The global experiment with artificial intelligence (AI) is just beginning. But the investment frenzy of big tech companies in their quest to build and lease data centers, the engine rooms of AI, is in full swing. They spent an estimated $105 billion last year on these massive, electricity-guzzling facilities.
That spending has led to increased demand for electricity and has raised environmental concerns. A recent headline in the publication The New Yorker He called AI’s use of energy “shameless.” But AI and the environment can be viewed from another perspective that focuses not on how the technology is made but on what it can do.
AI could help accelerate scientific discovery and innovation in all kinds of fields, thereby improving efficiency and reducing planet-warming carbon emissions in sectors such as transportation, agriculture and energy production.
This is what we need to know.
What makes these data centers consume so much electricity?
It is the emergence of so-called generative artificial intelligence.
Generative AI can do a lot of things, not just analyze data and make predictions, but also write poetry and computer code, summarize books, and answer questions—generally, with the same ability as a human being. And computing activity of that sort requires a lot of energy. When we ask ChatGPT a question, it uses nearly 10 times more electricity than a regular Google search, according to a recent calculation.
Although several researchers had been working on generative AI for years, it really took off in November 2022 when OpenAI introduced ChatGPT, the conversational chatbot that became a sensation. Microsoft has invested more than US$13 billion in OpenAI and wants to include AI features in its products as soon as possible. So are Amazon, Google and Meta, the owner of Facebook, Instagram and WhatsApp.
How much will electricity demand increase?
There are higher estimates, but experts generally expect energy consumption in data centers around the world to at least double in the coming years. Goldman Sachs estimates that data centers will use 160 percent more electricity by 2030. A recent forecast by the International Energy Agency projects demand will more than double by 2026.
These predictions represent considerable increases, suggesting that data centers will see a marked increase in greenhouse gas emissions if they source their energy from fossil fuels such as coal and natural gas. But keep in mind that the global electricity sector is huge and diverse. Data centers are responsible for about one to two percent of total electricity demand. That share is projected to rise to three to four percent by 2030.
How can AI be presented as a green technology?
Some experts say artificial intelligence is a general-purpose tool that, if used wisely across the economy, could reduce greenhouse gas emissions by five to ten percent by 2030, according to a study commissioned by Google and the Boston Consulting Group.
For example, the technology promises to “spur biological design,” said Drew Endy, a professor of bioengineering at Stanford University. He said the result could be turbocharged biology if the right DNA formulas are discovered to make agriculture more efficient and less polluting, for example.
AI could also radically transform the way we find metals that are critical not only to the tech industry but to combating climate change. In one specific case, it helped find a vast deposit of copper in Zambia, a key component for electric vehicles.
And Zanskar, a Salt Lake City startup, is trying to improve the success rate of discovering geothermal energy for power plants with AI. The main reason about 90 percent of start-up geothermal projects fail is that they drill in the wrong place, said Carl Hoiland, Zanskar’s co-founder and chief executive. But combining AI with new geological data sets like satellite data and seismic sensors could make it possible to double or triple the field’s meager success rate.
In theory, that could make a big difference in the fight against global warming. Geothermal energy is a clean source of power available 24 hours a day, but currently accounts for less than one-half of one percent of electricity in the United States.
The conclusion
Although the electricity required for AI is expected to at least double in the coming years, the technology’s efficiency could rise at an even higher rate. There is historical precedent.
Consider what happened with cloud computing. Energy consumption increased in the early 2000s, and there were concerns that this increase would continue. But while computing in data centers around the world increased sixfold between 2010 and 2018, energy consumption only increased by six percent.
Several industry analysts point out that a similar trend could well be seen in the case of AI.
“After the mania dies down, other incentives kick in,” said Jonathan Koomey, a former scientist at Lawrence Berkeley National Laboratory who is now an independent researcher. “There is a huge incentive for the industry to become more efficient.”
CONTENT FOR SUBSCRIBERS
Big tech companies are looking for ways to optimize their software, hardware and cooling systems to reduce electricity consumption in their data centers. They have set up computing facilities in northern countries and use cold air drawn in from outside for cooling, in order to reduce electricity and water use. They are also investing in other energy sources.
If they are successful in these projects and if we use AI wisely, they could ultimately get a lot out of this investment.
#artificial #intelligence #ruin #save #planet
