Experts are skeptical about Google’s AI water consumption claims

Here is a 192-word summary of the news article: AI experts are questioning the accuracy of Google's claims about the water consumption of its Gemini AI model. According to the report, Google stated that a typical query to Gemini consumes only "five drops of water." However, this figure is facing criticism from several AI researchers, including Shaolei Ren, one of the authors of a report cited by Google. Ren estimated that Microsoft's data center consumed 700,000 liters of water to train OpenAI's GPT-3 model, and that a 20-50 message ChatGPT conversation can consume close to a pint of water - far more than Google's estimate. Experts argue that Google's figures only account for the water used to cool its data centers, while ignoring the water consumption of the power plants that supply the electricity. Researchers are also skeptical of Google's claims about the carbon emissions of its AI models, noting that the company used market-based emissions figures that exclude electricity certificates and carbon credits, and therefore do not accurately reflect the actual CO2 emissions. Overall, the article suggests that Google's water and emissions claims are oversimplified and do not provide the full picture of the environmental impact of its AI technologies.