The head of META* Mark Zuckerberg hinted that the company can change its approach to openness in the development of AI. In a recent appeal, he spoke about the plans to create “supernight” AI and noted that not everything will be available in the public domain due to security reasons.
Previously, META* actively promoted the idea of open AI – for example, released the LLAMA* model*. Last year, Zuckerberg wrote that open technologies make AI safer, because different sides have the opportunity to check each other.
However, now he admits: as the models are becoming more powerful, they become more difficult to use for the majority. And the publication of too advanced solutions can only help competitors or harm. He emphasized that Meta* will still share part of its developments, but not everyone.
*belongs to Meta, which is recognized in the Russian Federation extremist and prohibited
