Eighty years after bomb, is AI a new arms race in Los Alamos?

Yahoo

2 days ago

When Pat Fitch started working on artificial intelligence in the 1990s, he concluded it was a waste of time.
It now takes up a big chunk of his day.
About two years ago, Fitch, the deputy director of science, technology and engineering at Los Alamos National Laboratory, said he and other top lab officials started to feel a sense of urgency to get ahead of the emerging technology. The lab invested in AI infrastructure, including installing the Venado supercomputer last year. It started carving out millions of dollars for an AI initiative.
'It's gone from being useful to being at a scale and impact that we've got to get all over this,' Fitch said of AI.
Fitch isn't the only one with AI on the brain. When Chris Wright visited J. Robert Oppenheimer's home earlier this year, the newly minted federal energy secretary described the nascent technology as a new Manhattan Project. There are distinct similarities: While AI is being welcomed into the Land of Enchantment due to its potential for positive impacts on research, national security and economic development, it also poses risks — much as the World War II nuclear weapons project did.
'It's hard to overstate the importance and the impact AI will have in defense, in economics, in science,' Wright said at a February news conference in Los Alamos. 'This is just very powerful new tools being developed rapidly. We need to win the second Manhattan Project race, as well.'
022025_GC_EnergySecretary01rgb.jpg (copy)
U.S. Energy Secretary Chris Wright speaks to a small group of reporters at the J. Robert Oppenheimer house in Los Alamos in February 2025. 'It's hard to overstate the importance and the impact AI will have in defense, in economics, in science,' Wright said. 'This is just very powerful new tools being developed rapidly. We need to win the second Manhattan Project race, as well.'
It's a weighty comparison, especially in a state still grappling with the environmental and social impacts of the atomic bomb, which exposed unsuspecting residents of Southern New Mexico to radiation during the Trinity Test explosion 80 years ago — on July 16, 1945 — and caused environmental contamination that required billions of dollars worth of cleanup.
'Promise and peril'
Former President Joe Biden described AI in a 2023 executive order as a technology of 'promise and peril.'
Fitch agreed: 'It's your classic dual-use technology,' he said.
Fitch sees promise in the technology. Already, artificial intelligence has helped with everything from material science to seismic analysis, he said.
'That alone would be a big deal, but maybe it's not a Manhattan Project because it's a whole bunch of little things,' he said. 'If you're that individual scientist working on seismic signatures, the fact that you've never been able to do this before and now you can — that's a huge deal to you in your field, but maybe it's not as earth-shattering as the first atomic weapon or discovery of nuclear energy.'
But uses of AI are expected to grow, Fitch said, and the impacts are not decades away. The potential, in particular, for natural language models of AI — think ChatGPT — will affect every facet of life.
As it was during the Manhattan Project, the nation is in competition with other global powers. Then it was to get an atomic weapon first; now it is to develop the best AI models first.
The Trump administration has repeatedly drawn a link between AI dominance and national security; Fitch said the nation remains a hairsbreadth ahead of other countries when it comes to AI development.
'We've got all the pieces. We don't … have the scale yet,' Fitch said. 'And I think the sense of urgency is growing.'
New Mexico Economic Development Cabinet Secretary Rob Black sees promise, too, in the form of job and revenue growth.
Growing the economies of rural areas is 'one of the more difficult things' his department aims to do.
022425_md_layoffs2.jpg (copy)
Rob Black
'Rural economic development is not easy,' he said.
Black sees an opportunity for artificial intelligence infrastructure to jump-start rural communities, pointing to Los Lunas, which has seen major population growth since Meta built a massive data center in one of the town's business centers.
He also sees an opportunity for New Mexico, which has a high proportion of Spanish speakers, to correct an industry slant toward language models built on the English language.
Black is encouraged by New Mexico State University's June announcement it would offer the state's first bachelor's degree program for artificial intelligence.
But he thinks the state can do more to 'lean in' to develop the AI workforce, which, in turn, can benefit other industries in the state.
022525_GC_GovTechDeal01rgb.jpg (copy)
Gov. Michelle Lujan Grisham and Lanham Napier, chairman of BorderPlex Digital Assets, in February announce a deal for a multibillion-dollar artificial intelligence data hub and advanced manufacturing center. The Santa Teresa site could bring up to 1,000 jobs to the area and will include microgrid power generation and water treatment facilities meant to attract trade-centric businesses as clients.
'You're starting to solve very difficult challenges around pharmaceuticals, around cancer treatments, around climate change, material sciences that traditional computing can't do or hasn't been able to do,' Black said. 'If we're the ones building those technologies and the workforce to utilize them, we will be leading on those areas as well.'
Rising energy needs
Reliant on energy- and water-intensive data centers, artificial intelligence poses an environmental pickle. For New Mexico, the challenges are doubled: Not only is the state obligated to fully transition its utilities to carbon-free energy by 2050 under a 2019 law, but it is also mired in drought and battles over dwindling water supplies.
Data centers are some of the most energy-intensive buildings; the Department of Energy reports they can consume between 10 and 50 times more energy than the typical office building. And a 2018 report estimated data centers nationwide were consuming more than 400 million gallons of water a day.
To ride the AI wave and minimize environmental impacts, Black said New Mexico and its industries must continue investing in clean energy.
'We have to be able to walk and chew gum at the same time,' he said.
That could be a challenge. While the Trump administration has been a bulldog for artificial intelligence, it also has slashed funding for renewable energy and grid reliability.
The Department of Energy issued a recent report projecting more blackouts across the nation unless plans to shutter greenhouse gas giants like coal and natural gas facilities are halted. That's because of surging energy demand driven largely by artificial intelligence and the data centers that support them.
'This report affirms what we already know: The United States cannot afford to continue down the unstable and dangerous path of energy subtraction previous leaders pursued, forcing the closure of baseload power sources like coal and natural gas,' Energy Secretary Wright said in a July 7 statement. 'In the coming years, America's reindustrialization and the AI race will require a significantly larger supply of around-the-clock, reliable and uninterrupted power.'
8B9A0013.jpg (copy)
San Juan Generating Station in August 2024, the day before the scheduled implosion of the four smoke stacks in Waterflow, N.M. The federal Department of Energy issued a recent report projecting more blackouts across the nation unless plans to shutter greenhouse gas giants like coal and natural gas facilities are halted, due to surging energy demand driven largely by artificial intelligence and the data centers that support them.
New Mexico is already in the midst of its own energy transition. The coal-fired San Juan Generating Station near Farmington was shut down in 2022. The Four Corners Power Plant is scheduled to be shuttered next.
According to Data Center Map, an industry tool launched in 2007, there are 21 data centers operating in New Mexico, primarily in the Albuquerque metro area. There are a handful in Northern New Mexico as well.
Camilla Feibelman, director of the Rio Grande Chapter of the Sierra Club, thinks there are sufficient safeguards in place to make sure big energy users don't derail the state's shift to renewable energy.
Geothermal power is a potential win-win for buildings with high energy consumption, Feibelman said. But she said the state needs to ask the 'big questions' before opening the floodgates to AI development — namely, is it all worth it?
'Do those jobs actually appear?' she asked. 'Does the existence of AI end up helping society or does it end up harming society? And does powering these facilities put us at a greater risk of failing to meet the moment on the global climate crisis?'
Given cuts to renewable energy and grid reliability programs at the federal level, she believes the state may have to dig its heels in on making sure the transition actually happens.
'Society hasn't even been willing to say, 'we're willing to do what it takes to curb global climate change,' much less what the role of AI will be in our society and in our attempts to reduce our use of fossil fuels,' Feibelman said.
Workforce and resources
Earlier this year, the New Mexico Artificial Intelligence Consortium — a group made up of representatives of the state's national laboratories and higher education institutions — hosted an industry day.
Jen Gaudioso, director of Sandia National Laboratories' Center for Computing Research and the laboratories' consortium representative, said companies that attended were especially interested in AI education programs in the state.
'We have to both train our workforce into the capacity to utilize the technology, but we also want to be the place that is driving the development and invention of that technology,' Gaudioso said.
The consortium is focused on three pillars: AI workforce development, research applications and infrastructure.
Gaudioso sees potential in AI in materials sciences and resource management. It could help develop materials to replace persistent and hazardous 'forever chemicals,' find abandoned oil wells and bring advances in drug development.
Between the state's research institutions and national laboratories, she said the state has some unique resources for AI.
But there are still gaps.
Broadband access — patchy in rural areas in the state — is an obstacle, Gaudioso said.
And, like Feibelman, she said questions remain, including how much of the infrastructure needed to power AI must be located in the state.
Fitch, at Sandia's northern counterpart, said data centers to power AI don't necessarily have to be located in the same state in which the industry operates. Data centers are typically located near major highways, places where there's an abundance of water and electricity, he said.
'That's not a particularly good description of Northern New Mexico,' Fitch said.
Playing catch-up
State Rep. Christine Chandler, D-Los Alamos, said with two national laboratories, New Mexico is well-positioned to take advantage of artificial intelligence. But when it comes to wrangling the industry, with its high demand for electricity and water use for powerful data centers, the state is far from prepared, she added.
'I think we're behind,' Chandler said. 'There are states across the country now paying attention, and I'd like to see us be a leader in that effort.'
Chandler, who sponsored a bill in this year's legislative session to curb the use of discriminatory algorithms in rejecting or approving New Mexicans for jobs, housing or health care, is part of a group of legislators focusing on the technology.
Regulating AI is difficult work, requiring complicated and technical legislation. Definitions for key terms aren't currently a part of state statute. House Bill 60, the Artificial Intelligence Act, faced opposition from those who said the bill would stymie innovation in the state. Some lawmakers are unfamiliar with the intricacies of the technology.
Ultimately, the legislation failed. But Chandler thinks the state needs to act now, as efforts to resist and reverse regulation on the burgeoning industry crop up around the nation.
A 10-year moratorium on state AI regulations was proposed as part of the nation's budget bill. Although the provision failed — the measure was overwhelmingly rejected in a shockingly bipartisan 99-1 vote — Chandler said it could be part of a trend.
One of his first days in office, President Donald Trump overturned a Biden-era executive order on artificial intelligence that established several principles for 'responsible' AI development.
062325_md_govpresser2.jpg (copy)
Colorado Gov. Jared Polis speaks at the lectern alongside New Mexico Gov. Michelle Lujan Grisham, front left, South Dakota Gov. Larry Rhoden, back left, and Wyoming Gov. Mark Gordon in June 2025 outside of the Georgia O'Keeffe Museum to open the annual Western Governors' Association meeting. Polis signed an act regulating artificial intelligence in his state with 'reservations.'
Closer to home, a Colorado bill, which New Mexico's Artificial Intelligence Act was partially based on, was signed into law with 'reservations' by Colorado Gov. Jared Polis. Despite his signature, the tech entrepreneur and Democratic governor later signaled his support for the proposed moratorium on state artificial intelligence regulation 'to give Congress time to figure this out' and create nationwide standards that would overrule state law.
Chandler is also concerned about pushing a potentially resource-intensive technology while at the same time cutting funding for renewable energy and climate change research.
The National Renewable Energy Laboratory, credited with assisting the Department of Energy in developing report on grid reliability and the growing demand for power, largely driven by AI, is facing a budget less than half the size it was in the current fiscal year.
'The national labs should have a large portfolio of research,' Chandler said. 'My concern is that we are going to narrow our focus and simply focus on what appears to be AI and weapons development, nuclear development and perhaps civilian nuclear development.'
Although Chandler sees the potential for AI, she's not convinced the Manhattan Project is the right analogy. She thinks the urgency of climate change, which poses an 'existential' threat, is a better comparison to the pressures in the 1930s and 1940s that led to the creation of the Manhattan Project.
Facing down a race to develop the nuclear bomb, the government moved quickly, with few safeguards in place to prevent environmental damage, Chandler said. 'We're still kind of reeling' from the consequences, she added.
'There were a lot of unintended consequences as a result,' Chandler said. 'What I fear is all this hype, this excitement around AI could lead us to the same sort of thing. We don't really fully understand the implications of how it could be used or so forth.'