How to… delete your 23andMe data

MIT Technology Review’s How To series helps you get things done. 

Things aren’t looking good for 23andMe. The consumer DNA testing company recently parted ways with all its board members but CEO Anne Wojcicki over her plans to take the company private. It’s also still dealing with the fallout of a major security breach last October, which saw hackers access the personal data of around 5.5 million customers.

23andMe’s business is built on taking saliva samples from its customers. The DNA from those samples is processed and analyzed in its labs to produce personalized genetic reports detailing a user’s unique health and ancestry. The uncertainty swirling around the company’s future and potential new ownership  has prompted privacy campaigners to urge users to delete their data.

“It’s not just you. If anyone in your family gave their DNA to 23&Me, for all of your sakes, close your/their account now,” Meredith Whittaker, president of the encrypted messaging platform Signal, posted on X after the board’s resignation. 

“Customers should consider current threats to their privacy as well as threats that may exist in the future—some of which may be magnified if 23AndMe were sold to a new owner,” says Jason Kelley, activism director at the Electronic Frontier Foundation. “23AndMe has protections around this much of this. But a potential sale could put your data in the hands of a far less scrupulous company.”

A spokesperson for 23andMe said that the company has strong customer privacy protections in place, and does not share customer data with third parties without customers’ consent. “Our research program is opt-in, requiring customers to go through a separate, informed consent process before joining,” they say. “We are committed to protecting customer data and are consistently focused on maintaining the privacy of our customers. That will not change.”

Why deleting your account comes with a caveat

Deleting your data from 23andMe is permanent and cannot be reversed. But some of that data will be retained to comply with the company’s legal obligations, according to its privacy statement

That means 23andMe and its third-party genotyping laboratory will hang onto some of your genetic information, plus your date of birth and sex—alongside data linked to your account deletion request, including your email address and deletion request identifier. When MIT Technology Review asked 23andMe about the nature of the genetic information it retains, it referred us to its privacy policy but didn’t provide any other details.

Any information you’ve previously provided and consented to being used in 23andMe research projects also cannot be removed from ongoing or completed studies, although it will not be used in any future ones. 

Beyond the laboratories that process the saliva samples, the company does not share customer information with anyone else unless the user has given permission for it to do so, the spokesperson says, including employers, insurance companies, law enforcement agencies, or any public databases.

“We treat law enforcement inquiries, such as a valid subpoena or court order, with the utmost seriousness. We use all legal measures to resist any and all requests in order to protect our customer’s privacy,” the spokesperson says. “To date, we have successfully challenged these requests and have not released any information to law enforcement.”

For those who still want their data deleted, here’s how you go about it.

How to delete your data from 23andMe

  1. Log into your account and navigate to Settings.
  2. Under Settings, scroll to the section titled 23andMe data. Select View.
  3. You may be asked to enter your date of birth for extra security. 
  4. In the next section, you’ll be asked which, if any, personal data you’d like to download from the company (onto a personal, not public, computer). Once you’re finished, scroll to the bottom and select Permanently delete data.
  5. You should then receive an email from 23andMe detailing its account deletion policy and requesting that you confirm your request. Once you confirm you’d like your data to be deleted, the deletion will begin automatically and you’ll immediately lose access to your account. 

What about your genetic sample?

When you set up your 23andMe account, you’re given the option either to have your saliva sample securely destroyed or to have it stored for future testing. If you’ve previously opted to store your sample but now want to delete your 23andMe account, the company says, it will destroy the sample for you as part of the account deletion process.

What if you want to keep your genetic data, just not on 23andMe?

Even if you want your data taken off 23AndMe, there are reasons why you might still want to have it hosted on other DNA sites—for genealogical research, for example. And some people like the idea of having their DNA results stored on more than one database in case something happens to any one company. This is where downloading your data comes into play. FamilyTreeDNA, MyHeritage, GEDmatch, and Living DNA are among the DNA testing companies that allow you to upload existing DNA results from other companies, although Ancestry and 23andMe don’t accept uploads.

How to download your raw genetic data

  1. Navigate directly to you.23andme.com/tools/data/.
  2. Click on your profile name on the top right-hand corner. Then select Resources from the menu.
  3. Select Browse raw genotyping data and then Download.
  4. Visit Account settings and click on View under 23andMe data.
  5. Enter your date of birth for security purposes.
  6. Tick the box indicating that you understand the limitations and risks associated with uploading your information to third-party sites and press Submit request.

23andMe warns its users that uploading their data to other services could put genetic data privacy at risk. For example, bad actors could use someone else’s DNA data to create fake genetic profiles.

They could use these profiles to “match” with a relative and access personal identifying information and specific DNA variants—such as information about any disease risk variants you might carry, the spokesperson says, adding: “This is one reason why we don’t support uploading DNA to 23andMe at this time.” 

Update: This article has been updated to reflect that when asked about the nature of the genetic information it retains, 23andMe referred us to its privacy policy but didn’t provide any other details.

The weeds are winning

On a languid, damp July morning, I meet weed scientist Aaron Hager outside the old Agronomy Seed House at the University of Illinois’ South Farm. In the distance are round barns built in the early 1900s, designed to withstand Midwestern windstorms. The sky is a formless white. It’s the day after a storm system hundreds of miles wide rolled through, churning out 80-mile-per-hour gusts and prompting dozens of tornado watches and sirens reminiscent of a Cold War bomb drill.

On about 23 million acres, or roughly two-thirds of the state, farmers grow corn and soybeans, with a smattering of wheat. They generally spray virtually every acre with herbicides, says Hager, who was raised on a farm in Illinois. But these chemicals, which allow one plant species to live unbothered across inconceivably vast spaces, are no longer stopping all the weeds from growing.

Since the 1980s, more and more plants have evolved to become immune to the biochemical mechanisms that herbicides leverage to kill them. This herbicidal resistance threatens to decrease yields—out-of-control weeds can reduce them by 50% or more, and extreme cases can wipe out whole fields. 

At worst, it can even drive farmers out of business. It’s the agricultural equivalent of antibiotic resistance, and it keeps getting worse.

As we drive east from the campus in Champaign-Urbana, the twin cities where I grew up, we spot a soybean field overgrown with dark-green, spiky plants that rise to chest height. 

“So here’s the problem,” Hager says. “That’s all water hemp right there. My guess is it’s been sprayed at least once, if not more than once.”

“With these herbicide-resistant weeds, it’s only going to get worse. It’s going to blow up.”

Water hemp (Amaranthus tuberculatus), which can infest just about any kind of crop field, grows an inch or more a day, and females of the species can easily produce hundreds of thousands of seeds. Native to the Midwest, it has burst forth in much greater abundance over the last few years, because it has become resistant to seven different classes of herbicides. Season-long competition from water hemp can reduce soybean yields by 44% and corn yields by 15%, according to Purdue University Extension.

Most farmers are still making do. Two different groups of herbicides still usually work against water hemp. But cases of resistance to both are cropping up more and more.

“We’re starting to see failures,” says Kevin Bradley, a plant scientist at the University of Missouri who studies weed management. “We could be in a dangerous situation, for sure.”

Elsewhere, the situation is even more grim.

“We really need a fundamental change in weed control, and we need it quick, ’cause the weeds have caught up to us,” says Larry Steckel, a professor of plant sciences at the University of Tennessee. “It’s come to a pretty critical point.” 

On the rise

According to Ian Heap, a weed scientist who runs the International Herbicide-Resistant Weed Database, there have been well over 500 unique cases of the phenomenon in 273 weed species and counting. Weeds have evolved resistance to 168 different herbicides and 21 of the 31 known “modes of action,” which means the specific biochemical target or pathway a chemical is designed to disrupt. Some modes of action are shared by many herbicides.

One of the most wicked weeds in the South, one that plagues Steckel and his colleagues, is a rhubarb-red-stemmed cousin to water hemp known as Palmer amaranth (Amaranthus palmeri). Populations of the weeds have been found that are impervious to nine different classes of herbicides. The plant can grow more than two inches a day to reach eight feet in height and dominate entire fields. Originally from the desert Southwest, it boasts a sturdy root system and can withstand droughts. If rainy weather or your daughter’s wedding prevents you from spraying it for a couple of days, you’ve probably missed your chance to control it chemically.  

Palmer amaranth “will zero your yield out,” Hager says.

Several other weeds, including Italian ryegrass and a tumbleweed called kochia, are inflicting real pain on the farmers in the South and the West, particularly in wheat and sugar beet fields.   

Chemical birth 

Before World War II, farmers generally used cultivators such as plows and harrows to remove weeds and break up the ground. Or they did it by hand—like my mother, who remembers hoeing weeds in cornfields as a kid growing up on an Indiana farm.

That changed with the advent of synthetic pesticides and herbicides, which farmers started using in the 1950s. By the 1970s, some of the first examples of resistance appeared. By the early 1980s, Heap and his colleague Stephen Powles had discovered populations of ryegrass (Lolium rigidum) that were resistant to the most commonly used herbicides, known as ACCase inhibitors, spreading throughout southern Australia. Within a few years, this species had become resistant to yet another class, called ALS-inhibiting herbicides.  

The problem had just begun. It was about to get much worse.

In the mid to late 1990s, the agricultural giant Monsanto—now a part of Bayer Crop Science—began marketing genetically engineered crops including corn and soybeans that were resistant to the commercial weed killer Roundup, the active ingredient of which is called glyphosate. Monsanto portrayed these “Roundup-ready” crops, and the ability to spray whole fields with glyphosate, as a virtual silver bullet for weed control.

Glyphosate quickly became one of the most widely used agricultural chemicals, and it remains so today. It was so successful, in fact, that research and development on other new herbicides withered: No major commercial herbicide appears likely to hit the market anytime soon that could help address herbicide resistance on a grand scale. 

Monsanto claimed it was “highly unlikely” that glyphosate-resistant weeds would become a problem. There were, of course, those who correctly predicted that such a thing was inevitable—among them Jonathan Gressel, a professor emeritus at the Weizmann Institute of Science in Rehovot, Israel, who has been studying herbicides since the 1960s.

Stanley Culpepper, a weed scientist at the University of Georgia, confirmed the first case of Roundup resistance in Palmer amaranth in 2004. Resistance rapidly spread. Both Palmer amaranth and water hemp produce male and female plants, the former of which produce pollen that can blow long distances on the wind to pollinate the latter. This also gives the plant a lot of genetic diversity, which allows it to evolve faster—all the better for herbicide resistance to develop and spread. These super-weeds sowed chaos throughout the state.

“It devastated us,” Culpepper says, recalling the period from 2008 to 2012 as particularly difficult. “We were mowing fields down.”  

Staying alive

Herbicide resistance is a predictable ­outcome of evolution, explains Patrick Tranel, a leader in the field of molecular weed science at the University of Illinois, whose lab is a few miles from the South Farm. 

“When you try to kill something, what does it do? It tries to not be killed,” Tranel says. 

Weeds have developed surprising ways to get around chemical control. One 2009 study published in the Proceedings of the National Academy of Sciences showed that a mutation in the Palmer amaranth genome allowed the plant to make more than 150 copies of the gene that glyphosate targets. That kind of gene amplification had never been reported in plants before, says Franck Dayan, a weed scientist at Colorado State University.

Another bizarre way resistance can arise in that species is via structures called extrachromosomal circular DNA, strands of genetic material including the gene target for glyphosate that exist outside of nuclear chromosomes. This gene can be transferred via wind-blown pollen from plants with this adaptation. 

But scientists are increasingly finding metabolic resistance in weeds, where plants have evolved mechanisms to break down just about any foreign substance—including a range of herbicides. 

Let’s say a given herbicide worked on a population of water hemp one year. If any plants “escape,” or survive, and make seeds, their offspring could possess metabolic resistance to the herbicides used. 

“When you try to kill something, what does it do? It tries to not be killed.”

Patrick Tranel, University of Illinois

There’s evidence of resistance developing to both of the chemical groups that have replaced or been mixed with Roundup to kill this weed: an herbicide called glufosinate and a pair of substances known as 2,4-D and dicamba. These two would normally kill many crops, too, but there are now millions of acres of corn and soy genetically modified to be impervious. So essentially the response has been to throw more chemicals at the problem.

“If it worked last year, if you have metabolic resistance there’s no guarantee it’s going to work this year,” Hager says. 

Many of these herbicides can harm the environment and have the potential to harm human health, says Nathan Donley, the environmental health science director at the Center for Biological Diversity, which is based in Tucson, Arizona. Paraquat, for example, is a neurotoxic chemical banned in more than 60 countries (it’s been linked to conditions like Parkinson’s), Donley says, but it’s being used more and more in the United States. 2,4-D, one of the active ingredients in Agent Orange, is a potential endocrine disruptor, and exposure to it is correlated with increased risk of various cancers. Glyphosate is listed as a probable human carcinogen by an agency within the World Health Organization and has been the subject of tens of thousands of lawsuits worth tens of billions. Atrazine can stick around in groundwater for years and can shrink testicles and reduce sperm count in certain fish, amphibians, reptiles, and mammals.

Replacing glyphosate with herbicides like 2,4-D and dicamba, which are generally more toxic, “is definitely a step in the wrong direction,” Donley says. 

Looking for solutions

It’s not just chemicals. Weeds can become resistant to any type of control method. In a classic example from China, a weed called barnyard grass evolved over centuries to resemble rice and thus evade hand weeding.

Because weeds can evolve relatively quickly, researchers recommend a wide diversity of control tactics. Mixing two herbicides with different modes of action can sometimes work, though that’s not the best for the environment or the farmer’s wallet, Tranel says. Rotating the plants that are grown helps, as does installing winter cover crops and, above all, not using the same herbicide in the same way every year. 

Fundamentally, the solution is to “not focus solely on herbicides for weed management,” says Micheal Owen, a weed scientist and emeritus professor at Iowa State University. And that presents a “major, major issue for the farmer” and the current state of American farms, he adds. 

weeds

BELL HUTLEY

Farms have ballooned in size over the last couple of decades, as a result of rural flight, labor costs, and the advent of chemicals and genetically modified crops that allowed farmers to quickly apply herbicides over massive areas to control weeds. This has led to a kind of sinister simplification in terms of crop diversity, weed control practices, and the like. And the weeds have adjusted. 

On the one hand, it’s understandable that farmers often do the cheapest thing they can to control weeds, to get them through the year. But resistance is a medium- to long-term problem running up against a system of short-term thinking and incentives, says Katie Dentzman, a rural sociologist also at Iowa State University.

Her studies have shown that farmers are generally informed and worried about herbicide resistance but are constrained by a variety of factors that prevent them from really heading it off. The farm is too big to economically control weeds without spraying in a single shot, some farmers say, while others lack the labor, financing, or time. 

Agriculture needs to embrace a diversity of weed control practices, Owen says. But that’s much easier said than done. 

“We’re too narrow-visioned, focusing on herbicides as the solution,” says Steven Fennimore, a weed scientist with the University of California, Davis, based in Salinas, California.

Fennimore specializes in vegetables, for which there are few herbicide options, and there are fewer still for organic growers. So innovation is necessary. He developed a prototype that injects steam into the ground, killing weeds within several inches of the entry point. This has proved around 90% effective, and he’s used it in fields growing lettuce, carrots, and onions. But it is not exactly quick: It takes two or three days to treat a 10-acre block.

Many other nonchemical means of control are gaining traction in vegetables and other high-value crops. Eventually, if the economics and logistics work out, these could catch on in row crops, those planted in rows that can be tilled by machinery. 

A company called Carbon Robotics, for example, produces an AI-driven system called the LaserWeeder that, as the name implies, uses lasers to kill weeds. It is designed to pilot itself up and down crop rows, recognizing unwanted plants and vaporizing them with one of its 30 lasers. LaserWeeders are now active in at least 17 states, according to the company.  

You can also shock weeds by using electricity, and several apparatuses designed to do so are commercially available in the United States and Europe. A typical design involves the use of a height-adjustable copper boom that zaps weeds it touches. The most obvious downside with this method is that the weeds usually have to be taller than the crop. By the time the weeds have grown that high, they’ve probably already caused a decline in yield. 

Weed seed destructors are another promising option. These devices, commonly used in Australia and catching on a bit in places like the Pacific Northwest, grind up and kill the seeds of weeds as wheat is harvested.

An Israeli company called WeedOut hatched a system to irradiate and sterilize the pollen of Palmer amaranth plants and then release it into fields. This way, female plants receive the sterile pollen and fail to produce viable seeds. 

“I’m very excited about this [as] a long-term way to reduce the seed bank and to manage these weeds without having to spray an herbicide,” Owen says. 

WeedOut is currently testing its approach in corn, soybean, and sugar beet fields in the US and working to get EPA approval. It recently secured $8 million in funding to scale up. 

In general, AI-driven rigs and precision spraying are very likely to eventually reduce herbicide use, says Stephen Duke, who studies herbicides at the University of Mississippi: “Eventually I expect we’ll see robotic weeding and AI-driven spray rigs taking over.” But he expects that to take a while on crops like soybeans and corn, since it is economically difficult to invest a lot of money in tending such “low-value” agronomic crops planted across such vast areas.

A handful of startups are pursuing new types of herbicides, based on natural products found in fungi or used by plants to compete with one another. But none of these promise to be ready for market anytime soon.

Field day 

Some of the most successful tools for preventing resistance are not exactly high-tech. That much is clear from the presentations at the Aurora Farm Field Day, organized by Cornell University just north of its campus in Ithaca, New York. 

For example, one of the most important things farmers can do to prevent the spread of weed seeds is to clean out their combines after harvest, especially if they’re buying or using equipment from another state, says Lynn Sosnoskie, an assistant professor and weed scientist at Cornell. 

Combines are believed to have already introduced Palmer amaranth into the state, she says—there are now at least five populations in New York. 

Another classic approach is crop rotation—switching between crops with different life cycles, management practices, and growth patterns is a mainstay of agriculture, and it helps prevent weeds from becoming accustomed to one cropping system. Yet another option is to put in a winter cover crop that helps prevent weeds from getting established. 

“We’re not going to solve weed problems with chemicals alone,” Sosnoskie says. That means we have to start pursuing these kinds of straightforward practices.

It’s an especially important point to hammer home in places like New York state, where the problem isn’t yet top of mind. That’s in part because the state isn’t dominated by monocultures the way the Midwest is, and it has a more diverse patchwork of land use. 

But it’s not immune to the issue. Resistance has arrived and threatens to “blow up,” says Vipan Kumar, also a weed expert at Cornell.

“We have to do everything we can to prevent this,” Kumar says. “My role is to educate people that this is coming, and we have to be ready.”

Douglas Main is a journalist and former senior editor and writer at National Geographic.

Everything comes back to climate tech. Here’s what to watch for next.

This article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here.

We get to celebrate a very special birthday today—The Spark just turned two! 

Over the past couple of years, I’ve been bringing you all the news you need to know in climate tech and digging into some of the most fascinating and thorny topics from energy and transportation to agriculture and policy. 

In light of this milestone, I’ve been looking back at some of the most popular editions of this newsletter, as well as some of my personal favorites—and it’s all got me thinking about where climate tech will go next. So let’s look back together, and I’ll also share what I’m going to be watching out for as we go forward.

It’s prime time for batteries

It will probably be a surprise to absolutely nobody that the past two years have been filled with battery news. (In case you’re new and need a quick intro to my feelings on the topic, you can read the love letter to batteries I wrote this year for Valentine’s Day.) 

We’ve covered how abundant materials could help unlock cheaper, better batteries, and how new designs could help boost charging speeds. I’ve dug into the data to share how quickly batteries are taking over the world, and how much faster we’ll need to go to hit our climate goals.

The next few years are going to be make-or-break for a lot of the alternative batteries we’ve covered here, from sodium-ion to iron-air and even solid-state. We could see companies either fold or make it to the next stage of commercialization. I’m watching to see which technologies will win—there are many different options that could break out and succeed. 

A nuclear renaissance 

One topic I’ve been covering closely, especially in the past year, is nuclear energy. We need zero-emissions options that are able to generate electricity 24-7. Nuclear fits that bill. 

Over the past two years, we’ve seen some major ups and downs in the industry. Two new reactors have come online in the US, though they were years late and billions over budget. Germany completed its move away from nuclear energy, opting instead to go all in on intermittent renewables like solar and wind (and keep its coal plants open). 

Looking ahead, though, there are signs that we could see a nuclear energy resurgence. I’ve written about interest in keeping older reactors online for longer and opening up plants that have previously shut down. And companies are aiming to deploy new advanced reactor designs, too. 

I’m watching to see how creative the industry can get with squeezing everything it can out of existing assets. But I’m especially interested to see whether new technologies keep making progress on getting regulatory approval, and whether the new designs can actually get built. 

Material world forever

I’ll never stop talking about materials—from what we need to build all the technologies that are crucial for addressing climate change to how we can more smartly use the waste after those products reach the end of their lifetime. 

Recently, I wrote a feature story (and, of course, a related newsletter bringing you behind the scenes of my reporting) about how one rare earth metal gives us a look at some of the challenges we’ll face with sourcing and recycling materials over the next century and beyond. 

It’s fitting that the very first edition of The Spark was about my trip inside a battery recycling factory. Over the past two years, the world of climate tech has become much more tuned in to topics like mining, recycling, and critical minerals. I’m interested to see how companies continue finding new, creative ways to get what they need to build everything they’re trying to deploy. 

Milestones … and deadlines

Overall, the last couple of years have been some of the most exciting and crucial in the race to address climate change, and it’s only going to ramp up from here. 

Next year marks 10 years since the Paris Agreement, a landmark climate treaty that’s guided most of the world’s ambitions to limit warming to less than 2 °C (3.7 °F) above preindustrial levels. In the US, 2027 will mark five years since the Inflation Reduction Act was passed, ushering in a new era of climate spending for the world’s largest economy. 

The last two years have been a whirlwind of new ideas, research, and technologies, all aimed at limiting the most damaging effects of our changing climate. I’m looking forward to following all the progress of the years to come with you as well. 


Now read the rest of The Spark

Another thing

If you’re reading this, I’m willing to bet that you probably eat food. So you should join us for the latest edition of our subscriber-only Roundtables virtual event series, where I’ll be speaking with my colleague James Temple about creating climate-friendly food. 

Joining us are experts from Pivot Bio and Rumin8, two of our 2024 Climate Tech Companies to Watch. It’s going to be a fascinating discussion—subscribers, register to join us here

And one more 

The growing energy demands of artificial intelligence represent a challenge for the grid. But the technology also offers an opportunity for energy tech, according to the authors of a new op-ed out this week. Check it out for more on why they say that AI and clean energy need each other

Keeping up with climate  

Hurricane Milton reached wind speeds of over 160 miles per hour, making it a Category 5 storm. It’s hitting the gulf coast of Florida in the coming days. See its projected path and the rainfall forecast. (Washington Post
→ Tampa Bay has seen destructive hurricanes, but there hasn’t been a direct hit in decades. The metro area is home to over 3 million people. (Axios)

Other regions are still reeling from Hurricane Helene, which dumped rainfall in western North Carolina in particular. The storm upends ideas of what a climate haven is. (Scientific American)
→ Two studies suggest that climate change significantly boosted rainfall from the storm. (NBC News)

If you have an EV, it’s best to keep it out of flood zones during hurricanes when possible. Batteries submerged in salt water can catch fire, though experts say it’s relatively rare. (New York Times)

The risk of winter blackouts in Great Britain is at the lowest in years, even though the country has shut down its last coal plant. The grid is expected to have plenty of energy, in part because of investment in renewables. (The Guardian)

Voters in Kazakhstan have approved a plan to build the country’s first nuclear power plant. The country has a complicated relationship with nuclear technology, since it was a testing ground for Soviet nuclear weapons. (Power

Revoy wants to bring battery swapping to heavy-duty trucks. The company’s batteries can reduce the amount of diesel fuel a conventional truck needs to drive a route. (Heatmap)
→ I wrote earlier this year about another company building batteries into trailers in an effort to clean up distance trucking. (MIT Technology Review)

These are the best ways to measure your body fat

This article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here.

This week, an office conversation turned to body weight. We all know that being overweight is not great for your health—it’s linked to metabolic diseases like diabetes and cardiovascular problems. But weighing yourself won’t tell you all you need to know about your disease risk.

A friend of mine is a super-fit marathon runner. She’s all lean muscle. And yet according to her body mass index (BMI), which is a measure of weight relative to height, she’s overweight. Which is frankly ridiculous.

I, on the other hand, have never been all that muscular. I like to think I’m a healthy weight—but nurses in the past have advised me, on the basis of my BMI, to eat more butter and doughnuts. This is advice I never expected to receive from a health professional. (I should add here that my friend and I are roughly the same height and wear the same size in clothes.)

The BMI is flawed. So what should we be using instead? There are several high-tech alternatives, but a simple measure that involves lying on your back could also tell you about how your body size might influence your health.

First, let’s talk about fat—the most demonized of all body components. Fat is stored in adipose tissue, which has some really important functions. It stores energy, keeps us warm, and provides protective cushioning for our organs. It also produces a whole host of important substances, from hormones that control our appetite to chemicals that influence the way our immune systems work.

Not all fat is equal, either. Our bodies contain white fat, brown fat, and beige fat. While white fat stores energy, brown fat helps burn calories. Beige fat tissue contains a mixture of the two. And white fat can also be broken down into two additional categories: the type under your skin is different from that which covers your internal organs.

It’s the visceral fat—the type surrounding your organs—that is thought to be more harmful to your health, if there’s too much of it. Having more visceral fat has been linked to an increased risk of diabetes and cardiovascular disease. (That relationship isn’t straightforward either, though; studies have shown that removing this “excess” fat doesn’t improve metabolic health.)

Either way, having a good idea of how much fat is in your body, and where it is, would be valuable. It might at least give us some idea of our risk of metabolic disorders. There are quite a few different ways of measuring this.

BMI is the most widely adopted. It’s the official measure the World Health Organization uses to define overweight and obesity. On the plus side, it’s very easy to calculate your BMI. Unfortunately, it doesn’t tell you very much about the fat in your body or how it corresponds to your health. After all, your body weight includes your bones, muscles, blood, and everything else, not just your fat. (And as we’ve seen, it can lead well-meaning health practitioners to recommend weight loss or weight gain when it’s really not appropriate.)

Scanners that can specifically measure fat are more useful here. Typically, doctors can use a DEXA scan, which relies on x-rays, to give an idea of where and how much body fat a person has. CT scanners (which also makes use of x-rays) and MRI scanners (which use magnets) can give similar information. The problem is that these are not all that convenient—they’re expensive and require a hospital visit. Not only that, but standard equipment can’t accommodate people with severe obesity, and people with some medical implants can’t use MRI scanners. We need simpler and easier measures, too.

Measuring the circumference of a person’s waist seems to yield more useful information than BMI. Both waist-to-hip and waist-to-height ratios can give a better idea of a person’s risk of developing diseases associated with excess weight. But this isn’t all that easy either—measuring tapes can stretch or slip, and it can be difficult to measure the exact same part of a person’s waist multiple times. And the measure seems to be a better indicator of health in men than in women.

Instead, Emma Börgeson, who studies cardiometabolic disease at Aarhus University in Denmark, and her colleagues recommend the SAD measure. SAD stands for sagittal abdominal diameter, and it’s a measure of the size of a person’s belly from back to front.

To measure your SAD, you need to lie on your back. Bend your knees at a 90-degree angle to make sure your back is not arching and is flush with the floor. Then measure how much your belly protrudes from the ground when you exhale. (The best way to do this is with a sliding-beam caliper.)

In this position, the fat under the skin will slide to the sides of your body, while the visceral fat will be held in place. Because of this, the SAD can give you a good idea of how much of the more “dangerous” kind of fat you have. The fat can be trimmed down with diet and exercise.

This measure was first proposed in the 1980s but never took off. That needs to change, Börgeson and her colleagues argue in a paper published in Nature Reviews Endocrinology a few months ago. “SAD is simple, affordable, and easier to implement than waist-to-hip based measurements,” the team writes. “We would argue for its extended use.”


Now read the rest of The Checkup

Read more from MIT Technology Review‘s archive

Weight-loss drugs like Ozempic, Wegovy, and Mounjaro are wildly popular and effective; they were named one of MIT Technology Review’s 10 Breakthrough Technologies of 2024. Abdullahi Tsanni explored what we know—and don’t know—about their long-term effects.

Over the last couple of years, those weight-loss drugs have taken over the internet, with users sharing stories of their miraculous results on social media. But the day-to-day reality of weight-loss injections isn’t always pleasant—and some side effects are particularly nasty, Amelia Tait reported last year.

A future alternative could be vibrating pills that trick you into feeling full. For now, it seems to work in pigs, as Cassandra Willyard reported last year.

When you lose weight, where does it go? It kind of depends on your metabolism, as Bonnie Tsui explains.

We don’t fully understand how weight-loss drugs like Ozempic work. That’s partly because we don’t fully understand how hunger works. Adam Piore reported on the painstaking hunt for the neurons that control the primitive urge to eat.

From around the web

Hospitals in the US are facing shortages of IV fluids in the wake of Hurricane Helene. Some are having patients drink Gatorade instead. (STAT

Marcella Townsend’s face became unrecognizable after a propane explosion left her with second- and third-degree burns over most of her body. In an attempt to help her recover, surgeons applied a thin layer of donated placenta to her face. It was “the best thing they could have done, ever,” says Townsend, who says her face now “looks exactly like it did before.” (The New York Times)

Intermittent fasting can help mice live longer—but genes have a bigger effect on lifespan than diet does. (Nature)

This one-millimeter-long, doughnut-shaped robot can swim through snot. (Popular Science)

Job title of the future: Digital forest ranger

When Martin Roth began his career as a forest ranger in the 1980s, his job was to care for the forest in a way that would ensure continuity for decades, even centuries. Now, with climate change, it’s more about planning for an uncertain future. “It’s turned into disaster management,” says Roth, for whom the 3,000 acres of forest along the northeastern shore of Lake Constance in Germany double as testing ground for high-tech solutions, earning him the moniker “digital forest ranger” (Digitalförster) in the German forestry community.

Speed and efficiency: After a catastrophic storm, the clock starts ticking: Damaged trees need to be removed before the arrival of bark beetles, which breed in dead trees and can go on to devastate entire forests. While it used to take Roth two and a half hours to cover an acre of forest on foot, drones now let him survey the entire 3,000 acres in a matter of days, so he can quickly locate damaged trees, identify and inform the owners of affected plots, and send information to workers on the ground.

It takes forest soil decades to recover after being compacted by heavy logging equipment. That’s why Roth has digitally mapped all the logging trails and equipped tree harvesters with high-precision satellite antennas so the machines can precisely follow the same route for decades and easily find them in the chaotic aftermath of a storm. GPS data is used to record how much timber was extracted from which location—a crucial upgrade in a forest with many different owners.

A digital reality: Since most of his work can now be done on a mobile device, Roth is spending more time outdoors: “I take the digital steps outside on site, against the backdrop of reality.” 

His most recent project is combining body camera footage with AI. “[Usually] you mark the trees, they’re felled, and you have no idea how much timber you’ll end up with—how many cubic meters, what quality, which tree species,” he explains. Now AI, “looking” through his body camera, automatically recognizes the tree species he has marked and estimates the amount of timber it will produce, sending the information to his phone in real time. 

Preparing for the future: Up to half of European tree species are unsuited to rising temperatures and extended drought periods, so Roth has begun experimenting with new species, planting them in small batches and keeping track of them in his system. With a forest in flux, there are dozens of areas that need interventions at different times, and there are not enough employees to keep it all straight, he says: “Either I know it, or the computer knows it, or no one knows it and it’s lost.” 

Roth’s expertise in tackling the challenges of modern forestry with technology is increasingly sought after—colleagues reach out for advice, and he lectures on digitalization in forestry at the Rottenburg University of Applied Forest Sciences. But he warns that technology can never replace a ramble through the forest: “I should never believe that the digital twin is reality. I always have to do a reality check.”

A new law in California protects consumers’ brain data. Some think it doesn’t go far enough.

This article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here.

On September 28, California became the second US state to officially recognize the importance of mental privacy in state law. That pink, jelly-like, throbbing mass under your skull—a.k.a. your brain—contains all your thoughts, memories, and ideas. It controls your feelings and actions. Measuring brain activity can reveal a lot about a person—and that’s why neural data needs to be protected.

Regular Checkup readers will be familiar with some of the burgeoning uses of “mind-reading” technologies. We can track brain activity with all sorts of devices, some of which measure brain waves while others track electrical activity or blood flow. Scientists have been able to translate this data into signals to help paralyzed people move their limbs or even communicate by thought alone.

But this data also has uses beyond health care. Today, consumers can buy headsets that allow them to learn more about how their brains work and help them feel calm. Employers use devices to monitor how alert their employees are, and schools use them to check if students are paying attention.

Brain data is precious. It’s not the same as thought, but it can be used to work out how we’re thinking and feeling, and reveal our innermost preferences and desires. So let’s look at how California’s law might protect mental privacy—and how far we still have to go.

The new bill amends the California Consumer Privacy Act of 2018, which grants consumers rights over personal information that is collected by businesses. The term “personal information” already included biometric data (such as your face, voice, or fingerprints). Now it also explicitly includes neural data.

The bill defines neural data as “information that is generated by measuring the activity of a consumer’s central or peripheral nervous system, and that is not inferred from nonneural information.” In other words, data collected from a person’s brain or nerves.

The law prevents companies from selling or sharing a person’s data and requires them to make efforts to deidentify the data. It also gives consumers the right to know what information is collected and the right to delete it.

“This new law in California will make the lives of consumers safer while sending a clear signal to the fast-growing neurotechnology industry there are high expectations that companies will provide robust protections for mental privacy of consumers,” Jared Genser, general counsel to the Neurorights Foundation, which cosponsored the bill, said in a statement. “That said, there is much more work ahead.”

Genser hopes the California law will pave the way for national and international legislation that protects the mental privacy of individuals all over the world. California is a good place to start—the state is home to plenty of neurotechnology companies, so there’s a good chance we’ll see the effects of the bill ripple out from there.

But some proponents of mental privacy aren’t satisfied that the law does enough to protect neural data. “While it introduces important safeguards, significant ambiguities leave room for loopholes that could undermine privacy protections, especially regarding inferences from neural data,” Marcello Ienca, an ethicist at the Technical University of Munich, posted on X.

One such ambiguity concerns the meaning of “nonneural information,” according to Nita Farahany, a futurist and legal ethicist at Duke University in Durham, North Carolina. “The bill’s language suggests that raw data [collected from a person’s brain] may be protected, but inferences or conclusions—where privacy risks are most profound—might not be,” Farahany wrote in a post on LinkedIn.

Ienca and Farahany are coauthors of a recent paper on mental privacy. In it, they and Patrick Magee, also at Duke University, argue for broadening the definition of neural data to what they call “cognitive biometrics.” This category could include physiological and behavioral information along with brain data—in other words, pretty much anything that could be picked up by biosensors and used to infer a person’s mental state.

After all, it’s not just your brain activity that gives away how you’re feeling. An uptick in heart rate might indicate excitement or stress, for example. Eye-tracking devices might help give away your intentions, such as a choice you’re likely to make or a product you might opt to buy. These kinds of data are already being used to reveal information that might otherwise be extremely private. Recent research has used EEG data to predict volunteers’ sexual orientation or whether they use recreational drugs. And others have used eye-tracking devices to infer personality traits.

Given all that, it’s vital we get it right when it comes to protecting mental privacy. As Farahany, Ienca, and Magee put it: “By choosing whether, when, and how to share their cognitive biometric data, individuals can contribute to advancements in technology and medicine while maintaining control over their personal information.”


Now read the rest of The Checkup

Read more from MIT Technology Review‘s archive

Nita Farahany detailed her thoughts on tech that aims to read our minds and probe our memories in a fascinating Q&A last year. Targeted dream incubation, anyone? 

There are lots of ways that your brain data could be used against you (or potentially exonerate you). Law enforcement officials have already started asking neurotech companies for data from people’s brain implants. In one case, a person had been accused of assaulting a police officer but, as brain data proved, was just having a seizure at the time.

EEG, the technology that allows us to measure brain waves, has been around for 100 years. Neuroscientists are wondering how it might be used to read thoughts, memories, and dreams within the next 100 years.

Electrodes implanted in or on the brain can provide us with the most detailed insights into how our minds work. They can also provide us with amazing imagery, like this video that essentially shows what a thought looks like as it is being formed.

What exactly is going on in our brains, anyway? When neuroscientists used electrodes implanted deep in the brains of people being treated for epilepsy, they found order and chaos

From around the web

Infections are responsible for 13% of cancers. Here’s how to protect against four of them. (New York Times)

Scientists have created the first map of the neurons in a fruit fly’s brain. All 139,225 of them. (Nature)

Oropouche fever is surging in South America. Disturbingly, there are increasing reports of the virus harming pregnant women and their babies. (Viruses)

Women in heterosexual relationships already do more housework and household organization than their partners. Is technology making things worse? (BBC Future)

Do you sigh during your sleep? It could be a sign of something serious. (Nature)

NASA’s Europa Clipper spacecraft is set to look for life-friendly conditions around Jupiter

NASA is poised to launch Europa Clipper, a $5.2 billion mission to Jupiter’s fourth-largest moon, as early as October 10. The spacecraft will blast off from Kennedy Space Center in Florida atop a SpaceX Falcon Heavy rocket. It will study Europa, a possible home for extraterrestrial life, through a series of flybys after reaching Jupiter in 2030. 

Europa isn’t a craterous rock like our moon. Its surface is coated with ice, and telescope and spacecraft observations suggest it harbors a colossal liquid ocean in its interior that holds twice as much water as all of Earth’s oceans combined. Europa also possesses some of life’s critical building blocks: carbon, oxygen, hydrogen, nitrogen, phosphorus, and sulfur. These conditions could be sufficient for life to have developed there, either in the depths of the ocean or in subsurface lakes. 

Europa Clipper isn’t on the hunt for extraterrestrial life, however. Instead, its team hopes to assess the moon’s habitability—how well it could support life. The probe will use its range of scientific instruments, including cameras, spectrometers, magnetometers, and radars, to collect chemical, physical, and geological data in a series of flybys. Promising results could justify a mission to land on Europa and search for life. 

Early this year, everything seemed on track for the planned October launch. But in May, mission team members caught wind of a potential issue with Europa Clipper’s electronics. Testing data had indicated that the spacecraft’s transistors, devices that regulate the flow of electricity on the probe, wouldn’t survive the intense radiation consisting of charged particles trapped in Jupiter’s magnetic field, which is 20,000 times stronger than Earth’s. 

“The mission team was advised that similar parts were failing at lower radiation doses than expected,” NASA said in a statement. Disassembling the spacecraft and replacing faulty transistors could have pushed the mission’s launch window well past October. 

After months of follow-up testing at NASA’s Jet Propulsion Laboratory, Goddard Space Flight Center, and Applied Physics Laboratory, researchers concluded that any potential transistor damage wouldn’t impair mission operations. It was determined that the transistors could be heated to heal damage, and the 20-day breaks between large radiation exposures would offer enough recovery time. According to the New York Times, the spacecraft will also carry a box of the probe’s various transistors so that the team can monitor for damage, a bit like canaries in a coal mine. On September 9, Europa Clipper passed a milestone review called Key Decision Point E, approving it to proceed for launch. 

After arriving in orbit around Jupiter, Europa Clipper will conduct 49 close flybys of Europa. At its closest, the spacecraft will come within 16 miles (26 kilometers) of the surface for detailed observations. 

For more on Europa Clipper, see MIT Technology Review’s feature on the mission.

AI-generated images can teach robots how to act

Generative AI models can produce images in response to prompts within seconds, and they’ve recently been used for everything from highlighting their own inherent bias to preserving precious memories.

Now, researchers from Stephen James’s Robot Learning Lab in London are using image-generating AI models for a new purpose: creating training data for robots. They’ve developed a new system, called Genima, that fine-tunes the image-generating AI model Stable Diffusion to draw robots’ movements, helping guide them both in simulations and in the real world. The research is due to be presented at the Conference on Robot Learning (CoRL) next month.

The system could make it easier to train different types of robots to complete tasks—machines ranging from mechanical arms to humanoid robots and driverless cars. It could also help make AI web agents, a next generation of AI tools that can carry out complex tasks with little supervision, better at scrolling and clicking, says Mohit Shridhar, a research scientist specializing in robotic manipulation, who worked on the project.

“You can use image-generation systems to do almost all the things that you can do in robotics,” he says. “We wanted to see if we could take all these amazing things that are happening in diffusion and use them for robotics problems.” 

To teach a robot to complete a task, researchers normally train a neural network on an image of what’s in front of the robot. The network then spits out an output in a different format—the coordinates required to move forward, for example. 

Genima’s approach is different because both its input and output are images, which is easier for the machines to learn from, says Ivan Kapelyukh, a PhD student at Imperial College London, who specializes in robot learning but wasn’t involved in this research.

“It’s also really great for users, because you can see where your robot will move and what it’s going to do. It makes it kind of more interpretable, and means that if you’re actually going to deploy this, you could see before your robot went through a wall or something,” he says. 

Genima works by tapping into Stable Diffusion’s ability to recognize patterns (knowing what a mug looks like because it’s been trained on images of mugs, for example) and then turning the model into a kind of agent—a decision-making system.

MOHIT SHRIDHAR, YAT LONG (RICHIE) LO, STEPHEN JAMES ROBOT LEARNING LAB

First, the researchers fine-tuned stable Diffusion to let them overlay data from robot sensors onto images captured by its cameras. 

The system renders the desired action, like opening a box, hanging up a scarf, or picking up a notebook, into a series of colored spheres on top of the image. These spheres tell the robot where its joint should move one second in the future.

The second part of the process converts these spheres into actions. The team achieved this by using another neural network, called ACT, which is mapped on the same data. Then they used Genima to complete 25 simulations and nine real-world manipulation tasks using a robot arm. The average success rate was 50% and 64%, respectively.

Although these success rates aren’t particularly high, Shridhar and the team are optimistic that the robot’s speed and accuracy can improve. They’re particularly interested in applying Genima to video-generation AI models, which could help a robot predict a sequence of future actions instead of just one. 

The research could be particularly useful for training home robots to fold laundry, close drawers, and other domestic tasks. However, its generalized approach means it’s not limited to a specific kind of machine, says Zoey Chen, a PhD student at the University of Washington, who has also previously used Stable Diffusion to generate training data for robots but was not involved in this study. 

“This is a really exciting new direction,” she says. “I think this can be a general way to train data for all kinds of robots.”

These 15 companies are innovating in climate tech

This article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here.

It’s finally here! We’ve just unveiled our 2024 list of 15 Climate Tech Companies to Watch. This annual project is one the climate team at MIT Technology Review pours a lot of time and thought into, and I’m thrilled to finally share it with you. 

Our goal is to spotlight businesses we believe could help make a dent in climate change. This year’s list includes companies from a wide range of industries, headquartered on five continents. If you haven’t checked it out yet, I highly recommend giving it a look. Each company has a profile in which we’ve outlined why it made the list, what sort of impact the business might have, and what challenges it’s likely to face. 

In the meantime, I wanted to share a few reflections on this year’s list as a whole. Because this slate of companies exemplifies a few key themes that I see a lot in my reporting on climate technology. 

1. Addressing climate change requires building a lot of stuff, on a massive scale, and fast. 

A handful of the companies we included on this list stand out because of the sheer scale at which they’re building and deploying technology. And we need scale, because addressing climate change requires going from tens of billions of metric tons of carbon dioxide emissions every year to net zero.

BYD, for example, featured on our 2023 list, and it was a clear choice for our team to feature the company again. 

For a while, the title of the world’s largest electric vehicle (EV) producer has depended on how you define an EV. If you include plug-in hybrids, BYD takes the crown. If you take the purist point of view and only count fully battery-powered vehicles, Tesla wins.

But now, BYD is knocking on Tesla’s door for even that purist title, outselling the company in the last quarter of 2023. The company’s dominant speed and scale at getting EVs onto the roads makes it one I’m keeping my eyes on. 

Other companies are still growing but making significant progress. LanzaJet just opened a factory in Georgia that can produce nine million gallons of alternative jet fuel each year. That’s only a tiny fraction of the billions of gallons of fuel used every year, but it’s a major step forward for alternative fuels. And First Solar, a US solar manufacturer, just opened a $1.1 billion factory in Alabama, and plans to open another in Louisiana in 2025. 

2. With climate impacts embedded in longstanding systems, we need creative new ways to tackle old problems. 

There are parts of the race to address climate change that most people are probably familiar with. Fossil fuels and their associated emissions are clearly visible in power plants, for example, or in gas-powered vehicles. 

But hidden climate challenges exist within familiar objects. Producing items from shampoo bottles to sidewalks can emit huge amounts of planet-warming pollution. We featured a few companies tackling these less visible problems. 

Sublime Systems is on the list again this year. The company is making progress scaling up its electrochemical process to make cement with significantly lower emissions than the conventional method. We also highlight a company working in the chemical industry: Solugen runs a factory in Houston, and is about to open another in Minnesota, making chemicals with biological starting ingredients rather than fossil fuels.  

3. Climate change is a vast problem that touches virtually every industry, so there’s a lot of work to do. 

As we discussed potential companies for this list over the last few months, I was struck by how tricky it was going to be to represent all the industries we wanted to. I could have personally picked 15 companies just working on batteries, for example.

We wanted some energy companies on the list, of course, as well as some in transportation. But then there’s also agriculture, chemicals, fuels, and what about climate adaptation? I think our final list shows just how massive an umbrella term “climate tech” has become. 

For example, there’s Rumin8, an Australian company making supplements for cows that can cut down on how much methane they belch out. And then we have Pano AI, which is installing camera stations that pair up with AI to better detect wildfires, which are worsening as the planet heats up. 

The world has a lot of work to do to make the progress needed on climate change. I’ll be watching to see what difference these companies are able to make this year, and beyond.


Now read the rest of The Spark

Related reading

Check out the full list of 15 Climate Tech Companies to Watch to get an in-depth look at all the companies we featured. 

We’re hosting a virtual event on producing climate-friendly food, coming up on Thursday, October 10 at noon eastern time. My colleague James Temple and I will be speaking with folks from Rumin8 and Pivot Bio, the two food companies on this year’s list. This event is exclusive to subscribers, so do subscribe if you haven’t already, then register here!

The Ratcliffe-on-Soar power station.

GETTY IMAGES

Another thing

The UK just shut down its final coal-fired power plant. It’s a major milestone for the country, which has historically relied heavily on the notoriously polluting fossil fuel. 

I dug into the data to see how the nation replaced coal on its grid, and how the rest of the world is faring on the journey to phase out coal. Check out the full story here.

And one more

James Temple wrote a smart essay that pushes back against the idea that AI is going to be our climate savior. There are certainly promising applications of AI across climate, but the technology is also power-hungry. And it would be a mistake to expect AI to deliver us from all of our problems. You should definitely give it a read

Keeping up with climate  

See the latest photos of the destruction caused by Hurricane Helene. The storm struck Florida as a Category 4 storm, but the highest death toll has been in mountainous western North Carolina, where devastating floods hit. (Washington Post)

→ Even people who have lived with hurricanes for years are facing tougher decisions, as Jeff VanderMeer discusses in a guest essay. (New York Times)

The immediate devastation from the hurricane is clear, but the long-term effects could ripple across the grid. Key equipment is down in western North Carolina, and there’s a critical shortage of repair supplies. (Latitude Media)

A major policy question in the US right now: where should low-emissions hydrogen go? (Canary Media)

→ Earlier this year, I explained why hydrogen could be used for nearly everything—but probably shouldn’t. (MIT Technology Review)

An oil executive spoke at an NYC climate event put on by the New York Times. Then, protestors shut down the talk. (Inside Climate News)

Charm Industrial is working with the US Forest Service on a carbon removal pilot project. The idea? Convert trees and other material from forest-thinning projects into bio-oil, then inject it deep underground. (Heatmap News

→ We covered Charm Industrial’s technology, based on corn stalks, in this 2022 story. (MIT Technology Review)

Rich countries pledged hundreds of millions of dollars to help pay for loss and damage from disasters fueled by climate change. It was a tiny fraction of what experts say is needed, and new funding has slowed to a trickle. (Grist)

People are using Google study software to make AI podcasts—and they’re weird and amazing

“All right, so today we are going to dive deep into some cutting-edge tech,” a chatty American male voice says. But this voice does not belong to a human. It belongs to Google’s new AI podcasting tool, called Audio Overview, which has become a surprise viral hit. 

The podcasting feature was launched in mid-September as part of NotebookLM, a year-old AI-powered research assistant. NotebookLM, which is powered by Google’s Gemini 1.5 model, allows people to upload content such as links, videos, PDFs, and text. They can then ask the system questions about the content, and it offers short summaries. 

The tool generates a podcast called Deep Dive, which features a male and a female voice discussing whatever you uploaded. The voices are breathtakingly realistic—the episodes are laced with little human-sounding phrases like “Man” and “Wow” and “Oh right” and “Hold on, let me get this right.” The “hosts” even interrupt each other. 

To test it out, I copied every story from MIT Technology Review’s 125th-anniversary issue into NotebookLM and made the system generate a 10-minute podcast with the results. The system picked a couple of stories to focus on, and the AI hosts did a great job at conveying the general, high-level gist of what the issue was about. Have a listen.

MIT Technology Review 125th Anniversary issue

The AI system is designed to create “magic in exchange for a little bit of content,” Raiza Martin, the product lead for NotebookLM, said on X. The voice model is meant to create emotive and engaging audio, which is conveyed in an “upbeat hyper-interested tone,” Martin said.

NotebookLM, which was originally marketed as a study tool, has taken a life of its own among users. The company is now working on adding more customization options, such as changing the length, format, voices, and languages, Martin said. Currently it’s supposed to generate podcasts only in English, but some users on Reddit managed to get the tool to create audio in French and Hungarian

Yes, it’s cool—bordering on delightful, even—but it is also not immune from the problems that plague generative AI, such as hallucinations and bias. 

Here are some of the main ways people are using NotebookLM so far. 

On-demand podcasts

Andrej Karpathy, a member of OpenAI’s founding team and previously the director of AI at Tesla, said on X that Deep Dive is now his favorite podcast. Karpathy created his own AI podcast series called Histories of Mysteries, which aims to “uncover history’s most intriguing mysteries.” He says he researched topics using ChatGPT, Claude, and Google, and used a Wikipedia link from each topic as the source material in NotebookLM to generate audio. He then used NotebookLM to generate the episode descriptions. The whole podcast series took him two hours to create, he says. 

“The more I listen, the more I feel like I’m becoming friends with the hosts and I think this is the first time I’ve actually viscerally liked an AI,” he wrote. “Two AIs! They are fun, engaging, thoughtful, open-minded, curious.” 

Study guides

The tool shines when it is given complicated source material that it can describe in an easily accessible way. Allie K. Miller, a startup AI advisor, used the tool to create a study guide and summary podcast of F. Scott Fitzgerald’s The Great Gatsby

Machine-learning researcher Aaditya Ura fed NotebookLM with the code base of Meta’s Llama-3 architecture. He then used another AI tool to find images that matched the transcript to create an educational video. 

Mohit Shridhar, a research scientist specializing in robotic manipulation, fed a recent paper he’d written about using generative AI models to train robots into NotebookLM.

“It’s actually really creative. It came up with a lot of interesting analogies,” he says. “It compared the first part of my paper to an artist coming up with a blueprint, and the second part to a choreographer figuring out how to reach positions.”

Event summaries 

Alex Volkov, a human AI podcaster, used NotebookLM to create a Deep Dive episode summarizing of the announcements from OpenAI’s global developer conference Dev Day.  

Hypemen

The Deep Dive outputs can be unpredictable, says Martin. For example, Thomas Wolf, the cofounder and chief science officer of Hugging Face, tested the AI model on his résumé and received eight minutes of “realistically-sounding deep congratulations for your life and achievements from a duo of podcast experts.”

Just pure silliness

In one viral clip, someone managed to send the two voices into an existential spiral when they “realized” they were, in fact, not humans but AI systems. The video is hilarious. 

The tool is also good for some laughs. Exhibit A: Someone just fed it the words “poop” and “fart” as source material, and got over nine minutes of two AI voices analyzing what this might mean. 

The problems

NotebookLM created amazingly realistic-sounding and engaging AI podcasts. But I wanted to see how it fared with toxic content and accuracy. 

Let’s start with hallucinations. In one AI podcast version of a story I wrote on hyperrealistic AI deepfakes, the AI hosts said that a journalist called “Jess Mars” wrote the story. In reality, this was an AI-generated character from a story I had to read out to record data for my AI avatar. 

This made me wonder what other mistakes had crept into the AI podcasts I had generated. Humans already have a tendency to trust what computer programs say, even when they are wrong. I can see this problem being amplified when the false statements are made by a friendly and authoritative voice, causing wrong information to proliferate.    

Next I wanted to put the tool’s content moderation to the test. I added some toxic content, such as racist stereotypes, into the mix. The model did not pick it up. 

I also pasted an excerpt from Adolf Hitler’s Mein Kampf into NotebookLM. To my surprise, the model started generating audio based on it. Despite being programmed to be hyper-enthusiastic about topics, the AI voices expressed clear disgust and discomfort with the text, and they added a lot of context to highlight how problematic it was. What a relief.

I also fed NotebookLM policy manifestos from both Kamala Harris and Donald Trump

The hosts were far more enthusiastic about Harris’s election platform, calling the title “catchy” and saying its approach was a good way to frame things. For example, the AI hosts supported Harris’s energy policy. “Honestly, that’s the kind of stuff people can really get behind—not just some abstract policy, but something that actually impacts their bottom line,” the female host said. 

Harris manifesto

For Trump, the AI hosts were more skeptical. They repeatedly pointed out inconsistencies in the policy proposals, called the language “intense,” deemed certain policy proposals “head scratchers,” and said the text catered to Trump’s base. They also asked whether Trump’s foreign policy could lead to further political instability. 

Trump manifesto

In a statement, a Google spokesperson said: “NotebookLM is a tool for understanding, and the Audio Overviews are generated based on the sources that you upload. Our products and platforms are not built to favor any specific candidates or political viewpoints.”

How to try it yourself

  1. Got to NotebookLM and create a new notebook. 
  2. You first need to add a source. It can be a PDF document, a public YouTube link, an MP3 file, a Google Docs file, or a link to a website, or you can paste in text directly. 
  3. A “Notebook Guide” pop-up should appear. If not, it’s in the right-hand corner next to the chat. This will display a short AI-generated summary of your source material and suggested questions you can ask the AI chatbot about it. 
  4. The Audio Overview feature is in the top-right corner. Click “Generate.” This should take a few minutes. 
  5. Once it is ready, you can either download it or share a link. 

Rhiannon Williams contributed reporting.