Meta is all in on open source A(G)I

Mark Zuckerberg betting big on AGI's potential and isn't shy about it. Key players like OpenAI, Google's Demis Hassabis, and now Meta's Mark Zuckerberg are all gunning for artificial general intelligence (AGI)—the next pinnacle of AI.

What’s going on here?

Mark Zuckerberg’s new goal is to create artificial general intelligence and maybe open-source it.

What does this mean?

Zuckerberg has thrown Meta's hat into the ring for developing AGI, merging Meta’s AI research (FAIR) and generative AI product development teams. Meta's serious about this, dedicating immense resources like a massive stockpile of Nvidia’s GPUs—a whopping 340,000 H100s by year's end and in total 600,000 equivalent GPUs.

But here's the twist: there's no clear AGI definition or timeline, not even from Zuckerberg. Yet, he's all about the breadth of intelligence, from reasoning to intuition. Meta's working on Llama 3, a large language model, beefing up its code-generating and reasoning capabilities. And, Zuckerberg is championing an open-source approach to AGI, contrasting it with companies turning more secretive.

Why should you care?

The restructuring of Meta to fast-track AI into their products, aiming to reach billions of users seems similar to Google Brain and Deepming merger. But Zuckerberg isn’t thinking about saving revenue from search. For him, AI (or AGI) is a part of the “Metaverse” plan.

Llama 3 will again be open source, but AGI might not be—Zuckerberg doesn’t want to commit just yet but wants to lean towards open source. Fair I’d say.