Altman: Training Humans Takes Resources, Too

"People talk about how much energy it takes to train an AI model," OpenAI boss Sam Altman said, "but it also takes a lot of energy to train a human." The OpenAI chief executive made the comment while in India for an A1 Summit when addressing the resources the industry will require, the Guardian reports. Critics of AI's electricity use overlook how much food and time it takes for a person to become "smart," he told the Indian Express, over the course of about the 20 years needed. Nevertheless, Altman called concerns over AI's power demands fair and said the world needs to rapidly ramp up nuclear, wind, and solar power.

Datacenters already use an estimated 1.5% of global electricity, and the International Energy Agency expects that to grow roughly 15% a year through 2030—far faster than other sectors. On the other hand, concerns about AI's water use are "totally fake," Altman contended. He said modern data centers don't rely on evaporative cooling, despite some experts noting many facilities still do use significant amounts of water. Environmental groups have pushed for a US pause on new data centers, warning of threats to climate, communities, and water supplies.

Online criticism called Altman's human-versus-AI comparison dehumanizing. Some sustainability experts questioned whether the most common uses of ChatGPT—like writing help—justify its environmental cost. Alatman told the event that India has more than 100 million weekly ChatGPT users active, per TechCrunch, second only to the US, and that the country has the most students using ChatGPT.