A recent study has cautioned that the energy needs of the artificial intelligence (AI) industry could reach the level of a nation as large as the Netherlands by 2027.
Since ChatGPT made its debut last year, large tech companies have been rushing to implement AI-driven services.
They consume significantly more energy than normal programs, thus requiring much more power when going online.
The research additionally suggested that if the current rate of development for AI was moderated, then its environmental impact might not be as dire as previously thought.
Experts, among them the writer of this report, have stated that trying to make an accurate forecast is difficult due to the fact that tech companies do not give away enough information.
No doubt, AI necessitates hardware of greater strength than conventional computing activities.
Alex De Vries, a PhD candidate at the VU Amsterdam School of Business and Economics, conducted a study under the assumption that certain parameters such as the rate of AI growth, access to AI chips, and consistent server capability remain unaltered.
Mr De Vries reckoned that Nvidia, the chip designer, is projected to provide around 95% of the AI processing gear necessary by the industry.
He was able to calculate that AI would likely consume between 85 and 134 terrawatt-hours (TWh) of electricity annually based upon the projected delivery of these computers by 2027.
Approximately, the amount of energy consumed yearly by a tiny nation is equivalent to the number stated above.
He informed BBC News that the Netherlands would make up only a fraction of the total global electricity usage, estimating it at about half of a percent.
Nvidia chose not to make a statement.
Mr De Vries asserted that his research indicates AI should only be employed when absolutely necessary.
The journal Joule has published his peer-reviewed study.
Specialized computer setups known as data centres are necessary for AI systems such as large language models powering popular chatbots, such as OpenAI's ChatGPT and Google's Bard.
This implies that the machinery is more demanding of energy and, as with typical machinery, needs to be kept…
The research excluded the energy sourced for cooling. Several of the large tech companies don't quantify this particular energy use or water consumption. Mr de Vries is one of those asking for the industry to be more open about the same.
No doubt, the requirement for computers for AI is surging, and with it the energy required to keep the servers functioning smoothly.
Danny Quinn, the head of the Scottish-based data centre company DataVita, stated that since the beginning of 2023, the amount of inquiries about placing AI hardware in his facility has grown from a mere one or two every week to multiple hundreds.
He additionally highlighted the contrast in energy consumption between a tray of normal servers and one having AI chips.
A normal rack of equipment is estimated to require 4kWh of power, which equals what a family house requires. However, a rack of AI gear requires 20 times more energy, roughly 8kWh. Data centres can contain hundreds or even thousands of these racks.
He remarked that, although Scotland's cooler and damper weather undeniably does aid them in the process of cooling down the machines, it is still a daunting challenge.
Microsoft's latest sustainability report reveals that, with its heavy investments in AI, the company's water consumption increased by 34% from 2021 to 2022, reaching approximately 6.4 million cubic metres - equivalent to the size of 2,500 Olympic swimming pools.
Prof Kate Crawford, author of a book about AI and its impact on the environment, revealed that she finds the subject deeply troubling and it keeps her awake at night.
She told the BBC in July that these electricity and energy-consuming AI supercomputers require vast amounts of water for cooling purposes. Consequently, she continued, there is likely to be a big energy-extracting industry in the 21st Century.
It is anticipated that AI could be of assistance in addressing some of the environmental struggles confronting our world.
Google and American Airlines recently discovered that pilots could reduce the amount of contrails (vapor trails) that airplanes produce by half with the assistance of an experimental AI tool to select an optimal altitude. It is widely believed that contrails have a negative effect on the climate due to climate change.
The US government is investing millions of dollars in efforts to replicate nuclear fusion, which is what the Sun uses to generate energy.
Achieving success in this endeavor would be a major milestone, and would unlock a never-ending, renewable energy source. Artificial intelligence could be utilized to expedite the research that has plodded along since the 1960s with limited success.
In February of this year, Brian Spears, a university professor, declared he had utilized AI to foresee the result of an experiment, resulting in a major discovery.
He noted that for 100 trillionths of a second, they had generated ten petawatts of energy, making it the brightest thing in our solar system.
Commentaires