Meta boss Mark Zuckerberg wants to develop a “superintelligence”.

Enlarge image

Facebook founder Mark Zuckerberg at a product presentation in September 2023

Photo:

Andrei Sokolow / dpa

The three letters AGI electrify Silicon Valley.

The abbreviation stands for “Artificial General Intelligence”, which literally means general artificial intelligence.

There are varying definitions of the term in detail, a typical one being: a program that has the ability to understand or learn any intellectual task that a human can perform.

Meta's competitor OpenAI had set the goal from the beginning of wanting to achieve this form of AI.

“We were the first and probably the only ones to ever say that we were pursuing AGI,” OpenAI CEO Sam Altman told SPIEGEL last May.

Now Facebook founder Mark Zuckerberg is joining the chorus: “Our long-term vision is to develop general artificial intelligence,” he wrote in a post on the Threads platform, which is part of his Meta Group.

Zuckerberg, like other AGI advocates, was unable to clearly define what exactly this AGI would look like and only provided a rather cloudy explanation: “One can argue about whether it corresponds to human intelligence or whether it is a superintelligence in the distance future or a kind of human-plus,” he said in an interview with “The Verge.”

For him, the breadth of skills is crucial

everyone “for whom you have to be able to think and have intuition,” said Zuckerberg.

AGI as bait for talented developers

Zuckerberg also did not want to commit to a clear, measurable point at which AGI would be reached.

Because of the combination of great promise for the future and unclear definition, experts regularly point out that AGI is not just a technical term, but also marketing for companies.

In any case, Zuckerberg also has a concrete motivation for propagating AGI: the battle for talent raging in Silicon Valley.

It is important to convey that Meta is working towards AGI "because many of the best researchers want to work on the more challenging problems," he said in the "Verge" interview.

So that Meta can better implement its vision, Zuckerberg is restructuring the company internally.

The internal AI research laboratory FAIR is to be connected to the department that works on generative AI applications for the various meta-apps such as Instagram, WhatsApp and Facebook.

Indirect criticism of OpenAI

Without naming names, Zuckerberg also indirectly criticized the strategy of ChatGPT developer and market leader OpenAI.

He originally started with the promise to publish his AI models open source, i.e. to make them available to all developers in an open-source and transparent manner.

OpenAI has now moved away from this.

Zuckerberg now said: "I think you see the dynamic that people just realize: 'Hey, this is going to be a really valuable thing, let's not share it'."

Meta has so far taken a different approach and has made some AI models available open source.

The question of whether AI should be open source or not is controversial in Silicon Valley.

Critics fear that dangerous misuse of the technology could become easier.

Proponents argue that the public benefits when more people can contribute to progress.

hp