The Luddites want to get rid of oil, coal, nuclear power, gas appliances, forced used of solar panels. Now they are going after AI!
“For all its promised benefits, the artificial-intelligence boom is likely to prove costly to public health and even lead to hundreds of deaths a year in the U.S. alone, and those ill effects are likely to disproportionately hit poorer communities, according to a study by California researchers that’s believed to be among the first of its kind.
AI affects public health via air pollution created as a consequence of operating the data centers used to train and run the technology, researchers at UC Riverside and the California Institute of Technology laid out in their report “The Unpaid Toll: Quantifying the Public Health Impact of AI,” published in December.”
These are the folks that tried to stop electricity, manufacturing, innovations and independent action. The good news is that AI will not be stopped. The scare tactics will no longer work.
AI-induced pollution could kill hundreds, cost billions, researchers say
By Troy Wolverton,SF Examiner, 2/11/25 https://www.sfexaminer.com/news/technology/ai-induced-pollution-could-kill-hundreds-cost-billions-researchers-say/article_6449a044-e811-11ef-88d2-473a3ec5a724.html
For all its promised benefits, the artificial-intelligence boom is likely to prove costly to public health and even lead to hundreds of deaths a year in the U.S. alone, and those ill effects are likely to disproportionately hit poorer communities, according to a study by California researchers that’s believed to be among the first of its kind.
AI affects public health via air pollution created as a consequence of operating the data centers used to train and run the technology, researchers at UC Riverside and the California Institute of Technology laid out in their report “The Unpaid Toll: Quantifying the Public Health Impact of AI,” published in December.
Demand for computing power for AI models is prompting the construction of increasing numbers of data centers. The manufacturing of chips and other hardware used in those facilities, the fossil-fuel power plants often used to provide electricity to them, and the typically diesel-powered backup generators used to keep them running during power outages or times of high electricity demand all generate air pollution, the researchers noted in their study.
The researchers said exposure to that air pollution — in the form of particulates, sulfur dioxide and nitrogen dioxide — can lead to maladies such as lung cancer, asthma and cardiovascular disease, and premature deaths. Already disadvantaged communities are likely to absorb much of that pollution.
When it comes to who bears the public health costs of AI, “20% of the people are taking, like, 50% of the burden,” said Shaolei Ren, an associate professor of electrical and computer engineering at UC Riverside and one of the report’s co-authors.
There’s been a lot of focus in recent years on the effects of the AI boom on energy use, climate emissions and water evaporation, noted Alex Hanna, the director of research at the Distributed AI Research Institute. But Hanna said the researchers’ paper is the first she’s seen that offered a comprehensive look at how AI is affecting public health.
“It was a really fascinating read,” she said.
In their study, the researchers primarily focused on the pollution produced as part of generating the electricity needed to power U.S. data centers. In a report last fall, consulting group McKinsey & Company forecast that such facilities would account for 11.7% of the nation’s electricity demand in 2030, up from 3.7% in 2023. AI is going to drive much of that increase, McKinsey said.
To figure out how AI was likely to affect public health, the UC Riverside and Caltech researchers looked at the average pollutant emissions by power plants in different parts of the country. They also examined the emissions by so-called marginal power plants — the electricity generators that are turned on or ramped up in response to increased demand.
By looking at where power is being generated and the size of nearby populations, the researchers estimated the number of adverse health events that would likely be caused by AI-related air pollution. Using established formulas, they then translated those diseases and deaths into economic costs.
Already in 2023, air pollution related to powering U.S. data centers led to 360 premature deaths and total health costs of $5.6 billion across the nation, the researchers estimated. But those figures factored in just the average power-plant pollutant emissions.
When the researchers looked at marginal power-plant emissions, the figures jumped to 490 premature deaths and $7.6 billion of total health costs. The increase is due to the fact that much of the country’s dispatchable power comes from fossil-fuel plants, they noted.
By 2030, because of the expected increase in electricity demand for training and running AI models, data center-related air pollution could result in as many as 1,270 premature deaths and 600,000 cases of asthma nationwide on an annual basis, according to the study. All told, the pollution would cause as much as $21.5 billion per year in health-care costs, the researchers said.
People often associate air pollution with cars, Ren said. But that projected cost from AI-related pollution exceeds the health cost of the pollution generated by cars in California, he said.
Indeed, one of the things that surprised researchers was “the magnitude of the public-health burden” imposed by data centers, he said. It could also be understated, he said.
In December, the Lawrence Berkeley National Laboratory released a new report that estimated that data centers could account for as much as 12% of U.S. energy use two years earlier than McKinsey forecast — due largely to AI. That would mean that the public-health costs of AI would be larger and hit sooner than Ren and his colleagues forecast, Ren said.
Regardless of their size, those costs aren’t evenly shared, according to the report. The health effects of pollution depend greatly on the types of power plants, where they’re located, when the emissions happen and the prevailing wind patterns. Coal-fired plants located close to or upwind from major population centers are likely to result in greater health costs than, say, natural gas-fired ones set in more rural areas.
Taking such factors into account, the researchers determined which counties across the country would see the greatest health costs from AI-related electricity generation air pollution in 2030.
The top 10 counties were spread across West Virginia, Ohio and Pennsylvania, an area that has traditionally relied on coal-fired plants. The median income of residents in those counties ranges from 62% to 88% of the median national income, which was $78,538 per household in 2023, according to the U.S. Census Bureau.
The worst hit would be Mason County, West Virginia. On a per-household basis, the health impact of AI would be an estimated $986 there in 2030, the equivalent of more than seven months of electricity bills, according to the report. The median income in Mason County is 71% of the national figure.
The county was one of five in which the average health costs from AI would top $900 a year by 2030, according to the report.
As part of the study, the researchers took a look at the backup generators used by data centers in three Northern Virginia counties. They found that the pollution those generators produced could affect people as far away as Florida.
Even if the generators were only emitting 10% of the pollutants they are permitted for, each year they’d still be causing 14,000 cases of asthma, as many as 19 deaths and as much as $300 million in total health costs across several jurisdictions, including Virginia; Maryland; Washington, D.C.; Delaware; Pennsylvania; New Jersey; and New York, the researchers found.
Before working on the study, the researchers thought that the air pollution linked to data centers would be limited to the counties where the facilities or the power plants that provided their electricity were located, Ren said. They figured that people who lived far away from them wouldn’t be affected, he said.
“That’s actually not true,” he said. Pollutants, he said, “can travel hundreds of miles.”
Because of that dispersion, areas can be hit with AI-related health costs even if they don’t see any of the economic benefits, the researchers noted.
The paper helps highlight the disparate impact AI will have on health, said Hanna, the AI researcher. There’s been a lot of excitement and hype about the potential for the technology to help find new treatments for diseases.
To the extent that AI lives up to that promise, those treatments are likely to be expensive and only available to those with the best health insurance or the means to pay for treatments not covered by insurance, she said.
Meanwhile, the health cost of training the technology used to find those treatments is already falling disproportionately on people who are more likely to be uninsured or uninsured and living far from hospitals, Hanna said. Those people aren’t likely going to benefit from any of these new AI-produced treatments anytime soon.
Instead, as the paper points out, “these people are going to develop chronic illnesses or face premature death,” she said.
Amid all the hype about AI, there’s been a growing awareness of the downsides of the technology, said Digiconomist founder Alex de Vries, whose online publication focuses on technology’s environmental impact.
But it can be a challenge to catalog all the various damage that technologies such as AI can cause because there are so many dimensions to examine, he said.
“I’m very happy that there are researchers like this apparently thinking about one of these dimensions [of technology harm] and dedicating a whole paper for that,” said de Vries, who is also a PhD candidate at Vrije Universiteit Amsterdam.
But, he said, “there’s still room to improve even further.”
