Elon Musk says he’s building his own rival to OpenAI’s ChatGPT in the form of “TruthGPT,” which will be designed to counter the “political correctness” from other chatbot programs.
“What’s happening is that they [OpenAI] are training their AI to lie,” Musk claimed in an interview with Fox News pundit Tucker Carlson.
Musk is no fan of ChatGPT. In February, he accused it of spouting propaganda after the AI program produced incorrect responses or refused to comment on politically charged topics, such as former President Donald Trump.
“A path to AI dystopia is to train an AI to be deceptive,” Musk told Carlson. “So yeah, I’m going to start something called TruthGPT. Or a maximum truth-seeking AI that tries to understand the nature of the universe.”
Musk acknowedged that he's "definitely starting late" on the AI front, "but I will try to create a third option [that] hopefully does more good than harm.”
Musk has reportedly purchased 10,000 enterprise-grade GPUs to help train his own AI programs. At the same time, he's among the entrepreneurs and computer scientists who recently signed a letter demanding the tech industry pause development on AI programs for six months, citing the technology’s potential to disrupt society.
In the interview with Carlson, Musk said he’s concerned AI is dangerous because it holds the potential to outsmart human intelligence. This could lead to unpredictable effects, for example, creating an AI “super god” that could secretly take over or influence society. Hence, he says the technology needs to be regulated to ensure AI programs can benefit humanity.
“I think there should be some government oversight because it’s a danger to the public,” he said, later adding: AI "has the potential of civilizational destruction.”
Musk vs. OpenAI
Musk took plenty of shots at OpenAI, a company he helped found in 2015 before resigning from its board in 2018.
“The intention of OpenAI was obviously to do good, but it’s not clear if it’s actually doing good. Except that I’m worried about the fact that it’s being trained to be politically correct, which is simply another way of being untruthful, saying untruthful things. That’s a bad sign.”
In a tweet after the interview aired, Musk also signaled he might sue OpenAI.
Musk claimed OpenAI started with Google co-founder Larry Page, whom Musk would frequently talk to about AI safety. “My perception was that Larry was not taking AI safety seriously enough,” Musk claimed. “He really seemed to [want] a digital super intelligence. Basically, digital god, if you will, as soon as possible. And I agree with him that there’s great potential for good, but there’s also potential for bad."
But in a bit of irony, Musk’s solution to counter both Google and OpenAI is to create an AI program that can freely research and comment on any topic. “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. Hopefully, they would think that,” Musk said.
OpenAI didn’t immediately respond to a request for comment.