And An Open Source AI Giant Rises: Zuckerberg Targets AGI

In a surprise announcement that could change the pecking order in the global AI race, Meta CEO Mark Zuckerberg said he is setting his sights on building AI models that could reach artificial general intelligence or AGI.

In doing so, Meta is building what could be the open source community’s most formidable challenger to OpenAI, whose most powerful language models are proprietary and closed.

“It has become clear that the next generation of services requires building full general intelligence − building the best AI assistants, AIs for creators, AIs for businesses and more,” Zuckerberg said in a Facebook post.

“That needs advances in every area of AI from reasoning to planning to coding to memory and other cognitive abilities. This technology is so important and the opportunities are so great that we should open source and make it as widely available as we responsibly can.”

The idea of general intelligence, broadly defined as AIs being able to perform general tasks on par with or better than humans, is a goal long-sought-after by many in AI, including OpenAI and Google DeepMind.

But the definition of AGI is not set. “I don’t have a one-sentence, pithy definition,” Zuckerberg told The Verge. “You can quibble about if general intelligence is akin to human level intelligence, or is it like human-plus, or is it some far-future super intelligence. But to me, the important part is actually the breadth of it, which is that intelligence has all these different capabilities where you have to be able to reason and have intuition.”

Related:Meta Unveils New Image-Generation Model, AI Assistant as AI Race Heats Up

Aiming for AGI also is a carrot for top-notch talent as “a lot of the best researchers want to work on the more ambitious problems,” he said.

With this AGI focus, Meta is now doubling down on its already impressive AI work: LLaMA and the subsequent Llama 2 family of large language models have formed the basis of many open source AI models.

Zuckerberg said Meta is working on Llama 3, its next-generation large language model.

But if Meta’s models ever achieve AGI, Zuckerberg is not ready to commit to it being open source. “For as long as it makes sense and is the safe and responsible thing to do, then I think we will generally want to lean towards open source,” he told The Verge. “Obviously, you don’t want to be locked into doing something because you said you would.”

FAIR moves and loads of H100s

As part of its new focus, the company is consolidating its FAIR (Fundamental AI Research) lab with the AI product division, called GenAI. The move will see Joelle Pineau, who leads FAIR, and Yann LeCun, Meta’s chief AI scientist, report to Chris Cox, Meta's chief product officer.

Related:Meta to Infuse Generative AI into All Its Products

Zuckerberg also announced Meta is building a “massive compute infrastructure” to support its AI efforts, with plans to snap up 350,000 H100 chips from Nvidia by the end of 2024. That would bring Meta’s total computing power to almost the equivalent of 600,000 H100s, after counting other GPUs.

Sister research firm Omdia expects the company to continue being a top buyer of Nvidia’s AI chips. In 2023, Meta is estimated to have ordered at least 150,000 H100 chips to power its AI products.

Zuckerberg skeptical of AI rivals’ true motives

As for Meta's rivals in AI, Zuckerberg told The Verge that “there were all these companies that used to be open, used to publish all their work, and used to talk about how they were going to open source all their work. I think you see the dynamic of people just realizing, ‘Hey, this is going to be a really valuable thing, let’s not share it.’”

OpenAI was founded as a nonprofit, but as it discovered that it could not raise the amount of funding it needed for AI compute, it created a ‘capped-profit’ company under the nonprofit parent.

Zuckerberg thinks it is convenient for the biggest companies leading in AI to also be the ones calling for guardrails on the technology. OpenAI, Google, Microsoft, Anthropic and others have met with world leaders, lawmakers and regulators to ask them to craft rules to ensure AI safety.

“The biggest companies that started off with the biggest leads are also, in a lot of cases, the ones calling the most for saying you need to put in place all these guardrails on how everyone else builds AI,” he tells The Verge. “I’m sure some of them are legitimately concerned about safety, but it is a (heck) of a thing how much it lines up with the (business) strategy.”

Has Meta ditched the metaverse?

Meta’s entire company-wide focus has been on developing the metaverse – an interactive, virtual environment for users – so much so that it changed its company name from Facebook in 2021.

However, Meta focused heavily on AI at its Connect showcase last September. Now that the company is doubling down on AI, one question that arises is whether it is leaving the metaverse, at least for now, to the wayside.

But that is not the case – Zuckerberg touched on bringing together AI and the metaverse on his Facebook post: “People are also going to need new devices for AI and this brings together AI and the metaverse because over time, I think a lot of us are going to talk to AIs frequently throughout the day.”

Zuckerberg said that smart glasses are “the ideal form factor for letting an AI see what you see and hear what you hear.”

Meta teamed up with Ray-Ban to create Smart Glasses, which look like sunglasses but have embedded tech including a camera and Meta’s AI assistant. AI Business got to see them up close at the recent CES 2024.

The CEO said the glasses “are off to a very strong start.”

LeCun added that all future interactions in the digital world “will be mediated by AI assistants through our smart glasses and other devices.”

“We need these systems to have human-level intelligence if they are to understand the world, people, and tools, so as to help us in our daily lives,” the Turing Award winner said on X (Twitter).