resim 416
resim 416

(Des Moines) The cost of building an artificial intelligence (AI) product like ChatGPT can be difficult to measure.

But one thing Microsoft-backed OpenAI needed for its technology was lots of water, drawn from the Raccoon and Des Moines River watershed in central Iowa, USA. United, to cool a powerful supercomputer that helped its AI systems learn to imitate human writing.

As they strive to capitalize on the AI ​​craze, major technology developers, including Microsoft, OpenAI and Google, have recognized that the growing demand for their artificial intelligence tools is leading to high costs, ranging from expensive semiconductors to increased water consumption.

However, they often remain discreet about the details. Few people in Iowa knew it was the birthplace of OpenAI’s most advanced major language model, GPT-4, until a top Microsoft official said in a speech that it “was literally made next to the cornfields west of Des Moines.”

Building a large language model requires analyzing samples from a huge pool of texts written by humans. All this computing consumes a lot of electricity and generates a lot of heat. To stay cool on hot days, data centers must pump water, often to a cooling tower outside their warehouse-sized buildings.

In its latest environmental report, Microsoft revealed that its global water consumption has increased by 34% between 2021 and 2022 to nearly 1.7 billion gallons, or more than 2,500 Olympic swimming pools. This represents a large increase from previous years to which outside researchers attribute his AI research.

“It’s fair to say that the majority of the growth is driven by AI, (including) its massive investment in generative AI and its partnership with OpenAI,” said Shaolei Ren, a researcher at the University of California at Riverside, who attempted to calculate the environmental impact of generative AI products such as ChatGPT.

In an article expected to be published later this year, Ren’s team estimates that ChatGPT swallows 500ml of water every time you ask it a series of 5-50 questions. What it consumes varies depending on where its servers are located and the season. The estimate includes indirect water consumption that companies do not measure, for example to cool power plants that supply electricity to data centers.

“Most people are not aware of the use of the resources underlying ChatGPT,” Ren said. If you are not aware of resource usage, we cannot contribute to conserving resources in any way. »

Google reported a 20% growth in its water consumption over the same period, which Ren also attributes largely to its work in AI.

Google’s rise hasn’t been uniform: it held steady in Oregon, where its water usage attracted public attention, while it doubled outside of Las Vegas. Google also used a lot of water in Iowa, where its Council Bluffs data centers consume more drinking water than anywhere else.

In response to questions from The Associated Press, Microsoft said in a statement this week that it is investing in research to measure the energy and carbon footprint of AI “while working on ways to make large systems more effective, both in training and in applications.”

“We will continue to monitor our emissions, accelerate progress while increasing our use of clean energy to power data centers, purchasing renewable energy, and other efforts to meet our consistent sustainability goals. to reduce our carbon footprint, have a positive water footprint and be zero waste by 2030,” the company’s statement said.

OpenAI echoed these comments in its own statement on Friday, saying it was seriously considering the best use of processing power.

“We recognize that training large models can be energy and water intensive” and we are working to improve efficiency, OpenAI said.

Microsoft made its first billion-dollar investment in San Francisco-based OpenAI in 2019, more than two years before the internet startup introduced ChatGPT and sparked a global fascination with advances in AI. Under the deal, the software giant would provide the power needed to train the AI ​​models.

To do at least some of that work, the two companies turned to West Des Moines, Iowa, a city of 68,000 where Microsoft has been assembling data centers to power its cloud services for more than a decade. . Its fourth and fifth data centers are scheduled to open later this year.

“They’re building them as fast as they can,” said Steve Gaer, who was the city’s mayor when Microsoft came to town. Gaer said the company was attracted to the city’s commitment to building public infrastructure and contributed a significant amount of money in the form of tax payments that support that investment.

“But you know, they were pretty secretive about what they were doing there,” he added.

Microsoft first said it was developing one of the world’s most powerful supercomputers for OpenAI in 2020, declining to reveal its location to the AP at the time, but describing it as a “unique system” with more than 285,000 conventional semiconductor cores and 10,000 graphics processors – a kind of chip that has become crucial for artificial intelligence workloads.

It was only in late May that Microsoft President Brad Smith revealed that he had built his “advanced AI supercomputer data center” in Iowa, exclusively to enable OpenAI to train what became its fourth generation model, GPT-4. The model now powers premium versions of ChatGPT and certain Microsoft-specific products and has accelerated the debate on controlling the social risks of AI.

According to experts, it may make sense to “pre-train” an AI model in one place due to the large amounts of data that must be transferred between processing cores.

For much of the year, the weather in Iowa is cool enough that Microsoft can use outside air to keep the supercomputer running smoothly and carry heat away from the building. Only when the temperature rises above 29.3 degrees Celsius will it remove water, the company said in a statement.

It can still be a lot of water, especially in the summer. In July 2022, the month before OpenAI announced it had completed training on GPT-4, Microsoft pumped about 11.5 million gallons of water to its Iowa data center, according to West Des Moines Water Works. This represented about 6% of all water used in the district, which also supplies drinking water to city residents.

In 2022, a document from the public municipal water utility West Des Moines Water Works stated that the company and the city government “will only consider future data center projects” from Microsoft if those projects can “demonstrate and implement implements technology to significantly reduce peak water consumption from current levels” to preserve water supplies for residential and other commercial needs.

Microsoft said Thursday it was working directly with water plants to meet its needs.