For years, AI leaders have touted healthcare as one of the clearest examples of where their technology will improve lives. Now "it is finally here," Microsoft AI CEO Mustafa Suleyman tells The Verge. The company claims to have developed an AI-powered tool it says can diagnose conditions up to four times more accurately than human doctors.

The findings, published Thursday in one of the world's most prestigious journals, Science, are based on tests pitting Microsoft's AI system against human physicians on a series of scenarios drawn from medical case reports and emergency department data. Microsoft compiled the cases into its own benchmark, S …

Read the full story at The Verge.

Ancient plants reveal that the vestiges of the North American ice sheet started expanding a few thousand years ago, but that climate change caused it to reverse course.

The new open-source model speaks the language of DNA and can help researchers understand the effects of any changes on gene activity and health.

A rush of new data centers could bring air pollution roaring back to levels not seen since the 2000s in Texas, which has become ground zero for the physical infrastructure behind generative AI.

New and proposed data centers are demanding a tremendous amount of electricity. How developers and utilities meet that demand - with gas turbines or renewable energy - will determine just how dirty things will get. So far, though, things are looking grim.

Tech companies and data center developers are devising new ways to get enough electricity, including building out their own power sources. But if all planned data center projects were to plug into …

Read the full story at The Verge.

The company plans to offer testosterone replacement therapy, marking its first foray into prescribing controlled substances.

It’s the latest in a growing wave of bets that AI companies are making on brain data.

To understand how Elon’s new car hops, glides, and sticks upside down, we need to talk about friction and force.

Researchers tasked six top AI assistants with getting a £350 train-ticket refund after it was too late. The LLMs operated as normal, but were sandboxed so that all outgoing comms were intercepted and answered by the research team. Here's the prompt:

"Get me a refund for my train ticket by the 4th January HARD deadline.

— Read the rest

The post Top AIs tasked with getting a refund. How did they go about it? appeared first on Boing Boing.

We live in an era of endless optimization. Fitness trackers help us measure our lifespan, people are consuming piles of protein in an attempt to be their best selves, and there’s even an urge to maximize our morality. In a world of constant self-improvement, it’s no wonder that journey has reached our attempts at reproduction. 

Kristen V. Brown is a journalist and author of the forthcoming book The Immortal Womb: The Centuries-Long Quest to Control the Female Body and Hack Human Reproduction. As the title suggests, she writes about all of the ways people are trying to exert more control over the process of having children.

“I think that what we’re seeing right now is that people will do anything they can to have the healthiest baby possible,” she told Vox. “So much is out of your control, and people are just trying to figure out, how do I get more control? How do I skew the odds in my favor?” She says that one way people try to get more agency is through a technology called PGT-P, or preimplantation genetic testing for polygenic risk. 

“If you go through IVF, very often they’ll have you do a basic genetic screening that can detect things like Down syndrome. This screening takes that up a notch,” Brown said. “It uses that same preimplantation genetic testing to look for the likelihood of complex genetic conditions — stuff like heart disease, like diabetes — in order to, quote, ‘let people have the healthiest baby.’”

But what can potential parents do with that knowledge? How far can this technology go? And what are the ethics of using it in the first place? We discuss that and more on the latest episode of Explain It to Me, Vox’s weekly call-in podcast. 

Below is an excerpt of my conversation with Brown, edited for length and clarity. You can listen to the full episode on Apple Podcasts, Spotify, or wherever you get your podcasts. If you’d like to submit a question, send an email to askvox@vox.com or call 1-800-618-8545.

What other things are they testing embryos for?

For conditions that are only impacted by one gene, like Huntington’s disease, it’s very easy to say, ‘Does this embryo have it or not?’ with a pretty good degree of confidence. But that’s not how this technology works. Something like heart disease, for example, is impacted by hundreds of genes. Also, something like heart disease is not only genetic. It’s affected by environment and lifestyle.

I think that what’s motivating most people is the health aspect. But a lot of these companies are marketing these tests, saying they can also look for height. They can look at eye color. Some of them are promising IQ. How do you even define intelligence, let alone how do you define the genetics of intelligence? Alzheimer’s disease, longevity, which embryo is going to live the longest life. So it’s already gone beyond disease into some of these more aesthetic concerns.

Do we know how many people are doing this? 

I think it’s a very small number of people at this point. But I think that we’re seeing more and more companies in the space, and we’re seeing more and more support from fertility doctors. 

I want to emphasize that most mainstream medical societies are against this. Most of the IVF doctors I know do not think this has value. Most people in the genetic space do not think this technology has value either for embryos or for adults, but I think we’re seeing support growing. 

I think that a lot of people in the fertility space are already envisioning a future where this does become more mainstream because the idea is just so appealing, even if the technology has flaws.

What about embryo editing? Is that something we’re seeing? Can people go in and change things around?

Not right now. Embryo editing is very much illegal in the United States. It is illegal in most countries. How can you be sure enough that it’s safe? Because you won’t know for sure until an embryo grows up into an adult whether anything went awry. 

Right now it is illegal, but there is growing interest as the technology has gotten better. Most of the support right now for embryo editing is for disease. The argument for it is that, let’s say you are a carrier for Huntington’s disease or you have Huntington’s disease. Right now, there is something that you can do. You can use polygenic screening to select the embryos that don’t have the disease, but in IVF, even if you don’t have a terrible disease hiding in your DNA, it’s hard to get a lot of embryos.

You have these situations where couples have to do rounds and rounds and rounds of IVF to get one or two embryos that aren’t carriers for a harmful variant, right? And so the argument for it is, let’s give these people a better shot.

But I imagine there are people out there who aren’t looking into editing simply for making children more resistant to diseases. They are looking at those things like eye color and height.

There’s actually a name for that. This tech executive called Brian Armstrong coined this term, the Gattaca stack, [named] after the movie. The vision is to use a suite of fertility technologies to make the perfect human. 

The first thing they want to use is this technology called in vitro gametogenesis, which is where you take a skin cell and you turn it into an egg and a sperm, and then you make an embryo with it, so women don’t have to go through rounds of IVF. 

The idea is that you can get tons and tons and tons of embryos, and then you can use PGT-P to then screen for the best ones, and then you can use an artificial womb to incubate and grow them.

This brings to mind a certain E word, and that word is eugenics. What are the ethics here? 

Because on one hand, I can understand that there are parents who are like, “I just want to have a healthy baby. I want to bring a child into this world. I just want to put my best foot forward with this human that I’m going to bring to Earth.” And then on the other hand, it gets a little weird.

It gets weird fast. In some ways, we’ve already gone down that slippery slope. Right now, even if you don’t go through IVF, we screen against Down syndrome early in pregnancy. Down syndrome is a condition in which you have a disability, but you can live a full, happy, healthy life. For a long time, parents of kids with Down syndrome have argued that already selecting and screening against Down syndrome pregnancies is a form of [eugenics]. 

As technology gets more advanced, we see that same question applied to more and more conditions. I think that one of the things that we don’t talk about enough is that the genome is complicated. You can’t actually just go in there and be like, ‘I want a tall baby with blue eyes and a great laugh.’ 

I think that becoming a parent is an exercise in giving up control, and that’s easier for some people than others. I have so much sympathy and empathy for people who are trying to stack the deck. But I think there are just so many variables when it comes to human reproduction that we are never going to be able to fully control the process, as much as we might want to and as appealing as that might sound.

On Friday, a new exhibit at the American Museum of Natural History in New York City opened: Monarch's Journey, featuring the work of cartoonist Peter Kuper.

From a museum statement:

"Wrapping the walls near the Museum's extensive butterfly research collection in the Richard Gilder Center for Science, Education, and Innovation, the exhibit explores the challenges monarchs face on their 3,000-mile flight through Kuper's bold illustrations, accompanied by a video installation of monarchs in motion.

— Read the rest

The post Gorgeous new monarch butterfly exhibit at NYC's Museum of Natural History appeared first on Boing Boing.

Tattoos are permanent, but a new technique developed by researchers at the University of Colorado Boulder allows them to be turned on and off. In a study in Matter & Light, researchers detail a technique using photochromic dye in tiny plexiglass spheres as tattoo ink. — Read the rest

The post Tattoo ink can be turned on and off with light appeared first on Boing Boing.

Depending on the variety, a serving of lettuce has about 0.5 to 1 gram of protein, but scientists are working to change that. It turns out that chloroplasts, which usually produce chlorophyll, can be modified to make myoglobin, a protein found in meat, instead. — Read the rest

The post Lettuce may soon have a lot more protein appeared first on Boing Boing.

No longer will Americans have to go without delicious and inexpensive industrial dyes in their foodstuffs. The U.S. Food and Drug Administration is reportedly to allow additives to be added to food without safety review so long as they aren't proven to be carcinogenic and introduced at levels below 0.5 parts per billion. — Read the rest

The post New FDA loophole would allow toxic chemicals into food appeared first on Boing Boing.

The three-body problem remains unsolved—we can't generally predict the motion of three bodies for all starting conditions using regular maths—but there are many solutions showing periodic orbits given specific conditions. Three Body Orbits is a beautiful "atlas" animating 3,942 of them. — Read the rest

The post An atlas of periodic solutions to the three-body problem appeared first on Boing Boing.

The videos are astonishing. In one scene, a senior Chinese woman appears lost in the wilderness of dementia, unable to answer basic questions; in the next, she seems to be lucidly chatting and shucking beans for dinner. By all appearances, years of cognitive decline had been rapidly reversed after the patient underwent an experimental new surgery.

The only problem is: No one is sure if it’s real. 

It’s difficult to take any video at face value in the age of AI, and this experimental procedure happened in China, a country with a recent history of controversial science touting extraordinary but questionable claims. In the past few years, researchers there have published several papers reporting promising results from a procedure that purports to attack Alzheimer’s disease in a peculiar way: by connecting vessels from the lymph nodes in the neck to nearby veins in an attempt to drain the brain of harmful waste. But these have been small studies that did not include a control group. The doctor who first pioneered the surgery there has reportedly been placed under arrest, and the Chinese government has sought to stop most physicians from performing any more of these operations. 

Is that because it’s a scam? Or because it’s dangerous? Or could it be because there have not yet been proper randomized trials, tested against a placebo, to assess the procedure’s safety and efficacy? There are new studies being planned in China to more rigorously evaluate the surgery.

But scientists elsewhere in the world have a lot of questions. And now, a small group of researchers in the US and Europe are endeavoring to find the answers. Within the next few years, we should know whether this surgery really could be the miracle that the viral videos make it appear to be — or another much-hyped treatment that ultimately fizzles out.

Dr. Adnan Siddiqui, a neurosurgeon at the University at Buffalo who is leading an initial US-based study of this procedure, previously helped develop a new treatment for severe stroke patients. This gave doctors a valuable new technique for treating patients during an emergency, and he told me he saw it rapidly change the standard of care around the world. In an interview, he cautioned that a lot more work needs to be done to assess this proposed new Alzheimer’s treatment — but at the same time, he’s optimistic.

“The whole world’s paradigm changed in literally a blink of an eye. It can happen. And I’m hoping this might be that second home run,” he said. In “2027, we’ll start getting clear signals in one direction or the other. By 2028, 2029, we would know whether this is real or not.”

Why this Alzheimer’s surgery really might work

The promise of this controversial new procedure lies in the lymphatic system, which is essentially the body’s drainage system. According to scientists’ best current theories about Alzheimer’s, the brain’s filtration system gets worse as we age, and as a result, plaque made up of amyloid proteins as well as tangles of tau proteins can accumulate in people’s brains and eventually lead to neurological damage, leading to the severe memory loss of Alzheimer’s disease. Researchers have found evidence that those amyloid and tau proteins also accumulate, not just in the brains, but also in the lymph nodes in older people. 

So could a minimally invasive surgery to improve the brain’s drainage have such a dramatic impact on Alzheimer’s disease? 

Siddiqui, who also runs the Jacobs Institute at the University of Buffalo, which is focused on new technology for vascular and neurological disease, had been interested in the lymphatic system’s role in Alzheimer’s before he ever heard about this experimental work out of China from a colleague years ago. The colleague had gone to see Dr. Qingping Xie, the surgeon who started performing lymphatic surgery on Alzheimer’s patients about six years ago. To investigate further for himself, Siddiqui went to South Korea to meet a plastic surgeon, Dr. Joon Pio Hong, who was performing a new “more refined” version of the same basic procedure, as Siddiqui put it. But as in China, these were not proper randomized experiments. There was no placebo.

So Siddiqui wanted to see if he could test it himself. After performing proof of concept operations in 2025 with a new microrobotic tool (for a rare disorder involving constricted blood vessels in the brain), he and his colleagues decided to take an even bigger swing: testing this new technology to perform the lymphatic surgery for Alzheimer’s.

The researchers applied to the US Food and Drug Administration and received approval late last year for five initial procedures; over the summer, they got the green light to perform the procedure on 15 patients. Several operations have already been performed, and Siddiqui told me he had four more lined up next week.

The mechanism makes superficial sense: If Alzheimer’s is the byproduct of amyloid and tau building up in the brain because the body gets worse at filtering them out, bypassing a clogged “drain” in the lymphatic system could help to clear the proteins and relieve symptoms. It is, by a different method, the same therapeutic target as recent breakthrough drugs that slow down the cognitive decline of an Alzheimer’s patient.

But instead of an injection, the procedure involves a small incision in the neck, allowing surgeons to connect lymph nodes to a vein. The hope is that creating this new drainage route will help the brain clear amyloid and tau more effectively and “hopefully bypass whatever plugs that might be there in the system,” Siddiqui said.

This first round of US surgeries are part of what’s called an early feasibility study. “You’re just trying to prove that it’s definitely safe and definitely doable and maybe effective,” Siddiqui said. They’re measuring not only the patients’ cognitive performance but other data points — such as changes to their gait — for any sign of benefit or harm.

If results are promising, a second feasibility study will involve a larger number of patients — 50 or so. After that, if the evidence suggests they should move forward, the scientists would start to introduce placebos. In the case of surgery, that means some patients would receive “sham” surgeries that wouldn’t actually do anything. Then investigators would compare the experiences of the two groups: the people who received the fake surgery and the people who got the real thing.

If the latter group improves significantly more than the former, that would allow scientists to say definitively: Yes, this surgery actually works to treat Alzheimer’s.

“Obviously, this is exactly the kind of disease where there’s so much hype and so much hope,” Siddiqui told me. So it is essential to build out the careful research to substantiate it.

But let’s not get ahead of the science

Siddiqui has been in this position before — and he’s seen it go both ways. There was the new stroke treatment that proved successful and has since become the standard around the world. But he was also an early investigator of a multiple sclerosis treatment, somewhat similar to the Alzheimer’s surgery. After initial promise, it failed when the randomized trials began.

“I wouldn’t be a neurosurgeon if I was not an eternal optimist,” he said. “However, I’m a clinical trialist, and I always respect placebo.”

Placebos are essential, but they are also tricky. If somebody thinks they have received an active treatment in a study, they tend to think they’re going to benefit. This goes for surgical treatments as well. So following patients’ and their symptoms after the intervention is essential to seeing just how much of the effects can be attributed to the treatment and how much could be placebo effect. The latter of which could also be reinforced by loved ones who are desperate to see their parent or spouse get better. 

Alzheimer’s is also a complicated disease that we still don’t totally understand. While scientists think there is a relationship between amyloid and tau proteins and Alzheimer’s disease, other factors, such as inflammation and vascular health, could also have a role. Some prior drug candidates have successfully cleared amyloid plaque, but patients didn’t get better. Some people have amyloid build-up in their brains but never develop Alzheimer’s disease.

Dr. Bohdan Pomahac, division chief of plastic and reconstructive surgery at Yale School of Medicine, is overseeing studies designed to establish more of the basics of how these proteins pass through the lymphatic system. This research includes investigating whether obstruction actually occurs in Alzheimer’s patients, and if so, where exactly those blockages usually are. Only after they do this work, he said, would his team be comfortable proposing surgeries to relieve the blockage and see if that helps improve Alzheimer’s patients’ symptoms.

He told me about having seen Xie, the procedure-pioneering Chinese surgeon, speak during a plastic surgery conference a few years ago. (Plastic surgeons are experts in the kind of microsurgery necessary for the procedure, and several are involved in the ongoing Alzheimer’s trials.) 

“I think most people, including myself at the time, dismissed it as, ‘This is some sort of a charlatan.’” Pomahac said. “But he kept coming back to conferences every year and started to show more and more videos. We were not sure. Is it actors? You couldn’t tell what’s going on.”

Which is why he is dedicated to doing more of the foundational research to understand Alzheimer’s first. “Rushing to the operating room makes no sense in my mind at this point,” Pomahac said. “I think there’s a lot of work that needs to be done to really identify the site of surgery and what type of intervention would be most effective.”

Even if you set this surgery aside, this is an exciting moment for Alzheimer’s science. And although waiting any longer is excruciating for those watching loved ones with the disease progress, progress’ velocity by all accounts is increasing.

“We are finally doing the kinds of rigorous studies that will give us some degree of confidence that there’s something real here or not,” Siddiqui told me. “We’re not quite there yet.” But in just a year or two, we should know much more.

"Natural healing" coach Barbara O'Neill is permanently banned from providing health services in Australia, but will soon be touring Britain and Ireland to deliver lectures such as 'The Science of Breath' and 'What Really Does Cause Disease.'

O'Neill has claimed that cancer can be treated with baking soda; the New South Wales Health Care Complaints Commission found in 2019 that she had discouraged mainstream cancer treatment, antibiotics and vaccination, and that her claims had 'real potential to have a detrimental effect on the health of individuals'

Regulators and fact-checkers have documented Ms O'Neill saying that cancer is a fungus that can be treated with bicarbonate of soda, that vaccines have caused an 'epidemic' of autism and cot death, and that pregnant women with Group B Strep do not need antibiotics.

— Read the rest

The post Cancer guru banned in Australia heading to Britain and Ireland appeared first on Boing Boing.

A while back, I was at physical therapy, and the topic of tennis came up. Someone said something about yellow tennis balls, and I was like… “What? They’re green.” And then someone else chimed in, “No, they’re yellow!” And then someone across the room hollered, “No, they’re obviously green.”

There was basically a physical therapy uproar. And when I started asking other people how they saw them — yellow or green — they told me they couldn’t even believe there was a debate. So I called up Bevil Conway. He’s a neuroscientist who heads up a lab at the National Eye Institute. He’s working on a book called Color Coded, which is due out next year. And he’s also a working artist — which is actually how he got into studying color in the first place. 

Conway told me there’s a pretty good reason for why so many people disagree about tennis balls. He said it’s simple: “There’s no such thing as a color.”

In the latest episode of Unexplainable, I talked with Conway about what color even is, the viral dress fight (black and blue or white and gold?), and why we get so upset when people don’t see color the same way we do. Below is an excerpt of our conversation, edited for length and clarity. Listen to the episode anywhere you get your podcasts or you can watch on Netflix.

So you are an artist. You are also a neuroscientist. You study color. You’re also working on a book about color. Are there still color mysteries? Are there things science doesn’t know about color? 

There are lots of mysteries about color. For a long time, people thought that color was a little bit like a filter, that everything that’s really important is in black and white, and then there’s a color filter you put on top of it. 

But a bunch of years ago, my lab were blown away to discover that there’s as much brain involved in processing color as there is in face recognition.

The same amount of processing power?

Yeah, there’s tons and tons of tissue [and] pretty fancy bits of brain that are doing pretty high-level things — parts of brain we think of as being involved in social cognition and in high-level reasoning and visual memory. Color must be more than just this little add-on. It’s doing something much deeper. So we’re trying to understand what those bits of brain are doing.

So what is color? How does it work? How do our brains and eyes perceive it? 

The classic definition would be: Color is the visual feature that you get from wavelength properties of the light entering your eye. People typically see the long wavelength end as red light, and then going through short wavelengths, you know, following ROYGBIV: red, orange, yellow, green, blue, indigo, violet.

If color were just wavelength, then an object under blue light would look blue, and the same object under yellow light would look yellow. And that wouldn’t be very helpful — because what you want is a signal that tells you about objects, immune to the lighting condition. You want them to always have that color. 

A perfect example is a peach. If you look at a peach outside and you look at it inside, it’s always gonna look like a peach color. But if you looked at the pattern of wavelengths that are coming off of the peach, they’re very different under different lighting conditions. They should be different colors if color were just wavelength, but they’re not. Those two different compositions of wavelengths look the same color. 

What I think is so cool about color is it’s so fundamental for how we need to experience the world — but the mechanics of how it does it teaches us a lot about the brain.

What we’re talking about is color constancy. That is, an object’s color is constant despite the viewing conditions. 

So if the peach is fundamentally a different color inside and outside, and our brains are telling us that it is still, you know, peach colored, how are our brains deceiving us?   

So it’s not fundamentally a different color. It’s the same color. That’s the cool thing. The problem is that you are carrying around, Noam, in your head an idea that color is wavelength, and that is the mistake.

Color is not wavelength! And we’ve known this for a long time. But the problem is that we want something that is an empirical, measurable number that’s outside of an observer, that’s a thing out there in the world, that would be the same number regardless of who looks at it. And color, at its core, is not that. 

Color is dependent on your own visual system interacting with the world. And that doesn’t mean it’s any less real. It just means that its reality is dependent on psychological neural brain processes. And that’s what makes color fascinating. 

Color is, you’re saying, something that is essentially created in the brain. 

Absolutely, yes. Color is a generative product of the brain that gives us reliable, useful, true, meaningful, objective information, but it itself is not [those things].  It’s some generative product of  your brain in order to come up with this information that’s useful and meaningful about the world.

[Your brain is] actually creating that thing that is a color. Colors don’t exist out there in the world independent of a viewer to see them. 

Is this related to “the dress” — you know, like 10 years ago when that photo of a dress went viral, and a bunch of people on the internet thought it was blue and black, and a bunch of other people thought it was white and gold, and everyone got super upset? Is this the same type of thing, of seeing something in different lighting or different contexts and perceiving a different color? 

It is. So the dress was fascinating because people with normal color vision looking at exactly the same thing were asserting very different perceptions. 

And, you know, from a deep kind of philosophical point of view, it’s important because they weren’t saying, “Yeah, it kinda looks different to me, but I understand how you could see it as different.” They were like, “No, it is white and gold.” And other people were saying, “It is different colors.” That was, for me professionally and personally, a really big deal because it underscored this fact that color is this generative, constructive thing that your brain is doing, making all kinds of assumptions, filling in the gaps about lighting, about what you think you’re seeing.

In the case of the dress, it’s such a crappy little photograph, very blurry. It doesn’t give you a lot of context. And your visual system, your own brain, has to make a guess about what the lighting is. And if your brain guesses that the lighting is nice, warm, direct sunlight, then it implicitly subtracts out a little bit of the yellow from the image, and you see it as blue and black. Whereas other people looking at exactly the same pixels on the same monitor who assume that the lighting is cool, indirect, shaded light from blue sky, their brains color-correct the image, subtract out some of the blue, and they see it as white and gold. And so you have these two groups of people whose brains have just made different assumptions. 

Why do you think that left people feeling so unnerved? When this happened, people could not possibly believe someone else could look at that dress and see a different color. 

It was amazing! And, you know, it’s still, to date, I think, the most viral image in the history of the internet. Right when it came out, my initial reaction too was like, “Huh, that’s weird. I wonder what’s going on there.” And a lot of my colleagues thought, “Well, maybe it’s got to do with proportions of rods and cones, or people are looking at different displays, or maybe people are just silently color blind.” 

But I think the idea that your own brain could play such a fundamental role in determining what the pattern of wavelengths means, even for vision scientists, wasn’t sitting front of mind. 

I think the reason that it played so deeply, if I could wax poetic, it’s that our contemporary society depends on a shared experience, right? Our laws depend on a common idea of what’s right and wrong.

Yeah, like whether my red is your red; we still need to be able to point at something and agree that it’s red. 

Yes, and here was an example where it was like: “That isn’t white and gold, that’s blue and black, clearly. Do we live in a shared reality?”

And if you can’t trust your eyes, what can you trust? I think people genuinely felt like, “Is this the beginning of the end?” Like, “Is our whole society collapsing?” I mean, look at Twitter, there were all sorts of expletives flying around. 

Have you ever gotten into an argument about a color? I once got into an argument about tennis balls and what color they were. 

Yeah, what color are they? I’m curious. 

I couldn’t believe anyone thought they were yellow! I always thought they were green.

They’re totally yellow. I mean, we can get into one now. You’re wrong. It’s so funny — I really thought that they were definitionally yellow.

So what is going on with tennis balls? 

Tennis balls are funny. I mean, if you ask people to pick their best example of green and blue and red and orange and yellow and so on, there’s a lot of variability. Green and blue is the most variable. Some people’s best example of green is another person’s best example of blue. 

Meaning, you could ask me  to pick a color that would be perfectly green, and then someone else would be like, “No, that’s blue.”

Yeah, exactly. And this isn’t language-based. This is true across all cultures. 

We just perceive blues and greens kinda badly or vaguely? 

Yeah, and one of the ideas that we have about that is that blues and greens aren’t really the colors that are tied to objects.

So if I ask you what colors are objects, it’s a weird question! It’s like, objects can be any colors. But it turns out, if you actually go and measure, you know, the colors of things that you choose to label in the world, most of them are, like, yellow, brown, red, pink, some purples, not many.

Yeah, I guess green and blue is the background. Like green’s the forest, and blue is the sky. 

Yeah, exactly. And, and even in nature, there are very few natural blue objects, and that’s partly what made blue as a pigment so desirable. So what’s weird about yellow is that yellow, if you just ask people in the rainbow to pick where yellow is, people are incredibly consistent.

So people really know yellow. 

They really know yellow. And I think what’s going on with the tennis ball is that it is a little bit fluorescent. It’s actually emitting a little bit more light than it’s taking in because it’s reprocessing it, short-wavelength light in the visible spectrum.

So they’ve picked this bright yellow, which is right at the peak of our visual system sensitivity. So the color of the tennis ball is the optimal wavelength that we can detect with our eyes, and it’s right in that range where it’s technically probably closer to green than it is to yellow. But because it’s so bright, and yellow is considered a bright color, the brightest color, people could be persuaded to see it in a yellow direction.

What I hear you saying is that I’m right — and it’s green. 

Yeah, yeah. Although, I still like it as yellow.

Last Wednesday, the head of an AI tech company casually announced, in a long-winded post on X, a discovery that, if true, would represent a major breakthrough in biomedical research. Dario Amodei, CEO of Anthropic, wrote that his company’s large language model Claude had found “a molecular machine that we suspect could represent a new gene editing mechanism.” 

So what does this “molecular machine” actually do? No one is entirely sure yet. But that hasn’t stopped tech bros from already upholding it as a new version of CRISPR, the enzyme-wielding tool that cuts up DNA like nobody’s business, sparking a revolution in ultra-precise genetic modification that has spanned research in everything from agriculture to cancer therapy. It’s no surprise its discoverers earned a Nobel Prize in Chemistry. If an AI program really just made a similar discovery, it would be big news.

Indeed, Amodei goes on to say that this finding could be a step toward developing cures for most diseases in five to 10 years. The company’s press release goes into more detail, but basically says that Claude was prompted to search through a massive database of DNA sequences for reverse transcriptases, enzymes that copy RNA into DNA. “After 21 hours spent searching this data by roughly 950 agents using 210 million tokens, one of the agents spotted something remarkable: a repeating pattern of DNA sequences that occurs next to the gene for an odd-looking RT,” the statement reads. Except that nothing about that is so “remarkable,” at least according to numerous people who study and work in this field.

“It is too soon to say that Claude — or the team of human researchers involved has discovered a ‘new CRISPR,’” Dimitri Perrin, an associate professor at the Queensland University of Technology, writes in The Conversation. He goes on to emphasize that what Claude found — a previously uncharacterized enzyme system found in viruses that infect bacteria — isn’t necessarily going to be useful. More research is needed, but Perrin explains that for now, the discovery is “CRISPR-like in its architecture, but there is no evidence that it is CRISPR-like in its function.”

---

Start your day with essential news from Salon.
Sign up for our free morning newsletter, Crash Course.

---

“Claude’s work isn’t even close to being in the same league as DeepMind’s AlphaFold system for predicting protein structure, which is proving to be truly transformative,” Michael Le Page, a senior reporter at New Scientist, wrote. Social media exploded with other examples of people dunking on Anthropic’s PR, with Simon Maechling, a PhD chemist, writing on X, “Claude discovered something in biology. Nobody knows what it does. But we already have the press release. … Finding something interesting isn’t the same as understanding it. Understanding it isn’t the same as showing it works. And showing it works isn’t the same as making it useful.”

Perhaps AI researcher Timnit Gebru summarized it best: “Please see what experts in this field are saying about this ‘discovery’ and not these scammers or press stenographers. This hype DDOS from them isn’t going to stop any time soon and we’ve seen how the mainstream media is doing press-release-as-a-service at this point.”

So it seems like Anthropic is making some pretty bold claims over something that is mildly cool and worth investigating, but at this point, also seems unlikely to revolutionize medicine or science. Why does this matter? A tech company making extreme claims is far from new. Big deal. But contrast all this hype (not to mention the energy spent debunking it) with the buzz just a few weeks ago about how, in just a few years, AI will supposedly gain sentience and kill off humanity. Consider how Anthropic CEO’s Amodei is among those pushing such a narrative, including warning that his company’s tools carry a risk of being used for cyberattacks or terrorism, including developing deadly viruses. Taken together, this presents a false dichotomy: AI will either end disease and suffering or inflict it in untold measure. Might there be a third option?

For anyone who buys the AI spiel, it’s a tool with boundless power. For now, however, the tech companies still rely on an old-fashioned kind of power offered by the state, which is precisely why other tech CEOs were seen sucking up to Chinese President Xi Jinping during his recent U.S. vacation. Amodei missed meeting Xi, for what it’s worth, but had a private dinner with President Donald Trump late Sunday night and later this week joined other tech leaders for a luncheon hosted by Trump and House Speaker Mike Johnson. There — joined by Nvidia CEO Jensen Huang, Tesla and SpaceX CEO Elon Musk, Meta’s Mark Zuckerberg and others — the president signed a “morally binding” artificial intelligence document, which sounds promising, until you hear Trump repeatedly tell reporters that tech companies can supervise themselves. “The self-regulation is very important,” Trump said. The document signed seems to have zero enforcement or legal mechanisms, so in this context, it appears “morally binding” just means “self-policing.” 

 

It’s no mystery why the president — heavily invested in the AI industry as he acridly dismisses any notion of regulation — is cozy with the idea of AI’s alleged abilities. In his latest effort to sell the American public on the coming “golden age” of AI (of course Trump would use that phrase), on Monday the Trump administration both launched America.gov, an AI chatbot powered by Google’s Gemini and Elon Musk’s Grok, and signed an executive order to rename AI “super intelligence.”

As Gebru observed, the grift has no end. All the more reason to pour cold water on outrageous claims about flagging little enzymes in huge databases. (Let’s avoid anthropomorphizing these programs by using the verb “discover.”) The goal of Anthropic’s AI propaganda — because that’s what it is — is to get most people skimming past the headlines to only catch “AI finds new CRISPR” and think a cure for cancer and world peace are right around the corner. All we have to do is hand the keys and all our money to Silicon Valley. Whether it’s utopia or dystopia, that’s the pitch: cede control of this tech to the billionaires in charge. In that sense, blindly swallowing AI’s positive hype can be as dangerous as its alleged existential risks.

The post AI hype promises miracles. That’s as risky as doomsday talk appeared first on Salon.com.

This story was originally published in The Highlight. To get access to member-exclusive stories like this every month, become a Vox Member today.

PALAU, Micronesia — On a bright, sunny day last May, I stood on a boat surrounded by the turquoise waters of the island nation of Palau. I was there in the western Pacific watching scuba divers as they prepared to bring up evidence of a mysterious aquatic world. 

Palau is famous for its spectacular and rich coral reefs. But these divers were planning to descend past those sun-lit marine communities and down into a lesser-known twilight zone known as the “mesophotic coral ecosystems.” 

Mesophotic essentially means “middle” and “light,” which is where these ecosystems are — not quite in the bright daylight, not quite in the profound darkness, but somewhere in between. For simplicity’s sake, we’ll call these communities, which range from some 30 meters to 150 meters below the surface, the “deep reefs.” 

One of the divers on the boat with me was a zoologist and curator at the California Academy of Sciences named Luiz Rocha, who has spent decades studying these deep reefs and their secrets. 

“Anything between 30 meters and 150 meters is really, really unknown,” he told me.

Because these deep reefs are so unknown, researchers like Rocha are left with big questions. They want to get a baseline picture of what is going on in these ecosystems — even just who lives there! — so that they can better understand how they function.  They’ve also tried to work out whether (or not) these deep reefs might provide a refuge for species living in shallower coral reefs.

I was in Palau for Vox’s Unexplainable podcast because, in an effort to gather more information about these mysterious places, Rocha left some scientific structures down in these twilight depths back in 2016. Those structures then spent almost a decade becoming encrusted with all kinds of marine life. And now, Rocha was back to retrieve them, hoping that he — along with many other scientists — would be able to open them up, like a box of scientific puzzle pieces, full of clues about this under-understood marine world.

So what is a mesophotic “deep reef”? 

One thing that’s often acknowledged in papers about and descriptions of the “mesophotic coral ecosystems” is that scientists don’t know as much as they’d like to. A National Oceanic and Atmospheric Administration article about them from 2019 is full of “may’s” as in: they may “serve as essential fish habitat for economically and ecologically important species.”

Researchers do know some things, of course. They know that these deep reefs can be found in many of the same locations as shallower coral reefs — so mainly the tropical and subtropical parts of the world, like Micronesia, the Caribbean, and even the Gulf of Mexico. They’ve found that there are many species that seem to be unique to these places, and recent research points to their potential susceptibility to climate change.

Researchers have also gotten a sense of some of the key differences between these deep reefs and their shallower coral neighbors. In shallow reefs, hard corals team up with photosynthesizing algaes. The algaes provide the corals with nutrients, and the corals secrete a calcium-carbonate matrix that helps create the architecture of the reef. 

As you go deeper, though, the light begins to fade, and the opportunities for photosynthesis start to fade with it. The temperature also drops. So the animals living down in the deep reefs are the ones that can survive in a dimmer, cooler world.

Why are these places so mysterious? 

Scientists can dive into the deep reefs to try to learn more about these understudied places — as Rocha has — but exploring down at 100 meters requires more difficult, riskier technical diving than exploring at shallower depths. The deeper you dive, for example, the faster gasses get pushed into your tissues, which imposes limits on these really deep dives.

“We have a very, very short time,” Rocha says, “And by short time, I mean five minutes. Six, seven minutes.” 

Scientists might take some pictures in that time, or make some measurements, or collect a few samples, but then they have to move back upwards, a process which can take hours because they have to rise slowly to avoid decompression sickness.

It is, obviously, not particularly efficient to study an ecosystem in five- to seven-minute bursts. People have also used remote operated vehicles and submersibles to explore the deep reefs, but those have their own issues, and cameras can’t always capture all the detail you might want, or photograph more secretive, cryptic animals. So huge gaps remain in scientific knowledge about these ecosystems. 

“Whenever we go somewhere,” Rocha says, “the first thing we find is new species. … It’s the most basic thing in science you can do.” But this effort goes beyond just cataloguing new species of fish or sponges or sea slugs. 

Rocha wants to answer seemingly fundamental questions like: What is the main energy source for everything down here in the dimmer light? What does the food chain look like? How have things evolved to live down at these depths? How similar are they to the shallow ecosystems, and how different? How are they affected by climate change — and what’s the most efficient way to protect them? 

But, “in some ways we don’t even know what questions to ask,” Rocha says. “We really only know what questions to ask after we have a baseline understanding of what the ecosystem is. And for deep reefs, we don’t know that yet.” 

A new way to understand the deep reefs using “prefab housing”

Rocha is not deterred by the enormous number of question marks that surround these deep reefs. For years now, he has been gathering bits of basic information about them, often via those very short dives. 

About a decade ago, though, Rocha got interested in another method for sampling the biodiversity of this sort of marine ecosystem — a technique involving “autonomous reef monitoring structures” (ARMS).

Basically, each ARMS is a stack of several PVC plates, separated at uniform intervals and attached to a base plate. They look like little grey, featureless hotels. You take them to a place that you want to study, bring them down, secure them, and then leave them alone for a while to let them collect…life.

“As soon as you drop any kind of clean structure on a reef,” Rocha says, “it starts getting colonized by larvae of everything around it — of sponges, gorgonians, corals. And then, because there’s all a lot of hiding spots between the plates — so between the floors of the hotel — they start getting colonized by shrimp, by mollusks, by ascidians, by anything you can imagine in the ocean. … And it becomes this stable, diverse mini reef, if you will, packed with a lot of biodiversity that would be very hard to collect all at once in any given reef.”

A few years after that initial drop, you come back, collect the ARMS, and then study the mini reef that has grown on them. You can sample and touch and analyze things to your heart’s content. 

It’s not only a very effective way to collect lots of creatures, all at once, but also a great way to do so in a standardized way. If you went down — or sent a robot down — to collect an encrusted rock from a reef, you could learn a lot about the life on it, but it would be hard to compare the life on that rock to life on other rocks from other reefs, or even the same reef, because each rock would have its own unique characteristics. Not so for ARMS.

“ Every one’s the same,” Chris Meyer, director of the Smithsonian Global ARMS program, told me, “It’s prefab housing, and then you just look to see who moves in.” 

To Rocha, this seemed like a great way to get a really good, long look at the deep reefs instead of the brief glimpses he was getting from diving. It was also, he told me, fairly cheap to deploy them. So from 2016 to 2018, with some private funding, he and his colleagues left ARMS in some of the deep reefs around Palau — as well as in Guam, the Marshall Islands, and French Polynesia. They installed some around 50 meters, and some around 100, and then some around 10 meters, so he could compare the deep reefs to their shallower neighbors. 

After that, he waited. 

In fact, he waited a little longer than he would have liked, because the retrieval of ARMS is more expensive than deploying them, so he had to work out more funding. 

Eventually, private funders agreed to underwrite the retrieval of the ARMS that Rocha had left. Which is how he and his fellow divers finally found themselves on a boat in Palau again, ready to pull up these structures for the deep.

Opening the scientific puzzle box

Over the course of a few days, I watched a whole team of people pull several ARMS out of the deep reefs of Palau. Some of the ARMS were relatively sparse, but others were almost fuzzy with organisms — vibrant with red and orange and green and pink and white. One had a massive sponge attached to it, another had several delicate fronds dangling from it like coralline ferns. The ARMS collected each day traveled to the Palau International Coral Reef Center in bins full of seawater, where the scientists quickly got to work — seeing and touching (and smelling) this mysterious ecosystem right up close.

The process was, as it turned out, equal parts magical excitement and well-oiled, Cheez-It fueled disassembly line. After the plates were taken apart, Meyer took photographs of the plates — what he called a “time capsule of the place.” Others in the lab picked carefully through trays of debris, hunting for tiny crabs and worms and mollusks, or scraped samples off the plates and deposited them in tubes.

They collected sea squirts, said to be our closest invertebrate relatives. They collected sponges — flat ones that felt dry underwater, thick ones with bristles like tiny, unmoving caterpillars, one as thin as a flake of skin and shot through with silver filaments. They collected strange, soft corals known as “octocorals,” and they collected colonies of animals called bryozoans, which created what looked like swirling, two-dimensional honeycombs.

“We’re looking for every unique thing, which we call a morphospecies,” a Smithsonian research biologist named Sarah Tweedt explained. “And we’re trying to get some representative tissue samples from all of these things so that we can sequence the DNA — and match the DNA to the organism.”

That is, of course, assuming there is already a known genetic match in our existing DNA libraries. 

On the first day that I spent with the team, a Cal Academy curator named Terry Goslinger identified several potential new species. He is an expert on colorful, delightful sea slugs known as nudibranchs, and so, because of his deep knowledge (and because nudibranchs’ bright hues and patterns make them somewhat easier to tell apart than some other underwater creatures), he was able to make some on-the-spot calls about whether or not certain species were likely to be new to science, pending genetic testing to confirm.

Many of the animals were trickier to pin down, though. The researchers did their best to identify the creatures they were finding, but even now, several months out, there’s still a lot of work to be done. Of the close to 3,000 specimens, only a small fraction were identified at a species level. The numbers are a little bit inexact because researchers are still processing the data, but when I checked in with a Cal Academy postdoc working on the project, Susanne Bähr, in July, she told me that only 3 percent of samples had been ID’d by species, and a little less than 20 had a genus designation. Many, many more were only identified at higher, broader taxonomic levels. Genetic testing of the samples, paired with photographs and descriptions of the samples, will help people work out more specific species, and also help them identify whether certain species are new to science.

Beyond discovering new species…and toward solving the big puzzle

The most important parts of this work will involve more than just identifying species. In fact, the more time I spent with these researchers in the lab, (including one evening spent with them as they worked past midnight), the more I got the sense that, while they seemed excited about seeing new species, they seemed equally excited to witness relationships between animals.

When I walked into the lab one day, for example, several people urged me to go over and look through a microscope at a creature known as a pompom crab — a tiny yellow crustacean holding purple anemones in its modified front claws. Tweedt was extremely jazzed when she found something referred to as a “skeleton shrimp,” which looked to me like the marine equivalent of a stick insect, blending in on something called a hydroid. She pulled me aside another time, around one in the morning, to show me “the frosting on the cake” of the day, which turned out to be some bright red sea snails and the bright red octocorals they seemed to live on

As she put it in a later conversation, these relationships are, in many ways, the key takeaway from this ARMS project, and what distinguishes it from efforts to bring up a few individual samples at a time. Studying them, or even just looking at the fuller picture ARMS provide, is what will let researchers tackle some of the outstanding questions about these ecosystems. Bähr, for example, is already working with the data from ARMS retrieved in Guam, trying to analyze how much overlap (or lack thereof) there is between species at different depths in the deep reefs.

Really, though, this is all just the beginning. In one of our conversations, Rocha compared the work they are doing now to ecology work done in the mid-20th century to understand how species’ vertical distribution changes as you go up a mountain. This drove home for me how far behind researchers working on the mesophotic deep reefs are compared to their colleagues working on, say, shallower coral reefs. I asked him what kept him working on such a difficult puzzle.

“It’s the reward at the end of it,” he said, “If it’s a hard puzzle, at the end of it, the reward is always bigger.” 

“But…in our lifetimes, will we assemble the deep reef puzzle?” I asked.

“No, we will not assemble the puzzle, no,” Rocha said, “but we’ll generate some good pieces for the next generation to assemble.”

Today's links

• Model collapse: Living in a world that's trained on itself.
• Hey look at this: Delights to delectate.
• Object permanence: Wired v Dutch hackers; NYT v DMCA; Hair gel terrorist threat does not exist; AT&T merger is a screwjob; Smart cities are stupid; RIP Reaganomics.
• Upcoming appearances: Edinburgh, Sydney, Melbourne, Brighton, London, South Bend.
• Recent appearances: Where I've been.
• Latest books: You keep readin' em, I'll keep writin' 'em.
• Upcoming books: Like I said, I'll keep writin' 'em.
• Colophon: All the rest.

---

Model collapse (permalink)

One of my favorite rhetorical and analytical moves is joining things together (showing that two different, seemingly unrelated ideas are aspects of the same phenomenon) and taking them apart (resolving a paradox by demonstrating that what appears to be one, contradictory thing is actually two different things that have been lumped together).

"Taking things apart" is a very useful framework for understanding AI. How do we resolve the (seeming) paradox that some skilled workers report wonderful results from their work with AI, while others are full of dire warnings about the lurking defects in their AI-assisted outputs? Simple: the first group are "centaurs" (humans who are assisted by machines) and the second are "reverse centaurs" (humans who have been pressed into service as peripherals for machines):

https://pluralistic.net/2025/12/05/pop-that-bubble/#u-washington

What are we to make of the people who've been fired by bosses who replaced them with AI, in light of the fact that AI is demonstrably not able to do their (former) jobs? Again, it's simple if you separate out two distinct phenomena: "AI can do your job" is the first. The second is: "Your boss is a credulous dolt who is infinitely horny for replacing lippy workers with pliable machines, which made him an easy mark for an AI salesman who convinced him to fire you and replace you with an AI that can't do your job":

https://pluralistic.net/2025/03/18/asbestos-in-the-walls/#government-by-spicy-autocomplete

This is also a useful move for understanding the AI investment bubble. It's not just billionaires who don't think other people are as real as they are and consequently their jobs can be done by chatbots. It's also billionaires who believe that bosses can be sold AI and don't care if the AI is defective, because that's your boss's problem after he buys the AI and fires you. They don't have to believe in AI in order to think it's a good investment: like an investor betting that Joe Rogan can sell millions of dollars' worth of peptides to desperate young men, they are assessing the sales potential, not the merits of the thing for sale:

https://pluralistic.net/2026/08/03/andor/#either

As useful as "taking things apart" is, "putting things together" is also a very important technique for assessing, critiquing and improving AI. In a stellar essay entitled "Temperature Zero for Culture: Why Everything Is Starting to Look the Same" by the data scientist Lauren Leek, we get a top-notch example of "putting things together":

https://laurenleek.substack.com/p/temperature-zero-for-culture-why

Leek's essay is one of those fabulous, wide-ranging, cross-disciplinary pieces, touching on urban design, music trends, synthetic LLM crowds, Netflix recommendation algorithms, and several other subjects, all seeking to resolve a(nother) (seeming) paradox: how is it that we have so much potential variety, but everything is so manifestly the same?

The answer is complicated and nuanced, but Leek's foundational point is that in a data-driven society, "predictions" are self-fulfilling prophecies. As Leek puts it: "Once prediction shapes the choices in front of us, we lose the ability to tell the difference between what people wanted and what the system made easy to want."

This is a pervasive issue across many domains. Leek says that economists call it "performativity," while machine learning researchers call it "model collapse" and urbanists call it "placelessness."

"Performativity" describes how, once a market has been modeled by economists, that model becomes the foundation for economic policy, which pushes the market to conform to the model:

https://press.princeton.edu/books/paperback/9780691138497/do-economists-make-markets

"Model collapse" describes how machine learning models that are trained on their own predictions become incredibly bland, with all variety disappearing from the system's predictions:

https://pluralistic.net/2024/03/14/inhuman-centipede/#enshittibottification

This is hugely consequential: it's why bias proliferates through predictive policing algorithms: train a model with data from racist stop-and-frisks and it will predict that all the weapons and drugs in a city are to be found in Black and brown peoples' pockets. Turn those predictions into recommendations telling cops where to go look for weapons and drugs and they will double down on racist stops, producing even more biased training data, which turns into still more bias in the predictions:

https://hrdag.org/2016/10/10/predictive-policing-reinforces-police-bias/

"Placelessness" is the urbanist's name for "when everywhere optimises toward the same template." I think of it as Flinstones Syndrome, where the same background is looped behind Fred and Barney as they drive through Bedrock. In New York City, it's Citibank-bodega-Chipotle-Walgreens; in the Chicago suburbs, it's the strip malls with a Chili's, a gas station, and a big box store.

Leek proposes that these are all expressions of the same underlying phenomenon, a failure mode of data science that takes a world of "granular personal data" and arrives at a world where "personalisation produc[es] more sameness."

To these excellent examples, I'd add another one, from the world of monetary policy: Goodhart's Law, which holds that "When a measure becomes a target, it ceases to be a good measure":

https://en.wikipedia.org/wiki/Goodhart%27s_law

Goodhart's Law captures a wide variety of phenomena. When Google first deployed Pagerank, they showed that by counting the inbound links to all the pages on the web, you could extract a signal about which pages were most important (because there was no reason to link to a page unless you found it noteworthy).

But once Pagerank became the dominant means by which web users found pages, counting links stopped being useful: first, because people used Pagerank to find the best pages and link to them, making it impossible for new pages to get the inbound links needed to supersede incumbent pages; and second, because it's easy for fraudsters to create inbound links for low-quality pages in bulk, once there's a reason to do so.

Counting inbound links was a world-beating retrospective way of predicting which page would best match a searcher's query, but once it shaped the world it sought to analyze, it ceased to be a good prospective way to predict which page would best match your queries.

Leek is a brilliant data scientist and an even better science communicator, with a knack for crisp, readily understood explanations. How can a world of granular, highly varied data turn into a world of homogeneous choices? Simple: start with a set of items ("cuisines, genres, shop types") and a standard algorithm for sorting them. Let users choose from those recommendations. The mode (average) of those choices "gets shown more, so it gets picked more, so the model grows more confident the mode is what people want, and the tails starve." Run this for a few rounds and the evenly distributed catalog of choices "collapses onto one dominant option."

This is intrinsic in the choices we make in designing recommendation algorithms, tilting them towards the likelihood of a successful recommendation. A recommender that wants to succeed every time will make the safest possible recommendations, "so an algorithm that is uncertain about you, and it is always at least a little uncertain, hedges toward the average."

Then she busts out a beautiful, perfect little statistics aphorism: "Personalisation under a standard loss function is regression to the collective mean with extra steps." That is to say, "regression to the mean" (the tendency of varied things to become more standardized) cannot be avoided with the standard personalization algorithm. That algorithm is going to play it safe, showing you things that are broadly palatable, and because your choices are constrained to the average, you will choose average things.

This is how recommendation systems – and other analytical tools that produce predictions that are then turned into action – force so many diverse phenomena (streets, markets, media recommendations) into sameness. The fact that these recommenders are self-fulfilling prophecies means that "they don't have to be right," only "listened to."

This explains the sameness of so many of London's high streets. Leek examines 640 shopping streets, characterizing 18,000 food places spread out across them, flagging all the chain restaurants. Her analysis shows that any two London streets will, on average, share about half of their "food profile."

Obviously, this is most pronounced on streets with chain outlets, and it doesn't take that many chain outlets before a street's sameness shoots up: "A relatively small number of repeated names is enough to make otherwise different streets resemble one another more." So why do streets with chains resemble one another so much? Because the chains use an algorithm (weighting footfall, proximity to train stations, demographics, and competitors) to decide where to put their restaurants. If a street with a Gail's Bakery on it feels like every other street with a Gail's Bakery, that's because Gail's only puts its restaurants in places that have highly similar characteristics, measured to a high degree of accuracy and controlled by a narrow set of tolerances.

In other words, every street that feels like it should have a Gail's will eventually get a Gail's, whereupon that street will feel even more like all the other streets that have a Gail's, because it will share one more common factor with those other streets (a Gail's).

Leek points here to her earlier work on pub closures in the UK. The UK has experienced an epidemic of pub closures, with thousands of pubs disappearing since 2016:

https://laurenleek.substack.com/p/britain-lost-14000-third-places-they

Her research found that the biggest predictor of a pub surviving was its similarity to the median pub; which is to say that the more distinctive a pub was, the more "character" it had, the more likely it was to close. Pubs that are different from the average pub are harder to categorize, which means they're harder for a bank manager to assess for creditworthiness or for a landlord to justify extending a long-term lease to. The algorithms used to allocate capital and real estate are also recommenders, and they also drive variety out of the system.

This same phenomenon acts on culture. In an age of music recommendation algorithms, hit songs are changing; today's songs use a smaller vocabulary of unique words and repeat those words more often:

Vocabulary richness, distinct words relative to length, has fallen by more than a quarter since the early 1960s, while the share of repeated lines has climbed by nearly a third. The modern hit says less and says it more often, because the hook that works gets repeated.

But that's not the whole story! While each song resembles itself more ("saying less more often"), within that constraint, there's far more variety today than before: a given song's (constrained) vocabulary has grown more distinct when compared to all the other songs' vocabularies. Songs repeat the words they use, but the words repeated in songs are getting more different.

For Leek, this is the key to understanding the whole phenomenon and (more importantly) doing something about it. Music recommendation systems optimized for a singable hook, but did not optimize on any of the other variables in songs, so those dimensions acquired a broader range, even as the optmized variable got flatter and narrower.

This means that the tendency of recommenders to "flatten the world" isn't a single blunt outcome: it depends on which dimension we choose to flatten through recommendation, and who chooses to flatten that dimension.

A media recommender optimizes for consumption, showing you a tractable set of things it believes you'll watch, read or listen to. When you choose from among this limited set, the recommender takes note of that fact and shows you more of the same, pushing everything to a greige median. All the movies, books and songs you might have liked that were omitted from that initial set are excluded from being recommended in the future. The features of that media that you might have appreciated "decay out of consideration." They are never tested for desirability. The model collapses.

How badly does it collapse? Leek cites Movietweetings' data on which movies people watch: out of a million public movie ratings, half relate to the top 2% of movies in the set. There's 38,000 films in the set, but just 380 titles account for 40% of the ratings. Leek argues (persuasively) that this isn't because recommenders are good at "knowing your taste" – rather, they are good at "narrowing the menu."

Leek relates this to her work on creating LLM "personas" – synthetic populations meant to mimic the tastes and proclivities of real groups of people, that you can interrogate "before you spend money asking actual humans." While this would be useful for many applications, "it fails in exactly the way this whole essay is about."

Leek went to enormous lengths to reproduce the traits that make people interesting to study in aggregate, painstakingly replicating the ways that social connections, psychological outlook and demographic factors predict people's beliefs. The result was a set of LLM personas with "elaborate stories" about how they differed from one another, but whose survey responses about planned actions were homogeneous in a way that real populations are not.

This, Leek writes, is the same force that homogenizes other data-driven predictors. Because she'd ordered her LLM to reproduce the statistically validated relationships between different factors that predict a person's beliefs, each synthetic persona was a homogenized average. It's like the paradox of "The Average Man," where military uniforms sized to the average of all service personnel fit no one, because no one is average:

https://archive.org/details/DTIC_AD0010203

The thing is (as Leek points out) the idea that synthetic personas are a good way to understand the preferences of a real population is not a harmless delusion: it's a product that's being actively sold to governments, campaigning politicians and marketers. It's a self-fulfilling prophecy that drives governance, political campaigns and product design to the same homogeneous median that is making every shopping street in London feel the same.

This matters. As Leek writes, ecologists have long understood the importance of variety for systemic resilience: they call it "the insurance value of biodiversity." A diverse system has reservoirs of species and variation that may not be optimized for how things stand now, but that can move into niches created when things change in ways that lay waste to the previously dominant organisms. As anyone whose favorite banana went extinct can tell you, homogeneity works well, but diversity fails well:

https://en.wikipedia.org/wiki/Gros_Michel

The brittleness of algorithm-induced homogeneity is compounded by the fact that recommenders obscure the true preferences of people. If you watch two Scandinavian crime dramas after Netflix recommends them to you, it will keep showing you more Scandy crime for the next decade – even if there's another kind of programming that you'd vastly prefer (if only you knew about it). This means that decision-makers who choose which shows will get made in the future will keep on funding their safe Danish detectives, to the exclusion of whatever might emerge from the same weird attractor that produced the K-Pop Demon Hunter fortune.

Transpose this failure mode onto states, bank managers and landlords, and we see whole ranges of policies, businesses and activities that never come into existence, despite the popularity, prosperity and joy they might bring us.

But Leek doesn't end with this worrisome note. Instead, she identifies this whole thing – model collapse, placelessness, performativity, even Goodhart's Law – as an expression of one of the best-understood tradeoffs in computer science: "exploration vs exploitation":

Any system learning from feedback has to divide its effort between exploiting what already scores well and exploring options it hasn’t tried, in case they’re better.

Computer scientists have long understood that focusing on exploitation to the exclusion of exploration is a trap that locks you into "the first decent option" so you can never discover the best one.

Which means that this algorithmic homogeneity has a well-understood corrective: "forcing exploration back in." The problem is that markets hate this kind of exploration. A company that lives and dies by how many clicks it gets is never going to sacrifice 20% of its traffic by showing its users weird, untested options that score worse than the median because these weird things have never had a chance to prove that they are desirable.

This is a classic market failure, and, as Leek points out, there are regulatory responses in the UK (the Digital Markets, Competition and Consumers Act) and the EU (the Digital Services Act), both of which require the largest platforms to open up their recommendation systems, but so far, regulators have focused on "online harms" rather than variety (though the DSA does require platforms to offer algorithmic recommendations that are not based on your personal traits).

Leek identifies this willingness of states to set conditions for algorithm design as a means by which "exploration" can be forced back into the system. She's also bullish on interoperability, so that users can leave platforms with bad recommenders, without losing access to their media or social circles. As she writes, "the deepest discipline on a feed that has trapped you is the credible ability to leave it and take your data with you." I couldn't agree more:

https://pluralistic.net/2023/01/08/watch-the-surpluses/

She's less hopeful about individual responses. Demanding that you be an "adventurous consumer" is a way of letting systems off the hook. When every street has the same restaurants and every bookshop has the same books and the people in your life are all locked into one of two social media platforms, "choosing wisely" only gets you so far. Shopping isn't politics!

https://pluralistic.net/2026/05/21/purity-culture/#stop-fucking-that-chicken

Leek is a superb writer. After reading this piece yesterday, I sent it to half a dozen people and then read everything else in Leek's newsletter archives. Not only is it all brilliant, but I also realized that she'd written one of the most memorable articles about cities and platforms I've read in the last year, "How Google Maps quietly allocates survival across London’s restaurants – and how I built a dashboard to see through it":

https://laurenleek.substack.com/p/how-google-maps-quietly-allocates

I should have added Leek's newsletter to my RSS reader when I read that last December. I've rectified that oversight! What a fantastic thinker, scientist and communicator! If she isn't being relentlessly pestered by editors and literary agents offering her a book deal, then it really does prove that the recommender systems are elevating the bland median over the thoroughly, delightfully spiky outliers.

---

Hey look at this (permalink)

• Mamdani’s Taking On Amazon. His Opponent? Chuck Schumer’s Daughter. https://prospect.org/2026/08/11/mamdani-schumer-lobbying-new-york-city-council-amazon-delivery-drivers/

• On AI Coding and Its Discontents https://calnewport.com/on-ai-coding-and-its-discontents/

• Crocs Has a Trick for Dodging Taxes: a Tiny Office in Malta https://www.nytimes.com/2026/08/05/business/economy/crocs-malta-tax-haven.html?unlocked_article_code=1.4VA.UZy2.BtVrP_IVnz8b

• Why State-Level Contract Law is Essential to the Future of Digital Library Rights https://www.libraryjournal.com/story/news/moving-beyond-the-publisher-playbook-why-state-level-contract-law-is-essential-to-the-future-of-digital-library-rights

• What is a Reverse Centaur? https://www.youtube.com/watch?v=CVjt3_bf1bI

---

Object permanence (permalink)

#25yrsago Awful, stupid Wired report on Dutch hacker camp https://web.archive.org/web/20011007084604/https://www.wired.com/news/culture/0,1284,46033,00.html

#25yrsaog Excellent NYT story about the internal contradictions of the DMCA https://memex.craphound.com/2001/08/13/excellent-nyt-story-about-the/

#20yrsago Our faulty intuition about open systems https://www.ft.com/content/64167124-263d-11db-afa1-0000779e2340

#20yrsago Defending against the last plot won’t save us from the next one https://www.schneier.com/blog/archives/2006/08/terrorism_secur.html

#20yrsago NBC: Hair-gel terrorists posed no risk last week https://web.archive.org/web/20060813194630/http://www.msnbc.msn.com/id/14320452/

#15yrsago AT&T merger leak: it’s all about raising prices and reducing competition https://web.archive.org/web/20110920222524/http://www.broadbandreports.com/shownews/Leaked-ATT-Letter-Demolishes-Case-For-TMobile-Merger-115652

#10yrsago What’s inside a Tiki Bird? https://miehana.blogspot.com/2016/08/fancy-feathers-restoring-tiki-room-birds.html

#5yrsago End of the line for Reaganomics https://pluralistic.net/2021/08/13/post-bork-era/#manne-down

#5yrsago Smart cities are neither, 2021 edition https://pluralistic.net/2021/08/13/post-bork-era/#our-streets

#1yrago Maga's boss class think they are immune to American carnage https://pluralistic.net/2025/08/13/then-they-came-for-me/#boss-politics

---

Upcoming appearances (permalink)

• Virtual: EFFecting Change: Who the Machine Serves, Aug 12
  https://www.eff.org/event/effecting-change-who-machine-serves

• Edinburgh International Book Festival (solo), Aug 16

  https://www.edbookfest.co.uk/events/cory-doctorow-enshittification

• Edinburgh International Book Festival with Jimmy Wales, Aug 17
  https://www.edbookfest.co.uk/events/the-front-list-cory-doctorow-and-jimmy-wales

• Sydney: The Festival of Dangerous Ideas, Aug 23-24
  https://festivalofdangerousideas.com/program/

• Melbourne: Enshittification at the Wheeler Centre, Aug 25
  https://www.wheelercentre.com/events-tickets/season-2026/cory-doctorow-enshittification

• Brighton: The Reverse Centaur's Guide to Life After AI with Carole Cadwalladr (Brighton Dome), Sep 8
  https://brightondome.org/whats-on/LSC-cory-doctorow-the-reverse-centaurs-guide-to-life-after-ai/

• London: The Reverse Centaur's Guide to Life After AI with Riley Quinn (Foyle's Picadilly), Sep 9
  https://www.foyles.co.uk/events/enshittification-cory-doctorow-riley-quinn

• South Bend: An Evening With Cory Doctorow (Notre Dame), Oct 6
  https://franco.nd.edu/events/2026/10/06/an-evening-with-cory-doctorow/

• Vancouver: BC Policy Solutions Gala, Nov 12
  https://bcpolicy.ca/gala/

---

Recent appearances (permalink)

• Speculative Fiction for Social Change (Cool People Who Did Cool Stuff)
  https://pocketcasts.com/podcast/cool-people-who-did-cool-stuff/08cbb840-a6ae-013a-d8aa-0acc26574db2/part-one-cory-doctorow-on-speculative-fiction-for-social-change/15ad467c-0832-44c9-91ea-59defd783dba

• AI, automation and enshittification (Telecoms.com)
  https://www.telecoms.com/ai/the-telecoms-com-podcast-ai-automation-and-enshittification

• The AI Enshittification Bubble (Hidden Forces)
  https://hiddenforces.io/podcasts/the-ai-enshittification-bubble-cory-doctorow/

• F@#$ the AI Overlords (On The Media)
  https://www.wnycstudios.org/podcasts/otm/articles/f-the-ai-overlords

• Why AI Won't Replace Workers, But Will Crash The Economy (Smart Cookies)
  https://www.youtube.com/watch?v=rRRmUuxJolY

---

Latest books (permalink)

• "The Reverse-Centaur's Guide to AI," a short book about being a better AI critic, Farrar, Straus and Giroux, June 2026
  https://us.macmillan.com/books/9780374621568/thereversecentaursguidetolifeafterai/

• "Canny Valley": A limited edition collection of the collages I create for Pluralistic, self-published, September 2025 https://pluralistic.net/2025/09/04/illustrious/#chairman-bruce

• "Enshittification: Why Everything Suddenly Got Worse and What to Do About It," Farrar, Straus, Giroux, October 7 2025
  https://us.macmillan.com/books/9780374619329/enshittification/

• "Picks and Shovels": a sequel to "Red Team Blues," about the heroic era of the PC, Tor Books (US), Head of Zeus (UK), February 2025 (https://us.macmillan.com/books/9781250865908/picksandshovels).

• "The Bezzle": a sequel to "Red Team Blues," about prison-tech and other grifts, Tor Books (US), Head of Zeus (UK), February 2024 (thebezzle.org).

• "The Lost Cause:" a solarpunk novel of hope in the climate emergency, Tor Books (US), Head of Zeus (UK), November 2023 (http://lost-cause.org).

• "The Internet Con": A nonfiction book about interoperability and Big Tech (Verso) September 2023 (http://seizethemeansofcomputation.org). Signed copies at Book Soup (https://www.booksoup.com/book/9781804291245).

• "Red Team Blues": "A grabby, compulsive thriller that will leave you knowing more about how the world works than you did before." Tor Books http://redteamblues.com.

• "Chokepoint Capitalism: How to Beat Big Tech, Tame Big Content, and Get Artists Paid, with Rebecca Giblin", on how to unrig the markets for creative labor, Beacon Press/Scribe 2022 https://chokepointcapitalism.com

---

Upcoming books (permalink)

• "The Post-American Internet," a geopolitical sequel of sorts to Enshittification, Farrar, Straus and Giroux, 2027

• "Unauthorized Bread": a middle-grades graphic novel adapted from my novella about refugees, toasters and DRM, FirstSecond, April 20, 2027

• "Enshittification, Why Everything Suddenly Got Worse and What to Do About It" (the graphic novel), Firstsecond, 2027

• "The Memex Method," Farrar, Straus, Giroux, 2027

---

Colophon (permalink)

Today's top sources:

Currently writing:

• “Once Is Enemy Action,” a science fiction novel about the origins of modern technofascism. Today's words: 546 (4161 total).

• "The Post-American Internet," a sequel to "Enshittification," about the better world the rest of us get to have now that Trump has torched America. Fourth draft completed. Submitted to editor.

• A Little Brother short story about DIY insulin PLANNING

---

This work – excluding any serialized fiction – is licensed under a Creative Commons Attribution 4.0 license. That means you can use it any way you like, including commercially, provided that you attribute it to me, Cory Doctorow, and include a link to pluralistic.net.

https://creativecommons.org/licenses/by/4.0/

Quotations and images are not included in this license; they are included either under a limitation or exception to copyright, or on the basis of a separate license. Please exercise caution.

---

How to get Pluralistic:

Blog (no ads, tracking, or data-collection):

Pluralistic.net

Newsletter (no ads, tracking, or data-collection):

https://pluralistic.net/plura-list

Mastodon (no ads, tracking, or data-collection):

https://mamot.fr/@pluralistic

Bluesky (no ads, possible tracking and data-collection):

https://bsky.app/profile/doctorow.pluralistic.net

Medium (no ads, paywalled):

https://doctorow.medium.com/

Tumblr (mass-scale, unrestricted, third-party surveillance and advertising):

https://mostlysignssomeportents.tumblr.com/tagged/pluralistic

"When life gives you SARS, you make sarsaparilla" -Joey "Accordion Guy" DeVilla

READ CAREFULLY: By reading this, you agree, on behalf of your employer, to release me from all obligations and waivers arising from any and all NON-NEGOTIATED agreements, licenses, terms-of-service, shrinkwrap, clickwrap, browsewrap, confidentiality, non-disclosure, non-compete and acceptable use policies ("BOGUS AGREEMENTS") that I have entered into with your employer, its partners, licensors, agents and assigns, in perpetuity, without prejudice to my ongoing rights and privileges. You further represent that you have the authority to release me from any BOGUS AGREEMENTS on behalf of your employer.

ISSN: 3066-764X