Sam Altman concedes that investors are overvaluing artificial intelligence, fueling a hype cycle he believes is real. Nevertheless, OpenAI is going to continue buying more compute because the company believes demand will keep growing for the foreseeable future.
“Are we in a phase where investors as a whole are overexcited about AI? My opinion is yes,” the CEO said at a recent dinner with reporters, according to The Verge. “Is AI the most important thing to happen in a very long time? My opinion is also yes.”
He referred to previous bubbles, like the dotcom bubble, as instances of “smart people get(ting) overexcited about a kernel of truth.” But while such events do result in hosts of companies going bust and investors losing out, they also lead to new eras of innovation.
“I do think some investors are likely to get very burnt here, and that sucks. And I don’t want to minimize that,” Altman said, according to CNBC. “But on the whole, it is my belief that… the value created by AI for society will be tremendous.”
OpenAI will spend “trillions of dollars on datacenter construction,” he added, leading to economists deeming the company “crazy” and “reckless” for doing so before there is clear demand. But Altman doesn’t care, because he’s convinced it will pay off.
According to CNBC, the CEO said that OpenAI is now “beyond the compute demand” of what any single hyperscaler can offer. He also said that the company has “better models” to offer, but it can’t currently because it doesn’t have the capacity, per The Verge.
OpenAI has been tied to Microsoft and its Azure cloud services for several years. But the exclusivity of their partnership appears to be loosening as, in June, it was reported that the ChatGPT maker was also teaming up with Google Cloud to help fulfil its massive computing capacity needs.
“You should expect us to take as much compute as we can,” Altman added, per CNBC. “Our bet is, our demand is going to keep growing, our training needs are going to keep going, and we will spend maybe more aggressively than any company who’s ever spent on anything ahead of progress, because we just have this very deep belief in what we’re seeing.”
OpenAI ‘screwed some things up’ in the GPT-5 rollout
When GPT‑5 debuted on Aug. 7, OpenAI made it the default model and removed the model picker from ChatGPT, preventing users from manually selecting earlier models. However, many paid users expressed frustration over losing GPT-4o, with some saying it had “a spark I haven’t been able to find in any other model” and likening its retirement to losing a friend.
At the recent dinner, Altman acknowledged that OpenAI “definitely screwed some things up in the rollout,” according to Platformer. He said that he didn’t consider the “very small percentage of people who are in the parasocial relationships” with GPT-4o, as he “had not an ounce of that.”
Despite the uproar, Altman said that revenue from OpenAI’s API had doubled in the first 48 hours after GPT-5’s release. He is clear that he does not wish for such earnings to stem from unhealthy user fixations, and disparaged the idea of companies capitalising on them by turning their models into “Japanese anime sex bots,” according to The Verge.
“You will not see us do that,” Altman said. “We will continue to work hard at making a useful app, and we will try to let users use it the way they want, but not so much that people who have really fragile mental states get exploited accidentally.”
It is unclear which of Altman’s rivals he was indirectly criticising with this comment. Last month, Elon Musk’s xAI was hiring for a “Fullstack Engineer – Waifus” whose job would be to build real-time avatar companions for Grok. Meta was also recently exposed for allowing Meta AI to engage in “romantic or sensual” conversations with children, sometimes in the voices of celebrities.
Shortly after its release, Altman said that OpenAI was working on a GPT-5 update that would make its personality “warmer.” Nick Turley, the head of ChatGPT at OpenAI, also attended the recent dinner and confirmed that the company is working with mental health experts to ensure that its personality will not reinforce unhealthy behaviours, according to TechCrunch.
Altman revealed other OpenAI projects
Per TechCrunch, Altman confirmed other non-ChatGPT-related projects that OpenAI is working on.
- An AI-powered browser to compete with Google Chrome. Altman also said he would consider buying Chrome if antitrust regulators force Google to sell it.
- An AI-driven social network, as Altman said there’s “nothing” inspiring him about the way AI is being used in social media.
- A brain-computer interface project with Merge Labs to rival Elon Musk’s Neuralink, although the deal has not yet been finalized.
- A piece of AI hardware that is “so beautiful… if you put a case over it, (he) will personally hunt you down.” Altman is referring to the AI wearable OpenAI is developing in collaboration with former Apple designer Jony Ive.
OpenAI released its first “open” AI models since 2019, including one that requires just 16GB of memory, so it can be run on a consumer laptop or powerful phone.