OpenAI CEO Sam Altman took questions from the crowd.
His responses during a live Q&A on Tuesday were almost as illuminating as the questions themselves.
Altman outlined OpenAI’s shifting definition of AGI and his vision for the company’s future, acknowledged some recent blunders, once again tried to appease users who don’t wish to part with older models, and promised that adults would soon have greater freedom, well, to a point.
Here’s a look at his most revealing quotes.
‘Our aspiration is that we can build an infrastructure factory where we can create one gigawatt a week of compute.’
After a string of deals, OpenAI has current commitments for $1.4 trillion in spending on AI infrastructure for roughly 30 gigawatts of new compute in the coming years. Altman wants even more.
On Tuesday, he said that if OpenAI’s research continues to advance alongside sustained consumer demand, he would like OpenAI to build infrastructure that would produce 1 gigawatt of compute a week. For comparison, Nvidia CEO Jensen Huang has said that 10 gigawatts is roughly equal to between 4 million and 5 million graphics processing units (GPUs).
“To be clear, we’re not committed to this yet, but we are having conversations about it,” Altman said. “Our aspiration is that we can build an infrastructure factory where we can create one gigawatt a week of compute. And aspirationally, we would like to get that cost down significantly to like $20 billion over that five-year lifecycle of that equipment.”
Related stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
‘I wish I had used an example other than erotica.’
During the roughly hourlong presentation and live Q&A, Altman reflected on his own leadership. In discussing OpenAI’s goal of trusting users, he said that he botched a recent post when he said that ChatGPT would allow more erotica for appropriately aged users.
Some notable figures, including Mark Cuban, questioned Altman’s post.
“Now, I made one of my many stupid mistakes when I tried to talk about this recently. I wish I had used an example other than erotica,” Altman said. “I thought there was an understandable difference between erotica and porn bots, but in any case, we were trying to show the point that we’re trying to get across is that people need a lot of flexibility and people to use these things in different ways, and we want to treat our adult users like adults.”
‘We are not going to promise to keep it around until the death of the universe.’
No recent OpenAI Q&A has been complete without users complaining about how OpenAI has treated its older models.
Altman previously commented on how some users’ attachment to OpenAI’s 4o model was “different” than how people felt about previous technology. Altman and OpenAI briefly removed 4o from ChatGPT following the launch of GPT-5 but later restored it for paid users amid a massive online backlash.
When asked about OpenAI’s plans for 4o, Altman said there are no plans to sunset 4o.
“We are not going to promise to keep it around to the death of the universe either, but we understand that it’s a product that some of our users really love,” he said.
‘We’re not going to do the equivalent of selling heroin or whatever, even if you sign a liability.’
Multiple questioners expressed uneasiness about ChatGPT’s age verification system. One posed a novel concept to Altman: would OpenAI essentially open the floodgates for verified adult users if they signed a liability waiver protecting OpenAI?
Altman was clear that age-verified users will have “flexibility” and more freedom than current safety routers within ChatGPT allow, but that is only to a point.
“Again, we’re not going to do the equivalent of selling heroin or whatever, even if you sign a liability,” he said. “But yes, on the principle of treating adult users like adults, if you are age-verified, you will get quite a lot of flexibility.”
‘It’s much more useful to say our intention, our goal is by March of 2028 to have a true automated AI researcher.’
Tuesday was a historic day for OpenAI, as it announced it had completed its restructuring and reached a definitive agreement with Microsoft, one of its earliest investors. One of the biggest threads throughout both the announcement and the livestream was artificial general intelligence, or AGI, the theoretical moment when AI can reason like humans.
AGI has been a motivating force for OpenAI since its days as a research lab. Beyond a potential historic moment, reaching AGI was also a key component of OpenAI’s partnership with Microsoft since it allowed OpenAI to isolate Microsoft from its technology. OpenAI’s new agreement with Microsoft adds an extra layer to any declaration of when AGI is reached.
Altman’s broader point during the livestream and Q&A that followed was to try to move beyond the long-running debate over what AGI even means. Instead, Altman and Jakub Pachocki, OpenAI’s chief scientist, said their internal goals are to deliver an automated AI research intern in September 2026 with more advancements to come.
“It’s much more useful to say our intention, our goal is by March of 2028 to have a true automated AI researcher, and to define what that means than it is to sort of try to, you know, satisfy with the definition of AGI.”
