OpenAI CEO Sam Altman spoke out about the debate surrounding the environmental impact of artificial intelligence, emphasizing that much information circulating on the internet about AI water consumption is inaccurate.
Speaking during a visit to India and attending an AI conference held on February 19-20 (local time) in New Delhi, Mr. Altman said that allegations that each ChatGPT query consumes dozens of gallons of water are "completely untrue".
Mr. Sam Altman admitted that water use was once a problem when some data centers applied evaporative cooling methods, but current technology has changed significantly.
You will see information like don't use ChatGPT because each query costs 17 gallons of water. That is completely unrelated to reality," Mr. Altman said.
OpenAI CEO also admitted that concerns about the overall energy consumption of AI are reasonable, because the demand for artificial intelligence is increasing rapidly globally.
According to him, the long-term solution is not to limit AI but to accelerate the transition to clean energy sources such as nuclear, wind and solar.
Currently, there are no legal regulations requiring technology companies to publicly disclose the amount of electricity and water consumed for AI systems, causing many scientists to conduct independent research.
Some reports also point out that the rapid development of data centers may contribute to increasing electricity prices in some areas.
When asked about the opinion mentioned in a conversation with Bill Gates that a ChatGPT query can consume energy equivalent to 1.5 times a iPhone battery charge, Altman denied: "It can't be that much.
He believes that current debates are often unfair when focusing only on the energy needed to train AI models, without comparing it with the energy cost for humans to achieve similar thinking abilities.
According to Altman, for a human to become intelligent, it takes about 20 years of development along with the amount of food consumed during that time, not to mention the evolutionary process of mankind.
A fair comparison is to see how much energy AI needs to answer a question compared to humans after training. If measured in that way, AI may have achieved equivalent energy efficiency," he raised the issue.
Altman's point of view reflects the growing debate about the environmental impact of artificial intelligence as this technology develops rapidly.
In the context of a sharp increase in computing demand, many experts say that the challenge lies not only in AI but also in the global energy infrastructure.
