“I’m going to start something which I call TruthGPT,” Musk said, “or a maximum truth-seeking AI that tries to understand the nature of the universe.”
Musk previously tweeted, “What we need is TruthGPT,” in February.
The interview, which was recorded earlier this month, potentially sheds light on the direction of a new start-up dubbed X.AI that Musk incorporated in Nevada last month, according to documents filed with the state. Musk acknowledged in an interview with the BBC last week that Twitter is accumulating sophisticated computer equipment, with the goal of pursuing generative AI.
Musk would be entering an increasingly crowded field and a fast-escalating AI race, where OpenAI scored an early lead by launching a test version of ChatGPT late last year, followed by an investment from Microsoft in January and integration with its Bing search engine in February.
Last month, Google made its chatbot, Bard, available — something expected to eventually be paired with the company’s search engine.
In a “60 Minutes” interview that aired Sunday, Alphabet CEO Sundar Pichai expressed concern about the rapid pace of AI’s evolution, but pointed to some hope.
“Compared to any other technology, I’ve seen more people worried about it earlier in its life cycle,” he said. “Hence, the conversations are starting in a serious way.”
The AI-powered chatbots draw on vast troves of information and knowledge of speech patterns to generate conversational answers to questions, acting as a sort of know-it-all personal assistant, though experts have warned about their potential unreliability and their tendency to engage in what are termed “hallucinations,” or nonsense answers.
Some were quick to point out the irony of Musk pursuing generative A.I. after signing onto a letter last month calling for a pause on its development, alongside other business leaders and academics. He has also broken away from his involvement with OpenAI and has been publicly critical of current CEO Sam Altman.
“AI is more dangerous than say mismanaged aircraft design or production maintenance or bad car production,” Musk said in the Fox News interview. “It has the potential of civilizational destruction.”
OpenAI did not immediately respond to a request for comment.
Musk told Carlson he was deeply concerned about the potential dangers and believed a regulatory agency should be put in place with the ability to gather insights, solicit industry opinion and propose rules.
If Musk pursues his AI bid, he’ll likely add to growing concerns about his effectiveness as the CEO of Twitter, Tesla, SpaceX and two smaller companies and concern from some investors that he already has too much on his plate.
Musk said last week that leading Twitter has been painful. He has so far declined to hand over the reins, despite a pledge to step down as CEO. (Musk said in last week’s interview that he had handed over the position — to his dog Floki.) Tesla, meanwhile, has shed hundreds of billions of dollars in value since Musk began pursuing Twitter, which he’s said itself is worth less than half of the $44 billion he spent to acquire it.
This week SpaceX is attempting to launch Starship, its nearly 400-foot-tall reusable rocket — in a test of a vehicle that officials hope will one day put astronauts back on the moon, something Musk has been touting in tweets.
Musk’s sudden pivot to pursuing generative AI follows a pattern for an excitable CEO, who is sometimes susceptible to the latest trend in tech. Musk began pursuing self-driving cars at Tesla after other tech giants such as Google had established projects in the field. Musk’s tunneling start-up, the Boring Company, and brain implant company, Neuralink, have yet to deliver on their ambitious promises.
Musk, whose electric car company Tesla is seeking to build a “friendly” humanoid robot, struck a familiar tone as he discussed the purpose of his latest pursuit. He has regularly warned about the potential dangers of uncontrolled artificial intelligence to humanity more broadly, citing it as one of the biggest risks to civilization.
His AI would be different, however, he said.
“I think that 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,” he said.
