Mark Zuckerberg’s Meta is spending billions on AI after its metaverse flop

Companies in the AI ​​breed run a new goal post: so-called superintelligence or a AI model that can exceed human intelligence.

The terminology here can become a little dizzying. Top -Ki companies had already tried to build what has long been called “artificial general intelligence” (AGI) or AI, of which they say they will be as smart as humans. Depending on who you are talking to, Superintelligence may be a striking marketing term that is supposed to pull even more risk capital – or it could be the obvious next step in developing AI models, some of which believe that already as intelligent as people.

META in particular has described his AI department about the goal of the superintelligence through its new “superintelligence laboratories”. Mark Zuckerberg has escalated the talent war of the industry in recent months. Accordingly, wage packages of up to $ 1 billion offer Recruit the Top -KI engineers from Silicon Valley.

Today, explained The co-moderator Sean Rameswaram recently spoke with Riley Griffin, Tech Reporter at Bloomberg, about why Meta is currently shooting so hard for top AI talents and what the company wants to build with these top technologists.

Below is an extract of your conversation, which was processed for length and clarity. There is much more in the full episode, so listen Today, explained Wherever you get podcasts, including Apple podcastsPresent PandoraAnd Spotify.

How new is the new Superintelligence laboratory from Meta? What is it and what are you trying to do with it?

The superintelligence laboratory is incredibly new, and the story of Metas Superintelligence laboratory is essentially a history of competition.

So you have Mark Zuckerberg, a known CEO, and last spring we learned through our reporting that he had started to feel quite acute that Meta fell back for AI in this everything consuming. Meta had just published the latest version of his large voice model – they call it Lama. This is a system that is supposed to compete with open AI from Chatgpt or Anthropics Claude. But when the rollout landed, it fell flat. And after our sources it fell internally. It had just become clear that Meta did not lead the pack among these AI companies.

The listeners may consider META as the parent company of Facebook, Instagram or WhatsApp, these popular apps that have billions of users. But Meta also sees itself as a AI company. And it didn't lead there. And so Zuckerberg, who saw that he fell back in this race, immediately took action. He began personally from his houses in Tahoe and Silicon Valley, and he built this new secret team, of which we now know that they are meta superintelligence labs.

What will this super mysterious team do? I thought Meta was just about that “Metaverse”, “ And that's why they have changed their name in Meta. It sounds like they should have changed their name in AI.

It's a really good point. When I talk to people on Wall Street, the metaverse is always the pain point. It is something that still bleeds cash. It is not a big revenue driver. It has not proven to be the dream that Mark Zuckerberg once painted.

The AI ​​ambitions of Meta are a bit of another animal. In the days when it was still called Facebook, they were in the AI ​​game. They brought up this guy named Yann Lecun, a large, award -winning thinker [AI]. If you watched this at the time, you may have thought that Meta was really well positioned to be the AI ​​leader.

But when Openai came out with Chatgpt, it all got out of the water, including AI researchers at Meta that I spoke to. They had built up their own great models, but the approach was still rather academic. It was not packed in a consumer product in the way it was successfully executed on Chatgpt. I think the race began, but this April really has what it feels like for many at Meta as a flop with its last publication of Lama.

And what exactly has Meta done in this room? Nothing, it sounds like?

At the moment we see a talent war. We see that Meta creates a new organizational structure for its AI talent. This is a loss of billions of effort, and Meta has the money for the provision.

The perspective of the company management is that you make a big bet and want to close this talent gap. They believe that there is only a small talent pool that can best use these expenses to put together these competitive models. Many focus on argument, and that means that they are specialists who can help create models that think step by step, in contrast to this probabilistic approach, in which they predict the next word in the sentence.

Many chatbots at the moment they look as if they understand what they say when they enter a question, but they really only forecast. You have taken these massive data and can predict the next word in the sentence. And what Meta, what Openai wants to do is the ability to think for yourself. So if you spend hundreds of billions of dollars to win the AI ​​race, what are a few billion for the talent that the ship can control? That is the reason from the company.

But of course there is pushback. One thing that I have heard of people who have left meta over the years is that if they are a doctoral student who has entered this really complicated question of AI argument and superintelligence, they were not necessarily inspired as Meta that they created Ki chatbots that sound like snoop dogg. So there was skepticism towards Mark Zuckerberg as a guide. He is someone who really wants to put AI into the hands of people.

I think it's too early to say whether it works or not. What was successful are Mark's efforts to vote for some of the biggest names out there. You take top talents away from competitors. But we also know that there are people who rejected Mark Zuckerberg's compensation packages at the sports team level.

I will say that pushback doesn't come from Wall Street. The Meta share has increased by almost 30 percent this year. It is clearly excitement of the investor community.

So it sounds like Meta is being restructured or completely new structures to play the long game for superintelligence. What about the rest of the industry?

Everyone – I mean everyone – had to respond to Mark Zuckerberg's crazy settings. His efforts have set a new bar for compensation in the entire industry. We have seen that other companies like Openaai have to make offers that are much higher.

You can even see this in the language, the public attitude of CEOs. Openai said that they will output trillions for the AI ​​race. Other companies are quite loud about the money they will throw. But we talk about the biggest players here – we don't talk about the Umstarts who have just received a venture money and have no income. Meta, Google, even Openai are in a different position because they have money to burn.