The ever-growing investments in artificial intelligence infrastructure do not account as "a bubble" but rather represent today's "new normal" to meet the rising user demand, Fidji Simo, OpenAI's de facto No. 2, said on Monday.
The French-born executive made her comments in an interview with Agence France-Presse (AFP), her first since assuming the role of chief operating officer (COO) of OpenAI's applications, including its flagship ChatGPT model.
In the past few weeks, her company, under the leadership of CEO Sam Altman, has made a series of significant investments in data centers and AI chips, despite no clear signs that the rapidly emerging AI business is close to breaking even.
Simo provided some insights related to this, including thoughts on the dangers and breakthroughs of AI. The answers were lightly edited for length and clarity.
"What I am seeing here is a massive investment in compute (or computing power), with us meeting that need for computing power so incredibly badly for a lot of use cases that people want. (Video AI generator) Sora is a great example right now – there's much more demand than we can serve," she said.
"From that perspective, I really do not see that as a bubble. I see that as a new normal, and I think the world is going to switch to realizing that computing power is the most strategic resource."
Simo went on to say that she sees her job "as really making sure that the good side of this technology happens and we mitigate the bad side."
"Take mental health, for example. I'm hearing tons of users say that they go to ChatGPT for advice in tough moments where they may not have other people to talk to," she said, citing that many people can't afford to go to a therapist.
"I talk to a lot of parents who are telling me: God, I got this really awesome advice that helped me unlock a situation with my child. But at the same time, we need to make sure that the model behaves as expected," she furthered.
Moreover, concerning mental health, Simo recalled a road map of the AI startup.
"On mental health, we have announced a very robust road map. We started with parental controls," she noted.
"We have plans to launch age prediction: if we can predict that the user is a teenager, we give them a model that is less permissive than we would give to an adult," she added.
"I think the breakthroughs are about models understanding your goals and helping you accomplish them proactively," said Simo.
She also provided an example by saying: "Not just give you a good answer to a question, not just have a dialog, but actually tell you, 'Oh, okay, you're telling me that you want to spend more time with your wife. Well, there might be some weekend getaways that would be helpful, and I know it's a lot to plan. So, I've already done all the planning for you and made some reservations. Just tap one button to approve and everything gets done.'"
"We're still very early, but we're on that journey to capture that," she highlighted.
Regarding the use of OpenAI's popular chatbot, ChatGPT, by children, Simo clarified that it is not intended for those under 13. She added that her 10-year-old child still uses it under her supervision.
"It's magical to see what she's able to create. Just this weekend, she was telling me about creating a new business," the OpenAI executive said.
"She was using ChatGPT to make banners for the new business, to create taglines," she added.
"In our childhood, we couldn't turn our imagination into something real that fast. And I see that really giving her superpowers, where she thinks anything is possible."
