Sam Altman is pushing back on the idea that ChatGPT consumes too much energy.
“One of the things that is always unfair in this comparison is people talk about how much energy it takes to train an AI model relative to how much it costs a human to do one inference query,” Altman told The Indian Express last week on the sidelines of a major AI summit. “But it also takes a lot of energy to train a human.”
Altman suggested it’s not an apples-to-apples comparison, arguing that it’s unfair to discount the years spent nurturing and educating someone to be capable of making their own inquiries.
“It takes a lot of energy to train a human,” he said, prompting some laughter in the crowd. “It takes, like, 20 years of life, and all of the food you eat during that time before you get smart.”
Altman said the clock really began thousands of years ago.
“It took, like, the very widespread evolution of the 100 billion people that have ever lived and learned not to get eaten by predators and learned how to, like, figure out science or whatever,” he said.
Altman also called out what he said were “totally insane” claims on the internet that OpenAI is guzzling down water to power ChatGPT.
“Water is totally fake,” Altman said, when asked about concerns AI companies use too much water. “It used to be true, we used to do evaporative cooling in data centers, but now that we don’t do that, you know, you see these like things on the internet where, ‘Don’t use ChatGPT, it’s 17 gallons of water for each query’ or whatever.”
In June, Altman said that the average ChatGPT query consumes roughly the amount of energy needed to power a lightbulb for a few minutes.
“People are often curious about how much energy a ChatGPT query uses; the average query uses about 0.34 watt-hours, about what an oven would use in a little over one second, or a high-efficiency lightbulb would use in a couple of minutes,” he wrote on X.
Altman said it is fair as a whole to point out the AI industry’s overall energy consumption because of the large growth in usage. He said it’s why he and other AI CEOs have pushed alternative energy sources like solar, wind, and nuclear.
Unlike other CEOs, namely xAI’s Elon Musk, Altman is dismissive of the idea that space-based data centers are realistic in the next decade, a concept that some companies have floated as a way to reduce energy consumption.
Outside of OpenAI, Altman is a major investor in nuclear energy. He previously served as chairman of Oklo, a nuclear energy startup, and has been a major backer of Helion, which plans to build what it calls “the world’s first fusion power plant” in Washington state.
In the US, data center energy consumption is becoming a major topic. Last month, President Donald Trump said he was working with tech companies on “a commitment to the American people” to ensure that citizens don’t pay higher energy bills because of a nearby data center.
Consulting firm McKinsey & Company estimated last year that data centers could account for 14% of total power demand in the US by 2050.
Read the original article on Business Insider
The post Sam Altman says concerns of ChatGPT’s energy use are overblown: ‘It also takes a lot of energy to train a human’ appeared first on Business Insider.
