. If this happens so that the meta can prioritize closed-source models, it will identify a large philosophical change for the organization.

Although Meta has internally developed more advanced closed-source models, such as strengthening its meta AI assistant, created Zuckerberg Open source A central part of the company’s external AI strategy – a way to keep AI development fast. He has been in loud position Lama family Openness from competitors such as OpenAI, which is the Zuckerberg Publicly criticized To get more off after being parted with Microsoft. However, the Meta ads are under the pressure of monetization because it gives billions in AI.

These include a lot of signature bonuses and nine-photo salary, creating new data centers, and covering a lot of cost of developing artificial General Intelligence (AGI), or “superintelization”.

Despite being one of the world’s top AI research labs, Meta is still behind the rivals like Open, ethnographic, Google Dipmind and Jai when it comes to commercialization of AI work.

If the meta gives priority to the closed models, it can suggest that exposure was a strategic game, not ideological. Comments from Zuckerberg’s hint to the promise of opening the Sourcing Meta models. Ay Podcast Last summer, he said:

“We are obviously Open Source Pro, but I am not committed to expressing everything we do. I am basically interested in thinking that Open Sourcing will be good for the community because we will benefit from the innovations at one point, but if we do not feel the responsibility of this matter, we don’t feel the responsibility, but it is not very responsible for us. I don’t feel, but it doesn’t feel much responsible for us, but it doesn’t feel like our source, but it doesn’t feel like our source, but it doesn’t feel like our source, but it doesn’t feel like sacrificing us, but it should not be dedicated to us. “

The closed models will give Meta more control and more way of cashing-especially if it believes that the talent it has achieved can provide competitive, best-class performance.

This national shift may re -shape the AI landscape. Open-source speed, originally driven by meta and lamma can be slow, even for the manifestation of Openai’s manifest Still delayed open modelsThe Power can move towards the main players, including electric ecosystem, on the other hand, open-source development may be a grassroots effort. Riple effects will continue throughout the startup ecosystem, especially for smaller companies, focusing on the alignment, protection and model alignment, which depends on access to foundation models.

On the World Stage, Meta’s retreat from Open Source could probably put China into the ground, which Open-Source accepted AI-like DIPSEC and it Munshot AI – as a way to create domestic power and global impact.