Mark Zuckerberg Is Working Toward AI Superintelligence. Does Anyone Actually Want It?

Here’s what you’ll learn when you read this story:

• In a new post, Meta CEO Mark Zuckerberg said that the company’s AI is on the slow road to superintelligence, a technological threshold beyond artificial general intelligence (AGI) when AI can begin improving itself.
• In an effort to sell a future of his own design, Zuckerberg waxes poetic about the technology’s possible benefits, and even floats the idea of AI-powered smart glasses as a not-at-all creepy technological layer between you and the real world.
• While there’s no peer-reviewed research to confirm Zuckerberg's claims, some AI researchers believe that superintelligence could arise only a few years after the arrival of artificial general intelligence.

Although we’re only a few years into a brave new AI era, artificial intelligence continues to wreak unrelenting economic and environmental havoc. A new report, published in July, shows that AI has wiped out at least 15 percent of entry-level jobs while another recognized the technology as one of the top five reasons for job losses. Meanwhile, AI data centers are draining water resources from local communities, and according to Bloomberg, two-thirds of data centers built since 2022 have been in high water-stress areas.

But according to Meta CEO Mark Zuckerberg, that all should take second place to excitement about AI superintelligence, about which he is “extremely optimistic.”

“Over the last few months we have begun to see glimpses of our AI systems improving themselves,” Zuckerberg wrote in a post published on July 30th. “The improvement is slow for now, but undeniable. Developing superintelligence is now in sight.”

Unlike the large language models (LLMs) of today—such as ChatGPT or Claude—superintelligence is a technological threshold at which AI models can improve upon themselves, theoretically ushering in a new era of AI sophistication. AI experts see superintelligence as an evolution even beyond what’s long been considered the “holy grail” of AI research: artificial general intelligence, or AGI, which is the moment our algorithmic creations surpass human intelligence.

Because no peer-reviewed papers were published detailing Meta’s “superintelligence” breakthrough, it’s impossible to know the validity of Zuckerberg’s claims, but recent studies (unaffiliated with Meta) have shown ways in which AI can improve upon itself. A paper uploaded to the preprint server arXiv in October of 2024, for example, details the inner workings of a theoretical device known as Gödel’s Agent, which implements self-recursive improvements when a formal proof asserts an improvement’s potential benefit. Right now, some experts believe that superintelligence could likely emerge in only a few years after AGI is achieved (thought that itself is far from certain).

Of course, not all AI is necessarily bad. In the hands of scientists, AI can be a valuable tool in the fight against cancer or in divining the long-lingering secrets of the quantum and cosmological realms. But for most tech CEOs, AI is just another way to make money for shareholders, which in Meta’s case, basically amounts to new ways to sell more ads.

In a recent interview in May, Zuckerberg summed up his hopes for AI as a means to improve the company’s ad business while keeping people scrolling through Instagram, according to Vox. He also talked about people having AI therapists and combatting the loneliness epidemic with AI friends, which left some AI experts calling the vision “dystopian” among other things:

“It was like if you took a left wing caricature of why Zuckerberg is evil, combined it with a left wing caricature about why AI is evil, and then fused them into their final form,” AI writer Zvi Mowshowitz wrote back in May. “Except it’s coming directly from Zuckerberg, as explicit text, on purpose.”

In Zuckerberg’s latest post, he’s definitely toned down some of the more unsavory aspects of his future AI plan, and instead focuses on how personal superintelligence will help “you achieve your goals, create what you want to see in the world, experience any adventure, be a better friend to those you care about, and grow to become the person you aspire to be.” This immensely vague description speaks to the difficulties in selling a future few people seem to want. These platforms come with the promise of immense disruption, so couching the idea in empty niceties about how great it is going to be is just a classic sales tactic.

Of course, any AI vision would be incomplete if we didn’t try to stick some shiny new tech into some sunglasses... again. Yes, Zuckerberg also envisions smart glasses that represent a physical AI barrier between you and the real world. Vox describes this vision as a “Meta-governed layer”—a natural extension of what Zuckerberg tried to do with the Metaverse, the company’s failed attempt at an alternate virtual reality.

So this all begs the question: Should we really trust the foundations of our dystopian AI reality to a company that once lived by the motto “move fast and break things?” If we leave the future of AI exclusively in the hands of Silicon Valley billionaires, that famous phrase may one day become a fitting societal epitaph.