Sam Altman would like remind you that humans use a lot of energy, too
OpenAI CEO Sam Altman this week expressed his concerns about the impact of AI on the environment speaking at an event organized by The Indian Express.
For starters, Altman — who was in India for a major AI summit — said concerns about AI’s water use are “completely bogus,” although he acknowledged it was a real problem when “we used to do evaporative cooling in data centers.”
“Now that we don’t do that, you see things on the internet where ‘Don’t use ChatGPT, it costs 17 liters of water for every search’ or whatever,” Altman said. “This is completely untrue, completely insane, has no bearing on reality.”
He added that it is “fair” to be concerned about “energy consumption – not per query, but in aggregate, because the world is using so much AI now.” According to him, this means that the world must “very quickly move towards nuclear, wind and solar energy.”
There is no legal obligation for technology companies to disclose how much energy and water they use, but scientists have done so try to study it independently. Data centers have also been linked to rising electricity prices.
Referring to a previous conversation with Bill Gates, the interviewer asked if it was accurate to say that a single ChatGPT query currently uses the equivalent of 1.5 iPhone battery charges, to which Altman replied, “There’s no way it’s even close.”
Altman also complained that many discussions about ChatGPT’s energy consumption are “disingenuous,” especially when they focus on “how much energy it takes to train an AI model, relative to how much it costs a human to perform one inference.”
“But it also takes a lot of energy to train a human,” Altman said. “It takes about twenty years of your life and all the food you eat during that time for you to become smart. And not only that, it took the very widespread evolution of the 100 billion people who have ever lived and learned not to be eaten by predators and learned how to figure out science and whatever to produce you.”
So, he says, the fair comparison is: “If you ask ChatGPT a question, how much energy does it take once the model is trained to answer that question versus a human? And it’s likely that AI has already caught up in terms of energy efficiency, measured that way.”
You can watch the full interview below. The conversation about water and energy consumption starts around 26:35.
