OpenAI chief Sam Altman: ‘This is genius-level intelligence’
The tech entrepreneur on the risks and opportunities of AI, his dispute with Elon Musk and why he has the ‘most important job maybe in history’
Sam Altman’s escape from the Silicon Valley pack is a sprawling farm at the end of a road that snakes through the vine-swept hills in Napa Valley. I spot the 40-year-old with slightly tousled hair in the open-plan kitchen of the wide bay-windowed house, and I step right in. His bemused look tells me that I wasn’t expected. I am, it turns out, nearly an hour early, but the man behind ChatGPT will finish up a meeting and join me in the garden. I wait under a grapevine-shaded pergola that runs along the house.
For a Lunch with the FT, Altman offered to cook a simple vegetarian meal at his farm instead of meeting me at a restaurant of his choice, where he’s likely to be hounded by selfie-seekers. Since OpenAI, the company he runs, released the generative AI model in 2022, Altman has been catapulted to the status of worldwide celebrity. Last year, he married his software engineer boyfriend and they’ve recently had a baby boy via surrogacy (he consulted ChatGPT on which crib to buy) so he’s been spending more time on the Napa farm.
Altman takes only a few minutes to join me outside. He has built OpenAI into one of the fastest-growing companies ever, with a staggering valuation of more than $250bn, and accelerated a fierce race for AI supremacy: the pursuit of artificial general intelligence, when a machine can surpass the cognitive abilities of humans, not only absorbing knowledge but reasoning and learning on its own.
But it has been a rollercoaster journey in which Altman has been fired and rehired by his own company, his character and commitment to the safe development of AI subjected to bruising questions. He’s wrestled with Elon Musk, with whom he co-founded OpenAI, and sparred with Scarlett Johansson, who accused him of using an “eerily similar” voice to her own to train a chatbot. Having stolen a march on more established competitors (Google in particular, given that it has long had the lead in AI research), he’s been courted by presidents and prime ministers and he has seduced some of the world’s biggest investors.
There are some people who are, like, ‘all AI art is terrible’, but then there’s a lot of artists who are, like, ‘this is the best tool ever’
I find Altman brimming with confidence as our conversation ranges from AI products to the existential question of an AI future that a handful of optimistic technologists are steadily leading us to, whether we like it or not. Radiating ambition, he sounds like a man convinced of his own destiny. He tells me that he has the “coolest, most important job maybe in history” and while he used to think AI was as consequential as the Industrial Revolution, he now reckons the “explosion in creativity” makes the Renaissance a more apt analogy.
We are meeting soon after the release of OpenAI’s o3, a more advanced AI model with improved capability to reason and generate images. It is, he says, an important step towards the creation of AI agents that can execute tasks on humans’ behalf, and which all leading AI companies are furiously pursuing. “People are saying like, this is . . . genius-level intelligence,” he gushes.
No sooner was the tool released than users flooded the internet with images generated in the style of the Japanese animation house Studio Ghibli. That gave Altman and OpenAI a sensational marketing boost, but renewed questions about the liberal use of others’ intellectual property to train AI models and generate art. Altman says compensation for artists may be required (his company has done licensing deals with publishers, including the FT), but he prefers to put the tools out into the world and then find answers to questions that arise.
“There are some people who are, like, ‘all AI art is terrible’, but then there’s a lot of artists who are, like, ‘this is the best tool ever, it’s like the invention of the camera’,” says Altman. “We agree we need a new business model for this kind of a world, but what it is, the community is still sort of feeling their way through. I know that we’ve got to converge on what it should be.”
We are back in the kitchen and I watch Altman season with cumin the yellow and orange carrots grown on the farm, which are then roasted in the oven. With impressive determination, he chops an enormous amount of garlic, which he tosses into a pan with red chilli peppers, walnuts, parsley and pecorino to make what looks like a Californian take on aglio e olio spaghetti. The salad leaves, with thinly sliced carrots and radishes, are in the fridge already and need only dressing. Altman visibly enjoys cooking and, as I will soon find out, is rather good at it.
As we talk, I search for clues in his upbringing that hint at his future stardom. He says there are none. “I was like a kind of nerdy Jewish kid in the Midwest . . . So technology was just not a thing. Like being into computers was sort of, like, unusual. And I certainly never could have imagined that I would have ended up working on this technology in such a way. I still feel sort of surreal that that happened.”
The eldest of the four children of a dermatologist mother and a father who worked in real estate, Altman read a lot of science-fiction books, watched Star Trek and liked computers. In 2005, he dropped out of Stanford University before graduating to launch a social networking start-up. In those days, AI was still in its infancy: “We could show a system a thousand images of cats, and a thousand images of dogs, and then it [the AI] could correctly classify them, and that was, like, you were living the high life.”
Soon, Altman was running Y Combinator, a start-up accelerator that had backed his first venture. He was still there when he started OpenAI in 2015 as a non-profit with a mission that artificial general intelligence should benefit all of humanity. Musk was a co-founder who financed it with tens of millions of dollars but then fell out with Altman and left its board in 2018 in one of Silicon Valley’s most closely watched feuds. Musk, who has been growing his own AI rival company, xAI, alleges in a lawsuit that Altman had strayed from the company’s founding mission. Altman has countersued: “It seems clear to me why he’s doing all of this,” he says of Musk. “Because he’s trying to slow down a competitor and he doesn’t like that he’s like not winning in AI.”
Musk may have his personal motives but the debate over how to “win” in AI — which requires massive computing power and investment — without loosening commitment to safety, has long divided OpenAI. The splits exploded into the open in November 2023, when the non-profit board suddenly fired Altman, with one board member later accusing him of misrepresenting information and in some cases lying to the board.
Altman prevailed. He was reinstated within days, after nearly all employees threatened to resign and Microsoft, then the company’s largest financial backer, offered to hire Altman and his team. “It was very painful and very embarrassing that this whole thing happened, and no one, including me, really knew all of it at the time. What I wanted was to go sit on a beach and recover but I had to just keep running the company and now clean up a gigantic mess.” That mess, he says, included customers and investors asking whether they could depend on the company.
What I wanted was to go sit on a beach and recover but I had to just keep running the company and now clean up a gigantic mess
The storm led to the departure of some of OpenAI’s best researchers and left lingering questions about Altman. But it also cemented his status as the company’s undisputed leader, with a new board that backs him.
His ambitions too have expanded: early this year, he starred in another headline-grabbing event, appearing at the White House with Donald Trump to announce a joint venture with Japan’s SoftBank that will raise hundreds of billions of dollars to develop AI infrastructure, including data centres.
But while he has shifted the narrative around OpenAI’s mission as its commercial potential has become more apparent, attempts to change the company structure to a more traditional for-profit business have met with resistance, from Musk as well as from AI experts who insist the company must remain under a non-profit board to fulfil its mission of developing technology to benefit all humanity.
I ask Altman whether he learnt from the attempted coup. One criticism of him is that he tells people what they want to hear, depending on what’s expedient. All he will admit to is that he does prefer to avoid conflict and he’s had to learn quickly how to run such a complex company. “In the last two years we have gone through a decade and half of a normal company’s growth.”
It’s a busy day at the farm. Altman’s mother is visiting, as are his in-laws. His husband and son, as well as a colleague, are in the house too. They come in and out of the kitchen. Altman lays the bowls of food on a table, we serve ourselves and go back to sit in the garden. The pasta is delicious, with just the right amount of spice, the carrots crunchy, and the salad tastes light and juicy.
As a chief disrupter who is hyperactive on social media and, by his own description, OpenAI’s “marketing division”, he generates a constant stream of news. Recently that has included a dispute with an estranged sister who has accused him of sexually abusing her when she was a child (the rest of the family has backed his denials and he says he feels both compassion and upset towards a sister who’s “had a hard time of it for a long time”).
Altman claims that being competitive is not one of his defining features — “Am I not an outlier in terms of competitiveness,” he asks aloud, “compared to other tech CEOs?” — yet he relishes talk of winning. He admits that he has entertained running for governor of California (not a presidential bid though, as some have claimed); his favourite way of describing the reach of ChatGPT is not 800mn users but “10 per cent of the world”.
But how much does the competitive race ultimately matter? While Silicon Valley has been sinking massive investment into AI, DeepSeek, a Chinese start-up, released a model this year developed on a limited budget. That suggested that AI models were becoming commoditised and the US technological edge over China was diminishing. Altman says there is an “asterisk” to the commoditisation narrative: “Most of these models will be commoditised. The frontier models I don’t think will be.” He is, of course, expecting to prevail in frontier models but also win in the commoditisation game, given how many users are already attached to ChatGPT.
And how will OpenAI deliver the returns on massive investment? Altman hints at his ultimate goal, but describes it as just one compelling idea: when a subscription to ChatGPT becomes a personal AI, through which users log into other services. “You could just take your AI, which is gonna get to know you better over the course of your life, have your data in and be more personalised and you could use it anywhere. That would be a very cool platform to offer.”
The sun is too strong so we go back inside. Altman is sipping his tea, curled up on the living room sofa, his arms wrapped around his legs. We talk about the AI-dominated future his son will inherit. Diseases may be more rapidly cured by AI, and sectors from education to banking transformed. But it is a world that also raises existential questions about the way we live. Why should society trust a handful of AI men to decide on the shape of the future? In a response unlikely to convince, he says those developing the technology are “committed to meeting the gravity of the moment with responsible technology”.
AI progress is moving at such breathtaking speed that some experts favour slowing down until internationally agreed norms and regulations are put in place. Two years ago, Altman himself signed a statement with others in the field cautioning that “mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war”.
Altman insists he hasn’t changed his mind and there will be moments when holding back may be required. For now, he seems satisfied with the rollout of tools for people to experiment with and assess the risks. “The world needs to know about it [AI], the world needs to weigh in on it, very heavily. By having our users help us decide what the limits should be, like learning this collective value, function and preferences of humanity,” he says.
Some advances do scare him. After releasing a memory feature that allows the AI to register past behaviour, he has heard of cases where users become too emotionally dependent on the AI. “People are like, this is my new best friend, you can never delete this version, I need this thing . . . I have no doubt that we, society, will figure out how to navigate this, but that’s a new thing that’s just happened and you can imagine all sorts of ways that it goes really wrong.”
More alarming, I note, is a future in which AI agents communicate with each other without instruction by humans. Altman explains that perhaps it’s not an agent that creates other agents but an AI system that is so good, so trusted, that it in effect controls what humans do. “It [the AI model] becomes just better than we sort of have a conception for.” This sounds so spooky that even he seems alarmed by his words.
Altman doesn’t strike me as a man who entertains doubt but I ask how his grand plan of building an AI giant could go wrong. Is he confident that OpenAI will exist in 10 years? “Fixing fences and taking care of cows” would be his plan B, he jokes. More seriously, he says: “We could make a wrong research bet, you know, we could fall behind on product to somebody else. It’s like we’re doing a very complicated thing.”
We’ve been talking for more than two hours and his husband, holding the baby, joins us in the living room, and after some fussing, the tiny infant is falling asleep. I ask whether Altman finds his brave new world, in which humans are not the most intelligent thing on the planet, threatening — if not for him, then for his son?
He is, predictably, too enthralled by his AI creation to feel menace. “Do you think you’re smarter than o3 right now? I don’t . . . and I feel completely unbothered, and I bet you do too,” he says. “I’m hugging my baby, enjoying my tea. I’m gonna go do very exciting work all afternoon . . . I’ll be using o3 to do better work than I was able to do a month ago. I’ll go for a walk tonight. I think it’s great. I’m more capable. He [his son] will be more capable than any of us can imagine.”
