OpenAI CEO Sam Altman is pushing back against criticisms of AI's environmental footprint. He argues that the energy consumption of AI models is comparable to the energy expended during a human lifespan, and dismisses concerns about water usage in data centers as overblown. This comes as AI development accelerates and its resource demands face increasing scrutiny.
Comparing AI and Human Energy Consumption
Altman argues it's "unfair" to compare the energy costs of AI training to a single human query because "it also takes a lot of energy to train a human." He elaborated, stating, "It takes like 20 years of life and all of the food you eat during that time before you get smart."He suggests that when factoring in the resources needed for human development, AI's energy footprint is already competitive. "Probably AI has already caught up on an energy efficiency basis" compared to humans, Altman said.
Dismissing Water Usage Concerns
Altman directly addressed concerns about water consumption in AI data centers. "Water is totally fake," he stated, claiming that evaporative cooling, a water-intensive practice, is no longer prevalent in data centers. He criticized claims that a single ChatGPT query requires significant water usage, suggesting such claims are outdated.Power-hungry data centers often require large amounts of water to cool electrical systems, but Altman suggests this is becoming less of an issue [1]. Some newer data centers don’t rely on water usage at all, yet concerns persist as computing demand continues to rise.
