Musk discusses his new company xAI, 'superintelligence,' and China

In an hour-long treatment connected Twitter Spaces connected Wednesday night, Twitter proprietor and Tesla and SpaceX CEO Elon Musk reflected connected the information down his newest artificial quality venture, China's narration with the U.S. and the likelihood of AI creating a dismal aboriginal for humanity.

Musk participated successful the treatment with 2 cardinal members of Congress who beryllium connected the House Armed Services Committee and the Select Committee connected the Chinese Communist Party: Reps. Ro Khanna, D-Calif., and Mike Gallagher, R-Wisc. The lawsuit was billed arsenic a speech astir the aboriginal of AI and came connected a time that Musk launched his ain caller AI company, xAI.

Toward the opening of the conversation, Musk explained his extremity of creating xAI, describing it arsenic a benignant of if you can't bushed 'em, articulation 'em situation.

"If I could property intermission connected AI oregon truly precocious AI, integer superintelligence, I would," said Musk, who co-founded 1 of the starring AI companies OpenAI, but left its committee successful 2018 and nary longer holds a involvement successful the firm. "It doesn't look similar that is realistic."

As a result, helium decided to "grow an AI successful a bully way." To Musk, this means grooming it successful specified a mode arsenic to beryllium "maximally curious, maximally truth-seeking."

"I deliberation a maximally funny AI, 1 that is conscionable trying to benignant of recognize the beingness is, I think, going to beryllium pro-humanity," Musk said. "From the standpoint that humanity is conscionable overmuch much absorbing than not humanity."

Musk said helium prefers this method to "trying to explicitly programme morality into AI." He said that attack requires judgement calls connected whose morality is being programmed into the strategy and besides creates an "inversion problem" wherever the morality tin beryllium flipped connected its head.

As acold arsenic regulating the exertion goes, Musk suggested that policymakers walk a fewer years learning astir the exertion earlier moving to "oversight, with consultation with industry." He besides suggested that immoderate benignant of manufacture radical oregon self-regulatory assemblage akin to the Motion Picture Association would beryllium useful.

The 3 spent a important portion of the treatment focused connected the menace of China getting up of the U.S. connected AI and the imaginable for immoderate benignant of subject confrontation implicit the sovereignty of Taiwan, a self-ruled portion that the Chinese authorities considers portion of its territory.

Musk said that successful his conversations with elder enactment successful China connected a caller travel there, helium spent a bully magnitude of clip discussing AI safety. He said the thought that a "digital superintelligence" could supplant the Chinese Communist Party itself seemed to resonate.

"No authorities wants to find itself unseated by a integer superintelligence," helium said. "So I deliberation they really are taking enactment connected the regulatory beforehand and are acrophobic astir this arsenic a risk."

Musk adjacent said helium believes the Chinese authorities would beryllium unfastened to collaborating connected an planetary model astir AI regulation.

Gallagher, who chairs the prime committee connected China, pushed backmost connected the thought that the CCP could beryllium a constructive subordinate of specified an planetary framework. He warned that adjacent if they took Musk's warnings to heart, helium fears it wouldn't beryllium capable to dilatory their AI efforts and that China's person Xi Jinping would usage it to cement "totalitarian control."

After concisely changing the topic, Musk returned to Gallagher's thoughts and called himself "pro-China." Musk acknowledged helium has "some vested involvement successful China" but yet believes "China is underrated" and that "the radical of China are truly awesome." Gallagher aboriginal said helium "fully supports the Chinese people," but it's the ruling enactment helium takes contented with.

"That's not to accidental that determination aren't immoderate precise important disagreements and that there's evidently going to beryllium a important situation connected the Taiwan question," Musk said, referring to China's stated tendency to bring Taiwan backmost nether its control. "I deliberation ultimately, erstwhile the precise hard question of Taiwan is resolved, I americium surely hopeful that determination volition beryllium affirmative relations betwixt China and the United States and the remainder of the world."

While tensions betwixt the U.S. and China stay high, the 2 powers person made immoderate diplomatic headway by resuming in-person communications, with Treasury Secretary Janet Yellen precocious making a trip determination and U.S. clime envoy John Kerry acceptable to bash so aboriginal this month.

Still, Musk predicted that resolving the Taiwan question volition beryllium difficult.

"I bash person this mentation astir prediction, which is that the astir entertaining result — arsenic seen by a 3rd party, not the participants — is the astir likely," Musk said. "Which does suggest it's astir apt going to get blistery successful the Pacific. Hopefully not excessively hot. But it's going to get hot."

Musk predicted the satellite volition spot integer superintelligence, oregon a signifier of artificial quality that is "smarter than immoderate quality astatine anything" wrong 5 oregon six years.

"That's not needfully smarter than the sum of each humans," Musk said. "That's a higher barroom to beryllium smarter than respective humans. And particularly fixed that it's the sum of each humans that are machine-augmented, successful that we volition person computers and phones and bundle applications. We already are de facto cyborgs, it's conscionable that the computers are not yet integrated with us."

While helium has focused connected the existential risks of AI, antecedently signing a letter for a intermission connected precocious AI improvement and calling it 1 of the "biggest risks" to civilization, Musk said he's yet an optimist astir the technology.

"If I were to delegate probabilities, I deliberation it is much apt to beryllium a affirmative script than a atrocious scenario," helium said. "It's conscionable that the atrocious script is not 0% and we privation to bash everything we tin to minimize the probability of a atrocious result with AI."

Musk astatine 1 constituent described the aboriginal arsenic if it were a TV series.

"If this was a Netflix bid oregon something, I'd accidental the play finale would beryllium a showdown betwixt the West and China," helium said, "and that the bid finale volition beryllium AGI," oregon artificial wide intelligence.

Subscribe to CNBC connected YouTube.

WATCH: Can China's ChatGPT clones springiness it an borderline implicit the U.S. successful an A.I. arms race?