Rachel Feltman: Happy Monday, listeners. For Scientific American’s Science Quickly, I’m Rachel Feltman. You’re listening to our weekly news roundup.
Last Friday marked the 25th anniversary of the 9/11 attacks. Even a quarter century after that one devastating day, people are still dying as a result of the terrorist incursion—and, at least in some cases, as a result of government inaction. The World Trade Center Health Program, which is authorized to provide no-cost medical monitoring and treatment for certain health conditions determined to be connected to the 9/11 attacks, has around 140,000 members. The Centers for Disease Control and Prevention estimates that some 400,000 people were exposed to toxic contaminants, risk of physical injury, and physically and emotionally stressful conditions following the attacks. According to previously undisclosed records released last week, New York City officials told the public that the air around Ground Zero was safe to breathe even when testing showed otherwise.
Mayor Zohran Mamdani announced the release of more than 170,000 pages of documentation at a news conference last Tuesday. In addition to air quality reports, the packet includes internal memos discussing elevated levels of asbestos and other toxins found in environmental testing.
On supporting science journalism
If you’re enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
[CLIP: Zohran Mamdani speaks: “ People got sick because the leaders they trusted lied and told them they were safe to breathe in toxic air. And then when they began to contend with the symptoms of the illnesses that followed, the government that had not hesitated to send them into harm’s way was nowhere to be found.”]
Feltman: The documents are available to the public via a searchable Web portal at sept11documents.cityofnewyork.us. Survivor advocates at the press conference on Tuesday said that this documentation could help jumpstart health research and support the legal claims of those in need of medical support. Mamdani also noted that many people who are eligible for assistance from the WTC Health Program are not registered for it.
[CLIP: Mamdani speaks: “ For so many New Yorkers who lived or worked or studied near Ground Zero, and for the thousands of others who worked on the pile, recovering those we lost and clearing the rubble, that was only the beginning of years of hardship and illness that would follow. Thousands of ironworkers and firefighters and office workers and so many more remember September 11th, not just each time we mark the passage of another year, but as Congressman Nadler once said, ‘Each time they take a struggling breath, each time they have to swallow a pill, each time another chemo dose is administered.’”]
Feltman: If you were below Canal Street in Manhattan or in the Dumbo or Brooklyn Heights neighborhoods of Brooklyn on 9/11 or in the immediate aftermath of the attacks, you can still sign up for medical assistance. You’re also eligible if you worked as a first responder for the attacks outside of NYC. Go to cdc.gov/wtc/apply.html for more information.
Now let’s move on to some AI news. You may have seen headlines last week about artificial intelligence solving one of math’s so-called million-dollar problems. But it turns out this story is actually pretty contentious. Luckily we’ve got just the guy to unpack it for you: Joseph Howlett. Joe is a senior reporter at Scientific American who covers physics, math, astronomy and more. He also happens to have a Ph.D. in particle physics from Columbia, so we’re in great hands. Take it away, Joe!
Joseph Howlett: Hey, Rachel! Last Tuesday was the craziest day in mathematics history, maybe. For the first time in 20 years, someone solved one of the seven Millennium Prize problems, which are, like, the biggest problems in math. If you solve one, you’re supposed to get a million dollars. But this time the person that solved it was an AI or a group of AIs.
It’s called the Navier-Stokes problem. The Navier-Stokes equations govern how fluids move. So, like, any fluid, like air or water, how it swirls around in space, there’s these simple equations that tell you how that’s supposed to work. But those simple equations are really hard to work with. Mathematicians like after 150 years had no idea whether the equations work all the time or whether they sometimes completely fall apart. They call this blowup, and they’ve been looking for a blowup for decades. That’s the millennium problem, to find one or prove that it can’t happen
So about a year ago, two mathematicians got together and started using AI to try to solve this problem. About a month ago, they made really significant progress, but then rumors started to spread. One of these two mathematicians worked for Anthropic. So they were mainly using Anthropic’s Fable AI, but they were also checking its work all the time with OpenAI’s ChatGPT and Codex. So they were going back and forth. About a month ago, they had this breakthrough.
There was also an effort underway to try to solve Navier-Stokes with an internal model at OpenAI. And at some point they heard news of this, and then they redoubled their efforts. The OpenAI model figured out the full problem, Navier-Stokes, in just a couple of days. And it did it with a method that was really similar to what those two mathematicians had been doing for a year. So to them, it seems like maybe the AI, kind of, was referencing their proof that they had been giving it in prompts, right? OpenAI insists that this isn’t what happened, that user data was not accessed, but it’s caused a big stir in the math community. Nobody’s really sure who to give credit to, and everybody’s really angry.
This is a huge deal for math for a couple of reasons. Like, any millennium problem being solved, like, it’s something people have been waiting for for decades, it’s what mathematicians, like, build their whole career trying to accomplish.
If Navier-Stokes were solved by a human, I would be on Science Quickly trying to tell you why math matters, right? That, like, fluids are in our everyday life, they determine how airplanes fly, and we don’t understand them, and understanding them could have all kinds of ramifications for climate modeling and all kinds of stuff like that, right? That’s far more remote. You should care about this because mathematicians are now having a conversation that we all are soon gonna have to have, which is, like: Why do we do what we do? What is the value we get from it? And is the purpose to get as much done as quickly as possible by whatever means? Like, is AI a boon to the human experience, or is it taking something away? And like this is a question mathematicians are being forced to confront today, like, more than any other day and maybe more acutely than we’ve seen any field. But it’s a question that’s coming for all of us.
Feltman: Thanks for that, Joe. Listeners, you can keep up with the latest on this story at scientificamerican.com.
Next up, we’ll check in with another one of our favorite correspondents from the SciAm newsroom: Chief Newsletter Editor Andrea Gawrylewski. She’s here to tell us about the backlash against SpaceXAI’s huge data center, which sits right at the border between Tennessee and Mississippi. Aptly dubbed Colossus 2, it’s an incredibly dense computing center that’s largely being powered by gas-fueled turbines. Thanks for coming on to talk about this, Andrea.
Andrea Gawrylewski: Thanks for having me, Rachel.
The data centers where SpaceXAI trains its chatbot Grok are in Memphis, Tennessee.
SpaceXAI decided to build its Colossus data centers in Memphis.
But the local power grid couldn’t deliver enough electricity on SpaceXAI’s timeline, so the company brought power generation onsite with portable gas turbines rather than waiting for a conventional power plant to spring up.
The temporary power plant for Colossus 2, the second of the data centers, backs up to a residential area in Southaven, Mississippi, which is just across the state line from the data center. So the power is generated in Mississippi and transmitted over the border into Tennessee.
As of a survey in July, the temporary power fleet was up to 69 turbines. And those temporary turbines are now being phased out as the permanent facility comes online.
But as for the people living next door? They say they’re plagued by the exhaust from the turbines and the relentless noise. The sound is like a jet plane revving up on the tarmac but it never takes off.
This is a story popping up around the country.
As AI companies build huge data centers, power becomes a constraint. Data centers now use around 4 to 5 percent of U.S. electricity, and the Electric Power Research Institute projects that that share could reach 9 to 17 percent by 2030.
As for the residents of Southaven, they filed a class action lawsuit alleging lost sleep from the noise and falling property values—the turbines are in their backyards and directly across from their houses.
Meanwhile, a preliminary analysis of air quality that was taken over four months around one Southaven home found that the fine particulate matter concentrations exceeded the EPA’s daily health-protection standard about a third of the time.
Plus, dozens of gas turbines in one place can produce very high levels of nitrogen oxides. Those pollutants can irritate the lungs, worsen asthma, and form ground-level ozone, the main ingredient in smog.
At least 10 schools, three daycares, two libraries, 11 parks and 27 churches sit within two miles of the plant, according to a local advocacy group.
In April the NAACP sued, calling the operation an unpermitted power plant.
But the Department of Justice moved to intervene on SpaceXAI’s behalf and asked the court to dismiss the lawsuit. They said that shutting down the power source for Colossus 2 would threaten AI innovation and national security.
One of the area residents told Scientific American, “It’s so easy for me to see just how our federal government is being leveraged against us”.
Feltman: Thanks, Andrea. Listeners, you can read SciAm’s feature article on Colossus online right now as well as in our October print issue. And don’t forget to subscribe to Today in Science to get stories like this one sent straight to your inbox.
That’s all for this week’s news roundup. Tune in on Wednesday for something special: a feed drop! That’s when one podcast airs an episode from another podcast in their feed. But don’t worry, this isn’t some random show. It’s an episode of Vox’s Unexplainable featuring yours truly along with a few other folks from Scientific American. We can’t wait to share it with you.
Science Quickly is produced by me, Rachel Feltman, along with Fonda Mwangi, Allison Rodgers and Jeff DelViscio. This episode was edited by Alex Sugiura. Marielle Issa and Aaron Shattuck fact-check our show. Our theme music was composed by Dominic Smith. Subscribe to Scientific American for more up-to-date and in-depth science news.
For Scientific American, this is Rachel Feltman. Have a great week!