Sam Altman defends AI's resource use, compares energy use to humans

Sam Altman, the CEO of OpenAI, defended the demands of artificial intelligence (AI) on resources, labelling concerns about the water usage of data centers as "fabricated" while likening the energy requirements of AI systems to those of humans.

Altman addressed these concerns in a conversation with The Indian Express on the sidelines of the India AI Impact summit. 

He tackled common criticisms, particularly focusing on AI's consumption of energy and water.

The CEO remarked that the online claims suggesting ChatGPT consumes significant amounts of water per individual query were "utterly false and ridiculous," with "no basis in reality."

Traditionally, data centers consume substantial water quantities to cool down electronic components and prevent overheating. 

Although cooling technologies for data centers have promised reduced use, some newer data centers have started operating without relying on water at all.

Despite these advances, a recent report by the water technology firm Xylem and Global Water Intelligence predicted a more than threefold increase in water drawn for cooling over the next 25 years, given the growing demand for computing, which could stress water resources.

While dismissing concerns related to water usage, Altman acknowledged that energy consumption remains a valid concern for AI. 

"Not per query, but overall – as the global use of AI increases...we must swiftly transition to nuclear, wind, and solar energy sources," he stated.

When asked to comment on previous remarks by Microsoft founder Bill Gates—who suggested that since the human brain is efficient, AI could also become more energy-efficient over time—Altman disagreed.

"The comparison often feels unfair. People frequently discuss the energy required to train an AI model...But training a human requires significant energy too," he said. "Think of the 20 years of life and all the food consumed during that time before someone becomes knowledgeable."

"A fairer comparison arises when you consider the energy needed for ChatGPT to answer a question once trained versus a human, and AI probably rivals humans in energy efficiency, measured that way," he added.

The process Altman describes is known as inference, involving trained AI models generating new outputs. Typically, AI inference demands much less power than initial training.

Altman's remarks, especially regarding the AI-to-human analogy, have sparked online debates amid worrying discourse about AI replacing human roles.

Sridhar Vembu, the co-founder and chief scientist of Zoho Corporation, an Indian software company, expressed criticism at the summit regarding human-AI comparisons. "I oppose the notion of equating technology with human beings," the billionaire expressed in an X post.

This discussion arises as both governments and companies heavily invest in new data centers to meet the computing needs of AI systems.