[ad_1]
AI’s relationship with climate change is complicated. The data centers that power AI models require enormous amounts of energy and water. Yet for some, AI’s environmental impacts are considered a part of the necessary cost of technological advancement—especially when AI could help humanity tackle a range of climate challenges down the line, from ecology to natural disasters. But is AI’s energy use and impact on the Earth today worth the cost tomorrow? Will AI ultimately exacerbate an energy crisis—or push humanity toward a renewable future, preserving forests and restoring oceans along the way?Trillion-dollar giants like Microsoft are planting flags in desert soil for new data centers and architecting a $100 billion “supercomputer” with OpenAI expected to have millions of specialized chips for intensive AI computations. However, researchers warn that AI’s insatiable appetite for electricity and resources isn’t sustainable, even if tech firms use renewable energy sources. In an interview, Robert Stoner, MIT researcher, energy council member, and director of the university’s Tata Center for Technology and Design, tells PCMag that tech firms are using up existing electricity sources and taking away from the grid faster than they can add to it.
Meta’s data center in Mesa, Arizona, will be 2.5 million square feet when complete. (Credit: DPR Construction)
To address their energy needs, many tech firms are entering power purchase agreements (PPAs), Stoner says. A PPA is when a third party provides energy—sometimes from renewable sources—to companies like Microsoft and Google.“The problem with that is it’s not necessarily additional,” Stoner tells PCMag. “The companies can make themselves feel good by entering into these sorts of arrangements, [but] they’re really just consuming renewable energy that others would happily consume as well.”Stoner believes that most tech firms are trying to be responsible. He admits that data centers’ “ethical frameworks” can vary, however, and more broadly, rampant energy consumption without adding new sources to the grid is “a big problem.” Because humans are developing artificial intelligence, we have to ultimately decide whether AI accelerates the destruction of our planet—and determine whether its tremendous energy consumption happening now is worth it in the long run.The Power Predicament
Facebook data center (Credit: Meta)
A seemingly obvious solution to this problem is to increase the world’s renewable energy infrastructure so we can meet AI’s energy demands without dumping more carbon into the atmosphere. But that’s easier said than done. Adding new, renewable energy sources in the US is time-consuming and full of red tape. “Utilities really don’t like adding renewable power plants to their systems,” Stoner says. “They stress them, they add a lot of peak current at times of the day that are not the same as they are if the system doesn’t have those sources. They resist, and they hum and they haw.”As a result, it can take four or five years for a new energy source to pass through an “interconnect queue” to be added to existing US power grids, Stoner explains. At the rate AI is developing and investment in the sector is growing, that won’t be enough—despite California’s success with renewables. The Washington Post reports that many states like Texas don’t have enough electricity as it is—and states like Georgia, Virginia, and Arizona will have to scramble to keep up with data centers’ energy demands.Tesla CEO and xAI founder Elon Musk declared last month that there won’t be enough electricity to meet demand as soon as next year. OpenAI CEO Sam Altman agrees and suggests that nuclear fusion breakthroughs could be the only way to truly overcome AI’s energy use problem (nuclear fission plants for AI are in demand, too). Meta CEO Mark Zuckerberg, who’s largely pivoted his tech firm from the metaverse to AI development, even admits that tech companies are more likely to run out of electricity before they run out of money.“Getting energy permitted is like a very heavily regulated government function,” Zuckerberg said in an interview this month. “You’re talking about many years of lead time. So if we wanted to stand up some massive facility, to power that—I think that’s a very long-term project.”Full Steam AheadDespite the looming energy crisis AI aggravates, tech giants show no signs of slowing down development. Some justify the push with vague promises of AGI, or artificial general intelligence, which doesn’t yet exist. Even if it did, though, it’s unclear how helpful it could be in mitigating climate change’s effects.But Chandra Krintz, computer scientist and professor at the University of California, Santa Barbara, tells us that much of the talk around AI right now is being presented in a sensationalized manner.
AI-specific servers could soon suck up 134 terawatt hours of electricity a year on their own—as much as an entire nation.
– UCSB Professor Chandra Krintz
“Right now, I think it is extreme and over-the-top,” says Krintz, who is also the chair of the Computing Community Consortium’s Climate Task Force. “However, I think in the end, it’s going to force technology companies to identify, ‘OK. What is the right hardware? What is the right software?’ There will be pressure from politicians as well as society.”Krintz says it will be up to voters and policy-makers to help prevent the worst predictions about AI’s future energy use from coming true. AI-specific servers could soon suck up 134 terawatt hours of electricity a year on their own—as much as an entire nation. Together, crypto mining and AI could double global data center energy consumption by 2026, according to an International Energy Agency report released earlier this year. And The Week reports that over 300 million people in the US and Canada could soon experience power outages, in part due to AI’s massive energy demands.AI’s Measurement ProblemAnother part of the issue is that we don’t actually know how much of this energy problem can be attributed to AI directly. The industry has struggled to establish effective benchmarks to compare models, and firms aren’t releasing data on their models’ carbon footprints, electricity use, or water consumption. “There is very little transparency,” Krintz says. “Nobody wants to reveal their data. But we cannot solve any of these problems—we cannot understand these problems, we cannot speak about these problems—unless we measure.”In a 2022 paper published on the Association for Computing Machinery’s website, researchers find that cloud computing’s lack of public data makes it difficult for scientists to determine AI’s impact—and propose a model to measure AI’s carbon emissions.“Public pressure needs to be put on these companies to say, ‘You’ve gotta tell us how much energy you’re using, how much water you’re using. What are you taking away in order to achieve the commercial benefits that you’re achieving from us?'” Krintz says.Finding a BalanceIn the short term, Krintz sees AI as a net negative for the environment. But given a longer timeline, she believes there’s plenty of cause for optimism.“It will eat our planet,” Krintz tells PCMag. “However, it has so much potential for offering solutions that help reduce our carbon footprint, that help improve our energy efficiency, that if you think about it holistically … there’s a tremendous potential for having a balance, not destroying the Earth [but] trying to protect it.”
Recommended by Our Editors
That potential isn’t purely theoretical, either. Organizations across the globe are already leveraging artificial intelligence to address some of the most painful symptoms of climate change, like wildfires and coral reef loss.
Networks of AI-powered sensors could help us detect and address widlfires before they get too big (Credit: Anand Purohit)
One such company is Dryad Networks: a tech firm that’s trained AI to “smell” a wildfire’s scent. The company’s small, solar-powered AI sensors are hung from trees every 300 to 650 feet. These are combined with larger, solar mesh network gateways placed approximately every 2 miles to create a distributed fire-detection system that covers an entire forest and can be customized to detect a specific tree’s burn scent, cigarette smoke, or a harmless campfire. “We basically use AI to reliably detect the smell of a fire and distinguish that from the smell of, say, a diesel truck coming by,” Dryad CEO and Co-Founder Carsten Brinkschulte tells PCMag.“We’ve spent thousands of hours training our machine learning model on the different types of smells that you might encounter in the forest,” Brinkschulte says. “We don’t need big data centers to do our computation.” Like all AI applications, Dryad’s tech could deliver false-positive signals, but it’s also possible the solar-powered “noses” could save water by alerting firefighters early enough that they’re able to stop budding flames before they spread.
UK scientists are using AI to track and coral growth and find ways to restore dying reefs (Credit: Jonas Gratzer / Getty Images)
AI could also help address habitat destruction and species loss in our oceans. Sokol Murturi, a researcher at Falmouth University in the UK, uses machine vision to track coral growth. Murturi’s AI-powered system uses a combination of new and existing AI models to help regenerate coral reefs in captivity. It’s designed to provide data on alkalinity, pH levels, and other factors impacting corals. The researcher’s AI tools include visual tracking features like edge detection to indicate whether the coral grows over time. Human scientists can intervene if the AI systems indicate stalled growth or water quality issues.At-risk corals in the wild could be monitored with AI-powered buoys that capture similar data in the future, Murturi posits, adding that corals in areas with sudden temperature spikes could be brought into labs, nurtured, and then reintroduced once temperature surges die down. But if ocean conditions don’t improve, Murturi and other researchers won’t be able to return corals to their homes until fundamental steps are taken to lower rising water temperatures that are fatal to corals worldwide.“My research is a Band-Aid to the problem. What we need now is better forms of generating electricity,” Murturi tells PCMag.Until then, Murturi plans to make a Svalbard Vault of sorts for corals in hopes that one day—when humans have gotten climate change under control and addressed AI’s energy problems—these at-risk organisms could once again thrive in the wild.While AI could help prevent further environmental and ecological damage, ultimately it will still be up to the humans who fund, research, and use AI to oversee its use.“Do I think AI is bad for the environment? It depends on what we’re using it for,” Murturi says. “If we’re using AI to train massive datasets to determine what we should be selling to people, that’s probably not a good use of energy.”
Get Our Best Stories!
Sign up for What’s New Now to get our top stories delivered to your inbox every morning.
This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our Terms of Use and Privacy Policy. You may unsubscribe from the newsletters at any time.
[ad_2]
