Mark Zuckerberg dreams of a world where everyone has access to ‘superintelligence.’ In that dream, the Meta CEO seems to be moving away from his belief in open-source AI.
Meta is heavily investing in AI. In addition to billions in chips and infrastructure, it is also strengthening its internal AI teams. Recently, Meta acquired Scale AI for 14.5 billion dollars. These investments aren’t for nothing: Meta is quietly building towards “superintelligence,” as Mark Zuckerberg reveals in a remarkable letter.
In Zuckerberg’s future vision, everyone will soon have access to “personal superintelligence.” This personal AI assistant will improve your life, although the Meta boss does not describe how. Meanwhile, Zuckerberg’s Metaverse dream has died a quiet death, so it remains to be seen how his AI plans will fare.
A Bit Open, Mostly Closed
Zuckerberg is already looking ahead to “new security concerns” that superintelligence might bring. Therefore, Meta will have to “carefully choose what it opens up.” This stance is contrary to Meta’s current approach, which makes its Llama models available as open-source.
read also
Meta Disagrees with AI Act Code of Conduct: “Europe is Heading in the Wrong Direction”
Meta and Zuckerberg have always positioned themselves as evangelists for open-source AI and have received praise from the open-source community for it. The company hitched its wagon to Red Hat to promote open-source as the way forward for AI adoption. That Meta is now citing security as an argument to pull AI out of the open sphere is, to say the least, striking.
During a Q with investors, Zuckerberg says his stance hasn’t made a U-turn and that even now “not everything Meta does is open.” He promises that Meta will continue to focus on open-source technology. Yet, it seems to be a trend that the more powerful AI models become, the more they disappear behind closed doors.