2023 was marked by artificial intelligence. And 2024 is going to be the same or worse. As we saw in the Davos Forum yesterday, AI is what is driving all current conversations.
Big tech companies have gone from wanting to create the metaverse to betting their entire business plan on artificial intelligence. Because sometimes, businesses are a matter of trends. And an example of this is Facebook, which changed its name to Meta for a total bet that has ended up in nothing.
As you well know, OpenAI’s mission is to create this artificial general intelligence or AGI. And Demis Hassabis, the leader of Google’s AI efforts, has the same goal.
Now Facebook (Meta) goes for AGI
Now, Mark Zuckerberg, CEO of Meta, enters the race. Although he doesn’t know when AGI will be achieved, not even an exact definition, he wants to build it. At the same time, he is shaking things up by moving Meta’s AI research group, FAIR, to the same part of the company as the team creating generative AI products in Meta’s applications.
The goal is for Meta’s advances in AI to reach its billions of users more directly: “We have come to the conclusion that, in order to build the products we want to build, we have to build for general intelligence. I think it’s important to convey this because many of the best researchers want to work on the most ambitious problems,” Mark said.
Here, Zuckerberg is saying the quiet part out loud. The battle for AI talent has never been so fierce, with all companies in the sector competing for an extremely small group of researchers and engineers.
Those who have the necessary knowledge can earn salaries that exceed one million dollars per year. Executive directors like Zuckerberg are forced to try to win over a key recruit or prevent a researcher from going to the competition.
“We are used to very intense talent wars,” he says. “But here there is a different dynamic, with several companies looking for the same profile and many venture capitalists and people investing money in different projects, which makes it easier for people to start different things outside.”
After talent, the scarcest resource in the field of AI is the computing power needed to train and run large models. In this regard, Zuckerberg is willing to show off his muscles. According to The Verge, by the end of this year, Meta will own over 340,000 Nvidia H100 GPUs, the industry’s preferred chip for creating generative AI.
But, what is AGI?
No one working in AI, including Zuckerberg, seems to have a clear definition of AGI or an idea of when it will arrive.
“I don’t have a concise definition in a single sentence,” he tells me. “It can be argued whether general intelligence is similar to human intelligence, if it is more human or if it is a superintelligence of the distant future. But for me, the important thing is its breadth, that is, that intelligence has all those different capacities that allow you to reason and have intuition.”
Consider that their arrival will be a gradual process, rather than a single moment. “I’m actually not so sure that any specific threshold feels so profound.”
According to Zuckerberg, Meta’s new and broader focus on AGI was influenced by the release of Llama 2, their latest large language model, last year.
The company did not think that the ability to generate code made sense for how people would use an LLM in Meta’s applications. But it is still an important skill to develop in order to build a smarter AI, so Meta built it anyway.
Meta is already training Llama 3, which will have code generation capabilities, it says. Like Google’s new Gemini model, it also focuses on more advanced reasoning and planning capabilities.
“Llama 2 was not a leading model in the industry, but it was the best open-source model,” he says. “With Llama 3 and beyond, our ambition is to build things that are state-of-the-art and eventually become industry-leading models.”
