TruthGPT: yet another approach to artificial intelligence

In late March 2023, Elon Musk co-signed a letter calling for OpenAI to pause training AI systems. In the same month, it was revealed that Musk set up his own artificial intelligence company: X.AI Corp. The filing, first reported by the Wall Street Journal, lists Musk as the sole director of the company, and the director of his family office, Jared Birchall, as secretary.

The privately held company has permitted the sale of 100 million shares for the newly created firm, which might come as a surprise given that Musk has implied that OpenAI’s profit incentives might interfere with the ethics of the models it creates.

What we need is TruthGPT

— Elon Musk (@elonmusk) February 17, 2023

A tweet posted in February seems to have been well-received enough to give Musk’s ChatGPT alternative its name: TruthGPT. Although likely a response to the controversy of advanced chatbots propensity to lie, or at least their ability to regurgitate false information, the name now signifies a more transparent approach to the development of AI models — despite the secretive filing of X.AI.

Musk’s AI won’t seek to destroy humanity

TruthGPT is being framed as a course correction to OpenAI’s efforts, which Musk believes would have led to the destruction of humanity. “I’m going to start something which I call TruthGPT or a maximum truth-seeking AI that tries to understand the nature of the universe,” Musk said.

“I think this might be the best path to safety in the sense that an AI that cares about understanding the universe is unlikely to annihilate humans because we are an interesting part of the universe.”

Discussions have been held with Tesla and SpaceX investors about helping to fund the project. Perhaps some of the funding could go towards explaining to Musk that, threatening to humanity or not, generative AI doesn’t form opinions, doesn’t care about the universe unless you ask it to, and won’t call him papa and protect him from scary AI robots.

In an interview with Fox News, Musk’s vision for the threat that AI poses includes the fact that it “potentially manipulates public opinion in a way that is very bad.”

If regulations are only enforced in the AI race after something goes wrong, it may already be too late. “AI might be in control at that point.” In fact, it’s “absolutely” conceivable that AI would reach a point where it’s in control and couldn’t be turned off; it would be making decisions for people.

Ironically, Musk compares the way that his TruthGPT would protect humanity to the way that humans have protected chimpanzees. Even though we could have killed them off, there’s a fondness for the animals that means they’re protected (or experimented on at Neuralink).

Despite his apparent reservations, it looks like Elon Musk is joining the AI race in competition with the organization he helped start. Cynically, his condemnation of their practices might only have been the starting moves in establishing his own, separate efforts, although it’s debatable whether he was thinking that far ahead.

Watch the interview with Tucker Carlson here.