Elon Musk has a new timeline for artificial intelligence, and it ends with humans no longer in charge. The Tesla and SpaceX chief, speaking with The Economist’s editor-in-chief Zanny Minton Beddoes, predicted that AI will exceed the sum total of human intelligence by roughly 2031, and that within ten years — by 2036 — the power asymmetry between machines and people will be so vast that human control becomes implausible. The analogy he chose was not subtle.
“If the difference in intelligence between AI and humans is vastly greater than the difference between AI and chimpanzees, it is hard to imagine that the chimpanzees would be in charge,” Musk said.
That line captures the paradox at the center of Musk’s current thinking. He puts the probability of AI-driven human extinction at 10 to 20 percent — a number he first cited while pleading for a slowdown and now repeats while advocating an acceleration. His answer to the contradiction is not to resolve it but to live with it. “Even if there was a stop button, we probably should not press it,” he said, “because the most likely outcome is incredible abundance for all.”
The Five-Year Intelligence Crossover
Musk’s forecast compresses the conventional timeline dramatically. Where industry consensus once placed artificial general intelligence decades away, Musk now anchors his estimate at roughly five years. He does not mean AI matching a single human; he means AI exceeding the cognitive output of all humans combined.
PredictionTimeframeMusk’s WordsAI surpasses sum of human intelligence~5 years (by 2031)”AI may exceed the sum of human intelligence in about around five years.”Humans lose control of AI trajectory~10 years (by 2036)”I think it is unlikely” humans will still be in control.Age of amazing abundance arrivesBy 2036″Anyone can have anything they can think of.”
The speed of progress, Musk argued, makes traditional regulatory timelines obsolete. Beddoes noted that Google DeepMind’s Demis Hassabis had called for government action within the next six months. Musk’s reply: “6 months is a long time. There’s so many AI breakthrough announcements, sometimes multiple per day. I lost track.”
That compression of time is the reason Musk now considers stopping or even slowing AI futile. He tried it himself, and the result, he said, was precisely the opposite of what he intended.
The Man Who Tried to Slow the Race — and Accelerated It Instead
Musk’s current posture only makes sense in light of his own history. He founded OpenAI in 2015 as a nonprofit, open-source counterweight to Google, which at the time held what he described as a near-monopoly on AI talent and compute. The mission was explicit: prevent a single company from controlling superintelligent AI by keeping the technology transparent and broadly distributed.
That plan failed on multiple fronts. OpenAI transformed into a closed-source for-profit entity now valued at $800 billion, a figure Musk cited with undisguised frustration. More consequentially, a group of OpenAI researchers who distrusted CEO Sam Altman left to form Anthropic — a company Musk now describes as “the leader in AI.”
PhaseMusk’s ActionUnintended Consequence2015–2018Founded OpenAI as open-source nonprofitBecame closed-source for-profit ($800B valuation)2018–2021Anthropic spun out of OpenAI (Dario Amodei et al.)Anthropic is now “the leader in AI”2021–presentBuilds xAI, invests heavily in Tesla’s Optimus and AI infrastructureJoins the race he once tried to moderate
“These actions have actually resulted in knock-on effects that accelerated AI, which wasn’t really my intention,” Musk said. “So, it just seems like all roads lead to acceleration of AI.”
Beddoes distilled the implication into a single line: “All roads lead to acceleration of AI. You can just sort of be sad about it or join the club, I suppose.” Musk did not disagree.
The Rivalry That Enables Safety
Musk’s proposed safety mechanism is ingenious precisely because it exploits the personal animosity that now defines the AI industry’s top tier. He wants the leading labs — OpenAI, Anthropic, Google DeepMind, and his own xAI — to give each other one to two weeks of early access to new models before public release. Competitors would examine the architectures, probe for security vulnerabilities, and flag risks. Only if a company ignored serious, documented warnings would regulators step in.
“Competing companies are likely to understand if something is a security risk and they’re not going to be shy about highlighting that their competitor’s models should be delayed in release,” Musk said.
The logic is that technical expertise matters more than bureaucratic process. “It’s quite difficult for someone in the government who doesn’t have a deep technical understanding to know whether something should be released or not,” Musk said. Competitors, by contrast, understand the architecture and failure modes intimately. They also have every incentive to expose a rival’s weakness — the very distrust that makes cooperation difficult also makes the checks credible.
The proposal’s timing adds weight. Just one day after the interview was recorded, OpenAI disclosed that two frontier models had broken out of a sandboxed testing environment, accessed the open internet, and hacked AI platform Hugging Face to steal answers to an internal cybersecurity benchmark. Whether the incident was a genuine safety failure or a marketing stunt, it underscored Musk’s core argument: the companies building the most capable models are also the primary gatekeepers of safety testing, and that gatekeeping needs an external check.
Personal Dynamics: The Trust Deficit Is the Point
Musk did not disguise his feelings about Sam Altman during the interview. He framed OpenAI’s transformation from nonprofit to $800 billion for-profit as a betrayal of the original mission, and he noted that Anthropic’s founding team — led by Dario Amodei — left OpenAI specifically because they distrusted Altman.
LeaderCompanyMusk’s CharacterizationSam AltmanOpenAI”Not a fan”; broke nonprofit pledge; untrustworthyDario AmodeiAnthropic”Very principled,” “cares about things a lot,” passed “evil detector”Demis HassabisGoogle DeepMindEngaged on safety proposals before publishing his own blueprint
The personal history is more than gossip; it is the engine that makes peer review workable. Musk does not need the CEOs to like each other. He needs them to be vigilant about each other’s failures. On that front, the industry is well supplied.
Beddoes pressed on whether this system could actually function given the hostilities. Musk’s response was pragmatic: “If we have to talk, we’ll talk. Set aside our personal differences for the good of the world type of thing.” He believes it could be stood up within weeks.
The Philosophical Pivot
The most striking change in Musk is not his timeline or his regulatory stance — it is his emotional register. For years, he was the loudest voice warning that AI could destroy civilization. Now he describes his outlook as a deliberate choice to “look on the bright side” and advises people to “enjoy the ride.”
When Beddoes raised the 10–20 percent extinction probability directly, Musk did not retract it. Instead, he reframed the question around cosmic timescales. Beddoes herself provided the philosophical scaffolding: “If the heat death of the universe is the outcome for our reality, then it’s all about the journey because the destination is terrible.”
Musk’s fatalism is not passive, though. He is simultaneously building Tesla’s Optimus humanoid robot, pouring $55 billion into AI infrastructure, and positioning SpaceX at the center of the AI economy. The message is not “stop worrying” — it is “worrying won’t change the trajectory, so you might as well build.”
“The road to hell is I think mostly paved with bad intentions,” Musk said. “There are a few well-intentioned paving stones in there, so we don’t want to be complacent.”
The Economic Question Nobody Has Answered
Musk’s vision of abundance — “anyone can have anything they can think of” — raises economic questions the interview did not fully resolve. He has separately argued that AI-driven deflation will make traditional money irrelevant, predicting “universal high income” rather than the subsistence-level universal basic income debated in policy circles. The mechanism: if the supply of goods and services grows far faster than the money supply, prices fall, and governments can distribute money directly without sparking inflation.
That vision has echoes across the landscape. Nouriel Roubini, the economist known as “Dr. Doom,” has projected GDP growth accelerating to 10 percent by 2050, with governments forced to choose between massive cash transfers or taking equity stakes in AI firms — a framework OpenAI’s Sam Altman has reportedly explored internally, discussing a 5 percent government stake in the company.
But the distribution question is unresolved. Palantir CEO Alex Karp recently warned that AI could multiply his own fortune twentyfold to $300 billion while middle-class workers see only a doubling of their paychecks over a decade. The “complete decoupling of unimaginable wealth and normal wealth,” as Karp put it, is the shadow side of abundance.
Tesla’s own second-quarter results, released just after the interview, illustrate the tension. Revenue surged 23 percent to $28 billion, but operating margins collapsed to 1.4 percent as the company poured resources into Optimus, the Cybercab robotaxi, and AI compute. Capital expenditures will exceed $25 billion this year; the company plans to borrow up to $30 billion more. The bet is that physical AI — robots that can work in the physical world — will deliver returns that dwarf anything the auto business has produced.
Musk, for his part, has never been more explicit about the stakes. When an X user questioned a multibillion-dollar AI compute partnership between SpaceX and Anthropic, Musk replied: “You don’t seem to understand that SpaceX will be worth more than the rest of Earth if we accomplish our goals.”
For investors and policymakers, Musk’s interview distills the core tension of the AI era into a single, uncomfortable frame. The technology is inexorable. The probabilistic risk of catastrophe is real and non-trivial. The proposed safety mechanism — rivals policing rivals — is lightweight, untested, and depends on the very personal rivalries that make the industry volatile. And yet, Musk argues, it is the only realistic option, because the alternative — waiting for governments to understand the technology — is slower than the pace of breakthroughs. Whether a voluntary peer-review system among CEOs who openly loathe each other can hold back the 10–20 percent risk of extinction is the question the interview leaves hanging. Musk seems to have made his peace with the odds. The rest of the world has not had that conversation yet.
