This lab robot mixes chemicals

Lab scientists spend much of their time doing laborious and repetitive tasks, be it pipetting liquid samples or running the same analyses over and over again. But what if they could simply tell a robot to do the experiments, analyze the data, and generate a report? 

Enter Organa, a benchtop robotic system devised by researchers at the University of Toronto that can perform chemistry experiments. In a paper posted on the arXiv preprint server, the team reported that the system could automate some chemistry lab tasks using a combination of computer vision and a large language model (LLM) that translates scientists’ verbal cues into an experimental pipeline. 

Imagine having a robot that can collaborate with a human scientist on a chemistry experiment, says Alán Aspuru-Guzik, a chemist, computer scientist, and materials scientist at the University of Toronto, who is one of the project’s leaders. Aspuru-Guzik’s vision is to elevate traditional lab automation to “eventually make an AI scientist,” one that can perform and troubleshoot an experiment and even offer feedback on the results. 

Aspuru-Guzik and his team designed Organa to be flexible. That means that instead of performing only one task or one part of an experiment as a typical fixed automation system would, it can perform a multistep experiment on cue. The system is also equipped with visualization tools that can monitor progress and provide feedback on how the experiment is going.  

“This is one of the early examples of showing how you can have a bidirectional conversation with an AI assistant for a robotic chemistry lab,” says Milad Abolhasani, a chemical and material engineer at North Carolina State University, who was not involved in the project. 

Most automated lab equipment is not easily customizable or reprogrammable to suit the chemists’ needs, says Florian Shkurti, a computer scientist at the University of Toronto and a co-leader of the project. And even if it is, the chemists would need to have programming skills. But with Organa, scientists can simply convey their experiments through speech. As scientists prompt the robot with their experimental objectives and setup, Organa’s LLM translates this natural-language instruction into χDL codes, a standard chemical description language. The algorithm breaks down the codes into steps and goals, with a road map to execute each task. If there is an ambiguous instruction or an unexpected outcome, it can flag the issue for the scientist to resolve.

About two-thirds of Organa’s hardware components are made from off-the-shelf parts, making it easier to replicate across laboratories, Aspuru-Guzik says. The robot has a camera detector that can identify both opaque objects and transparent ones, such as a chemical flask.

Organa’s first task was to characterize the electrochemical properties of quinones, the electroactive molecules used in rechargeable batteries. The experiment has 19 parallel steps, including routine chemistry steps such as pH and solubility tests, recrystallization, and an electrochemical measurement. It also involves a tedious electrode-precleaning step, which takes up to six hours. “Chemists really, really hate this,” says Shkurti.

Organa completed the 19-step experiment in about the same amount of time it would take a human—and with comparable results. While the efficiency was not noticeably better than in a manual run, the robot can be much more productive if it is run overnight. “We always get the advantage of it being able to work 24 hours,” Shkurti says. Abolhasani adds, “That’s going to save a lot of our highly trained scientists time that they can use to focus on thinking about the scientific problem, not doing these routine tasks in the lab.” 

Organa’s most sophisticated feature is perhaps its ability to provide feedback on generated data. “We were surprised to find that this visual language model can spot outliers on chemistry graphs,” explains Shkurti. The system also flags these ambiguities or uncertainties and suggests methods of troubleshooting. 

The group is now working on improving the LLM’s ability to plan tasks and then revise those plans to make the system more amenable to experimental uncertainties. 

“There’s a lot roboticists have to offer to scientists in order to amplify what they can do and get them better data,” Shkurti says. “I am really excited to try to create new possibilities.” 

Kristel Tjandra is a freelance science writer based in Oahu. 

Oropouche virus is spreading. Here’s what we know.

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.

There have been plenty of reports of potentially concerning viruses this last year. Covid is still causing thousands of deaths, and bird flu appears set to make the jump to human-to-human transmission. Now there are new concerns over Oropouche, a virus largely spread by bites from insects called midges (sometimes called no-see-ums in the US).

There have been outbreaks of the Oropouche virus in Latin America for decades. But this one is different. The virus is being detected in all-new environments. It is turning up in countries that have never seen it before. The spread is being described as “unprecedented.”

It may also be causing more severe disease. People with Oropouche fever typically have a sudden fever, aches and pains, and nausea. Most cases are mild, but some people have developed encephalitis and meningitis. And this year, two otherwise healthy young women who caught the virus have died.

Oropouche can be passed from mother to fetus, and it has been linked to stillbirths and birth anomalies. There are no treatments. There are no vaccines, either. This week, let’s take a look at why Oropouche is spreading, and what we can do about it.

Oropouche virus was first identified in 1955, in a person and a pool of mosquitoes from the village of Vega de Oropouche in Trinidad and Tobago. It was found in a sloth in Brazil in 1960. Since then, there have been over 30 outbreaks—in those countries as well as Peru, Panama, Colombia, French Guiana, and Venezuela. At least 500,000 cases have been reported in South America, largely in areas close to forest.

That’s probably because of the way the virus is transmitted. Oropouche virus is thought to be carried by some populations of sloths, and potentially some nonhuman primates. These animals can host the virus, which can then spread to people via insect bites, usually from midges or some types of mosquitoes.

Since late 2023, outbreaks have been reported in a number of countries in South America, Central America, and the Caribbean, including Cuba, a first for the country. 

There has been an especially large surge of cases in Brazil. Since the beginning of this year, 10,275 cases of Oropouche have been confirmed in the Americas, according to a situation summary report published by the Pan American Health Organization (PAHO) earlier this week. And 8,258 of them were in Brazil. Travelers have also imported cases to the US and Europe for the first time—90 such cases have been reported in the US, and 30 in Europe.

Another change is that this time around, the virus has been infecting people in urban settings far from forests. It is not entirely clear why, but there are probably a few reasons. Climate change, for a start, has led to increased temperatures and rainfall, both of which can help create breeding grounds for the insects that transmit the virus. And deforestation and urbanization, both of which have caused people to encroach on the habitats of wild animals, have also raised the risk of transmission to people, says Ana Pereiro do Vale, a veterinarian and microbiologist at University College Dublin in Ireland.

The virus itself also appears to have changed, according to new research published this week. William de Souza at the University of Kentucky and his colleagues analyzed blood samples taken from people with an Oropouche diagnosis between 2015 and 2024, enabling them to compare the form of the virus that is currently circulating with a historical strain.

The team found evidence that the virus has swapped genetic material with a related one, creating a new “virus reassortment.” It is this new form of the virus that has spread since the end of 2023, the team says.

That’s not all. The genetic changes have endowed the virus with new features. The current reassortment appears to be better at replicating in mammalian cells. That might mean that infected people—and sloths—have more of the virus in their blood, making it easier for biting insects to pick it up and pass it on.

The new form of the virus also seems to be more virulent. The team’s lab tests suggest that compared with the historical strain, it appears to cause more damage to the cells it infects.

We are still getting to grips with how the virus can spread, too. We know midges and mosquitoes are responsible for spreading Oropouche, but the virus can also pass to a fetus during pregnancy, with potentially harmful consequences. According to the PAHO report, Brazil has reported “13 fetal deaths, three spontaneous miscarriages, and four cases of birth anomalies” linked to Oropouche infections.

In a separate study published earlier this week, Raimunda do Socorro da Silva Azevedo at the Evandro Chagas Institute in Ananindeua, Brazil, and her colleagues assessed 65 unexplained cases of microcephaly—a birth anomaly in which babies have an unexpectedly small head—that had been recorded in Brazil between 2015 and 2024. The team found evidence of an Oropouche infection in six of the babies—and in all three that had been born in 2024.

It’s still not clear whether or how the virus might affect fetuses and babies, and research is ongoing. But the US Centers for Disease Control and Prevention (CDC) recommends that pregnant travelers “reconsider non-essential travel” to Cuba

Some scientists worry that the virus might also spread via sex. In August, a 42-year-old Italian man who fell ill after returning from a trip to Cuba was found to have Oropouche virus in his semen. And it was still there 58 days later. The CDC currently recommends that men diagnosed with Oropouche should use condoms or not have sex for at least six weeks from the start of their symptoms. They should avoid donating semen, too, according to the organization.

There are a lot of unanswered questions when it comes to Oropouche. Some scientists have suggested that this is because outbreaks have historically been seen in poorer countries in the Global South.

“There is sufficient colonialism in disease research—if it doesn’t affect the industrial world and Western business interests, it’s not important,” Shahid Jameel, a virologist at the University of Oxford, told Gavi, an organization focused on global vaccination efforts. “Now that the virus has been found in Cuba—not far from Miami—the wheels of public health will turn.”

Let’s hope they get in gear quickly. As Vale says: “We don’t know what will happen with the virus, the mutation rate of the virus, or if the virus will jump to another host. We need to be careful and pay attention.”


Now read the rest of The Checkup

Read more from MIT Technology Review‘s archive

Oropouche infections can look similar to dengue—another viral disease, also spread by mosquitoes, that affects people in Brazil. The country is attempting to tackle the problem with bacteria-infected mosquitoes, Cassandra Willyard reported in March.

The spread of bird flu in dairy cattle in the US has virologists worried. The virus could stick around on US farms forever and is raising the risk of outbreaks in mammals—including humans—around the world.

Flu season is officially upon those of us in the Northern Hemisphere. This year, it could enable the creation of an all-new bird flu, too. 

Could gene editing help curb the spread of bird flu? Abdullahi Tsanni explored the possibility of using CRISPR to make chickens resistant to the virus.

Another option, of course, is vaccines. Most flu vaccines are made, ironically, in chicken eggs. mRNA vaccines could provide an alternative, egg-free approach.

From around the web

A fertility clinic in London has helped two transgender individuals have a baby in a process that involved egg freezing, donated sperm, IVF, embryo storage, and surrogacy. “To our knowledge this is the first report of family building by a transgender couple in which both partners had successfully achieved gender reassignment and the creation of a family through surrogacy,” write the team. (Reproductive BioMedicine Online)

“They showed me them in a mirror … and I looked like a witch,” says one woman who has experienced the horror of dental veneers gone wrong. Veneers have become as routine as Botox and lip filler. But what can people do when their dream of a perfect smile turns into a nightmare? (The Guardian)

Thinking about deleting your 23andMe data? The company will hold on to some of it regardless, to comply with legal regulations. Some of your genetic information, your date of birth and your sex, and data linked to your account deletion request will all be retained. (MIT Technology Review)

Pet dogs are spending more time indoors, in environments they aren’t suited to. Service dogs, on the other hand, are uniquely well adapted to life in the 21st century, say two researchers at the Duke Canine Cognition Center. Humans need to breed and train more puppies like service animals, they argue. (The Atlantic)

Azalea: a science-fiction story

“This is simply a question of right and wrong.”

“You can’t deny the costs, though. You keep saying that just one more year of taxes will solve—

“We’re not solving—we’re mitigating!”

“Then what’s the point?”

The shrill back-and-forth fills the kitchen, where Xia is busy making breakfast, some kind of awful cricket-protein smoothie with kale. Tascha squeezes into the small space behind her, kisses her on the cheek. 

“Can you maybe put that in your head?”

Xia doesn’t put it in her head, but she at least lowers the volume with a click of her tongue. ElectoPod’s never-ending shouting match becomes something more akin to ocean noise, where only occasional angry waves splash through the kitchen. 

Tascha digs through the fridge, looking for something that isn’t kale or crickets. Finds a BitterBucketBrew and cracks it open. The coffee comes from gengineered plants that can survive higher latitudes. Caffeine in heaps, plus a proprietary process to filter out almost all endocrine disruptors, phthalates, microplastics, arsenic, and lead. Xia says it isn’t as good as coffee grown from heirloom stocks. It’s not natural, she says. Xia also says that a 99% filtration rate isn’t all that great when hormone mimics are dangerous in parts per trillion. 

For a refugee, Xia can be awfully picky. 

“You should be paying attention to the election,” Xia says. “This is your country.”

Tascha sips her coffee. “That’s why I have you.”

“Why don’t you run for district carbon board?” Xia presses. 

“Because then I’d have to deal with people. Anyway, I don’t have time. I’m trying to make bonus so we can get into Azalea.”

“If you don’t make time, someone stupid will. Gribaldi is running again. In Texas—”

“Don’t worry.” Tascha kisses Xia on the forehead. “Even Gribaldi’s not as stupid as Texas.” 

Xia makes a biting motion at Tascha, deliberately turning up the volume in the kitchen.

Is that passive-aggressive? 

Or aggressive-aggressive? 

Regardless, ElectoPod once again floods the kitchen with the latest depressing news. A new generation of nearly undetectable AI proxies are battling it out for mindshare as November approaches. An ocean of microtargeted content is pouring into people’s feeds, custom-generated on-the-fly ads and entertainment based on mountains of tracking data—all of it illegally obtained offshore, all of it tailored to sway public opinion—and no one knows who or what is generating it. It’s enough to make Tascha think she should have fought through her ADHD and gotten her programming degree. Someone has to be making money off that. 

Instead, Tascha clicks her tongue and turns on her own feed. Peace instantly envelops her, as ElectoMute smothers ElectoPod. Custom-tuned bone-conducted vibrations hum through her skull, perfectly canceling out the sound waves of Xia’s obsession. ElectoMute is Tascha’s only paid subscription. It’s not even sound, Xia complains every time she sees the monthly bill. A symptom of Late-Stage Capitalism. Paying to make the noise of another feed go away. 

Tascha calls it the best $50 a month she’s ever spent. 

Sometimes, it’s just nicer to shut things out. If Tascha’s honest, it’s always nicer to shut things out. Ever since she got her first bone implants on recommendation of the school counselor to help her focus and calm herself, she’s been a fan of shutting things out. People are both distracting and a hassle. Tascha is still sort of amazed that Xia doesn’t get on her nerves more. Sure, she also has another—very secret—mute feed tuned to Xia’s voice … but doesn’t everyone put their relationship on mute sometimes?

Xia is pushing a smoothie across the table at her. Her lips make noise shapes. “No smoothie?” 

Tascha shuts off ElectoMute and XiaMute. “Did you know the plywood they’re using on the worksite is made of mushrooms?” 

“So?”

“It’s like, mushroom-hemp composite. I could bring some back for your smoothies.”

“Very funny.”

“It’s carbon negative. You’d love it.” 

Xia gives her a sharp look. “Don’t be cute. I’ll take it from the kid who lost her whole town to a tornado, not from you.” 

Xia volunteers at the Georgia Displacement Authority. She doesn’t have a full work permit yet, but she can volunteer, so, of course, she does. 

Xia, always looking out for everyone. 

Tascha nurses her smoothie. Her father says that relationships are about compromise. If the worst thing about Xia is kale-cricket smoothies, Tascha knows she’s a winner. She forces down the last of the smoothie and gets up from the table. 

“I’m late.” 

Xia’s lips move again, making more mouth shapes. 

Tascha tunes back in. “What?” 

“I said, make sure your frigrig’s charged. It’s hot today.”

“It’s always charged. They charge them every shift.”

Xia is undeterred. “And swap out your mask filters. Canada’s burning up again.”


“How is there even any forest left?” 

Janet’s voice crackles in Tascha’s ear as they dangle off their rappelling lines, swinging from point to point in their harnesses. The gray soup of Canada’s burning forests envelops them. 

“How are we in Atlanta, sucking smoke from fucking Canada?” Janet continues. “Are we not America? Do we not have a long and honorable tradition of blockades, border checkpoints, and deportation? If we can keep Texas and Florida out, why not Burnt Canadians? Put up a sign: BC: Not Fucking Welcome.” 

They’re stitching solar, dangling 300 feet in the air, working their way down Tower 3 of the new Azalea Arcology. Tying window electrics, solar paint, and cell panels together. The mix is meant to make a lovely pattern (azaleas, in fact) on the face of the arcology structure, but the architect should be shot, because the electronics are a hassle and Tascha’s crew is behind schedule.

Most of the arcology is fast-attach, standardized like Lego blocks, built in factories, then autonomously shipped to the site and popped together, as simple as a kit. In the early stages, Azalea was just swarms of bots digging, grading, and auto-­assembling according to plan, knitting together the bones and skin of an entire new city of 10,000. Now that they’re at final finishing stages, though, humans are taking back the site. The complex patterns of varied electrical components still need a clever human touch, which is why Tascha’s crew is out in the Atlanta swelter, nearly mummified in Day-Glo frigrigs to keep the heat at bay. 

 It’s one thing to bike to work with just a filter mask and a chilled helmet to keep you cool; it’s a whole other thing to hang off the side of a building all day when wet-bulb temps push into the 40s. 

“I heard someone was camping up in Alberta and lit a bunch of beetle kill on fire. Whole state’s going up.” 

“Is it a state or a province?”

“How the fuck would I know?”

“Can we get off the coms, people?” Latoya, their crew lead, interrupts. “Some of us are trying to make bonus.”

“Yeah, Janet, get to work.”

A whole chorus of agreement follows from the rest of the crew.

“Yeah, Janet, get to work.” 

“Yeah, Janet, get to work.”

“Where the fuck is Janet? I can’t even see her in this smoke. It’s like pea soup.” 

Tascha slides across to a bank of PV windows and takes a sip from her frigrig reservoir. Cool water, sucked out of the humid air by the suit. Far below, cyclists zip down the street, their chilled jackets boosted by their bike batteries, popping in and out of view from under the street’s shading solar shelters and tree foliage. They look like schools of minnows. Heavy cargo vans and zip buses string out in a line, following programmed one-way routes to give most of the space to cyclists. AItlantis, the city management software, must have detected an increase in people waiting at bus shelters, because a surge of robotic zip buses is swarming toward Obama Greenway. 

Tascha starts splicing and soldering a cluster of energy-­generating windows to the solar paint that surrounds them, and then hooking the whole into wires that will carry the energy into the main arcology grid. It’s fussy work. Janet keeps bitching about the smoke. 

“I swear,” she says, as she spools down and jerks to a stop next to Tascha. “I’m just going up there and lighting the rest of that fucking forest on fire. I mean, shit, let’s get it over with—Hey. Tascha. You doing okay?” 

“I’m fine, why?”

“You missed a connection.” 

Tascha blinks away sweat that’s dripping in her eyes. “Oh. Thanks.”

Janet reaches into the nested wires. Tightens screws. Runs her own diagnostic. Everything glows green. “Can’t have our pretty solar design fail before it even gets hooked up, right?”

“Right. Yeah.” Tascha takes another sip of water. The frigrig’s reservoir and heat pump should be keeping it ice cold, but it’s more lukewarm. “Hey, can you check my frigrig battery?” 

Janet spins her around, checks her back. “Looks good to me. Seventy percent.” 

“Plugged in tight?” 

Tugs and jerks. “Yeah. All good. All tight.” 

“Let’s keep it moving, people. I want to finish this wall today,” Latoya says over the com. “Let’s get our bonus, right? For once?”

“If they designed these connections for frigrig gloves, this would go faster,” Janet says.

“Get back to work, Janet.” 

They all say it at the same time, and laugh.

Tascha’s father claims people didn’t wear any kind of cooling clothing in the old days. They wore tank tops and shorts, and sure they sweated buckets, but people didn’t have to completely hide from the heat. Tascha can’t imagine it. The only person Tascha knows who spends any time out in the heat willingly is Xia. Sometimes Xia lies nude on their balcony, letting the sun burn down on her, her skin sheened with sweat, salt jewels trickling lazily down the curves of her ribs. 

There’s something seductive about the contrast between Xia’s sun-browned flesh and the pale spiderweb of lines where the filter mask and its straps have hugged her face.

It’s fascinating and horrifying, like watching someone cook in an oven. 

Xia says it’s like a giant natural sauna, so why wouldn’t she take advantage of it? Saunas are good for you. Just ask the Finns. Xia also claims she’s epigenetically advantaged thanks to the Texas grid constantly failing during her childhood. Her body has been trained to survive any heat—which is so categorically bullshit that Tascha doesn’t even bother to argue. But Tascha does like Xia’s tan lines. There’s something seductive about the contrast between Xia’s sun-browned flesh and the pale spiderweb of lines where the filter mask and its straps have hugged her face. When Tascha ran her fingers over the tan lines one night, Xia told her it was a fetish. 

“It’s called FMT,” Xia said. “Filter mask tan lines. Very Rule 34.”

For reasons Tascha can’t fully explain, she’s annoyed that something that felt personal and private is actually a well-trodden porn search. Even when she’s alone in her own mind, tentatively feeling her way into real intimacy with Xia for the first time in her life, she’s still surrounded by people. At this very moment, there’s probably some algorithm custom-tailoring political attack ads based on FMT. It probably already knows about Tascha. 

People ruin everything.


“Goddamn, that looks like the good life.” 

Janet is peering in through another cluster of windows that they’re supposed to be hooking up. “Check it. You can see the waterfall and the river from here. It’s finished!”

Tascha realizes that she’s been leaning her head against the glass. She wipes sweat out of her eyes and peers through the PV glaze. Sure enough, the artificial river meanders along under high glass gallery arches, doing its job of water repurification and cooling as it winds through the arcology, then out under the dome of a semi-wild park, where lots of fast-growing carbon-sink cypress and citrus are growing, then cascading and pooling down through a series of rapids down into the artificial canyon the arcology uses for geocooling. Deep down in the shadows, Tascha glimpses a series of artificial lakes where mercury-free fish are destined to be raised.  

“Someone’s kayaking!”

A bright-yellow kayak has entered the top of the cataracts, some lunatic with a red helmet paddling down through the water features. Now that Tascha is looking closely, she can see that another whole part of the canyon is destined to be climbing walls. 

“I’m definitely buying in,” Janet says. “You buying in? We get top slots in the lottery, since we worked on it.” 

“I guess it depends if we get bonus.” Tascha’s hands feel clumsy. She drops the leads. “Are you hot? I think my suit’s fritzing.” 

“You want to tap out early, get it checked?”

“It’s just another couple hours. I’m fine.”

“Don’t try and muscle through—”

“I’m fine. Xia keeps telling me saunas are good for you. Let’s get our bonus.” 

Now that she’s sure her frigrig is fritzing, the heat becomes more bearable. She just imagines Xia, sunbathing, sweating it out intentionally. If Xia can take it, Tascha can take it. Christ, Xia complains if Tascha even turns up the A/C in the condo. She’s got the poverty mentality of all Texans, where people dying for lack of electricity is one of the independent territory’s founding principles. 

“It’s fine,” Tascha had explained, the first time they got in a fight over what constituted a reasonable temperature. “The grid’s in surplus. We’re doing them a favor by using it.”

“You’re making that up.”

“I can literally air-condition the balcony if I want. If you just open the doors I can knock it down 15 degrees. I can make you comfortable out there.”

“Don’t you dare.”

It makes Tascha want to move into Azalea even more. The whole place is kept at reasonable temperatures all day, every day. Outdoor Living, Indoors! is the arcology’s tagline, and it sounds like heaven. No wildfire smoke. Controlled temps. All those parks and rec trails and outdoor cafés. The energy systems connected to the cooling systems connected to the hydroponics systems, all of it managed by the unfortunately named AIzalIA Management Software that should, according to the brochure, make the entire arcology not only function as a carbon sink but also run an energy surplus that all the residents will profit from. 

Xia hates the idea of it. 

“It’s more privatization. You can’t privatize municipal services. It drains support for centralized government and general infrastructure. The rich live great, and the poor die like flies.”

“This isn’t Texas. That’s not how we do it here.”

“It’s not Texas … yet. If you let the rich live apart from the rest, eventually they start to undermine everything.”

“Can you just enjoy things, for once? Maybe practice a little optimism?” 

“It happened with schools. It will happen with infrastructure if you let it.”

“You know, this is exactly why the High Reverend of Texas has a warrant out for you. You’re lucky we don’t extradite.”

Xia makes a face. Tascha feels bad. Xia worrying is the same as Xia getting involved is the same as Xia making trouble is the same as Xia taking care of people is the same as Xia taking care of Tascha. It’s what she does. Tascha kisses her on the forehead. “Not everything is a plot to destroy the world.”

“This is exactly how Florida drowned itself. The rich got rich, and then they got on their private jets and flew away when Miami drowned. They always planned on kissing off to somewhere else. You can’t let these people undermine everything and then run away to hide with all their wealth.”

“I don’t think that’s what Azalea is about—”

“Yeah? What’s the buy-in?”

“That’s not fair. You know how much it costs to build. This ain’t cheap tech.”

“You know what would have been cheap? Just fixing the problem in the beginning so we all could just have gone on ­outdoor living, you know, outdoors. But rich people figured they’d be protected, so they didn’t give a shit. They’d move to New Zealand, right? They’d make their own personal compounds. They’d hire guards. They’d make Azaleas and they’d be fine—”

“But we can buy in too! If I make bonus, we’ll have enough—” 

Xia bites her teeth, hard. ElectoPod streams into Tascha’s head, bypassing ElectoMute: a pair of commentary hosts, haranguing one another.

“I think we need to remember that people in Florida had incomplete information.”

“Bullshit. They had everything they needed.” 

“Come on, Sunita. No one sets out to drown themselves! The people who drowned weren’t the people who made the disaster plans. Florida’s governor didn’t care how many people died. His real estate donors didn’t care. They had the numbers. They knew how much bigger storm surges were going to get—”

“So no one had a clue at all? They were just sitting in the dark like mushrooms? Come on, Maria. Let’s listen to this.”

A news announcer cuts in, old news coverage:

 “That’s the South Beach seawall. We can see the water coming up, coming through. We don’t know how many people are still in lower Miami. Obviously, this brings to mind the levee break in New Orleans in the early 2000s. Our thoughts and prayers are with the people of Florida in this trying time.” 

The argument between the hosts resumes. 

“Reminiscent of New Orleans! They had 70 years of warning! Literally everyone knew. Not just the governor! Not just his real estate donors. Don’t bother defending them, Maria. People got exactly what they signed up for. They deserved it.”

Tascha wants to argue with Xia. To point out that ElectoPod is saying that it wasn’t just rich people, that everyone was stupid, that everyone went along. Bottom line, people in general are just stupid, but Xia keeps talking at her, and XiaMute doesn’t seem to be working.

“Wake up, Tascha! You can’t just seal yourself off from people.”

“Wake up, Tascha! You have to be involved. If you don’t get involved, stupid people will.”

Wake up, Tascha! If you don’t pay attention, other people will decide for you.”

“Wake up Tascha! I know about XiaMute.”

“Wake up!”

“Wake up!”

“Wake up!”


Tascha comes awake, gasping. Water rushes around her. She thrashes, trying to swim, trying to keep her head above water.

“Whoa, girl! Take it easy!” Janet is cradling her in her arms, along with some woman in a red helmet. 

The kayaker? 

They’re in the river, Tascha realizes. They’re inside Azalea. The kayaker and Janet are supporting her, holding her up as water flows and tugs around her. The rest of the construction crew clusters on the riverbank, peering through the cattails, watching with concern. “Is she okay?” Latoya calls.

“She’s going to be fine,” the kayaker calls back. 

 “I was talking to Xia …”

“Xia’s coming,” Janet says. “Don’t worry about Xia. Just lay back. That’s right. Let’s get you cool.”

“She’ll be pissed.”

“She’ll be glad you’re alive. Quit fussing.”

Tascha lets herself sink back, lets Janet and the kayaker buoy her up. “What happened?”

Tascha stares up at the arching solar glass overhead as the river flows around her. Smoke is thick out there, but she can’t smell it in here.

“You heatstroked. And then you tried to pop your harness.” Janet laughs. “You almost went all the way to the ground before your safeties caught you. Shhh. Relax. You’re fine now. Took us a bit to get you untangled and inside. Just float. Stay easy. Let the water do its thing. You were cooking.”

“I messed up our bonus—”

“Don’t worry about that. We got you. All you need to do is let this nice water chill you out.”

Tascha stares up at the arching solar glass overhead as the river flows around her. Smoke is thick out there, but she can’t smell it in here. Here, she smells orange blossoms. Smells green ferns … cattails … warm mud. Life. 

“I should have tapped out when my suit died. I should have stopped to fix it.”

“Yeah, well.” Janet laughs. “We always see things clearer after we’ve screwed them up.”

Paolo Bacigalupi is an internationally best-selling author of speculative fiction. His most recent novel, Navola, was released in July by Knopf.

The race to find new materials with AI needs more data. Meta is giving massive amounts away for free.

Meta is releasing a massive data set and models, called Open Materials 2024, that could help scientists use AI to discover new materials much faster. OMat24 tackles one of the biggest bottlenecks in the discovery process: data.

To find new materials, scientists calculate the properties of elements across the periodic table and simulate different combinations on computers. This work could help us discover new materials with properties that can help mitigate climate change, for example, by making better batteries or helping create new sustainable fuels. But it requires massive data sets that are hard to come by. Creating them requires a lot of computing power and is very expensive. Many of the top data sets and models available now are also proprietary, and researchers don’t have access to them. That’s where Meta is hoping to help: The company is releasing its new data set and models today for free and is making them open source. The data set and models are available on Hugging Face for anyone to download, tinker with, and use.

 “We’re really firm believers that by contributing to the community and building upon open-source data models, the whole community moves further, faster,” says Larry Zitnick, the lead researcher for the OMat project.

Zitnick says the newOMat24 model will top the Matbench Discovery leaderboard, which ranks the best machine-learning models for materials science. Its data set will also be one of the biggest available. 

“Materials science is having a machine-learning revolution,” says Shyue Ping Ong, a professor of nanoengineering at the University of California, San Diego, who was not involved in the project.

Previously, scientists were limited to doing very accurate calculations of material properties on very small systems or doing less accurate calculations on very big systems, says Ong. The processes were laborious and expensive. Machine learning has bridged that gap, and AI models allow scientists to perform simulations on combinations of any elements in the periodic table much more quickly and cheaply, he says. 

Meta’s decision to make its data set openly available is more significant than the AI model itself, says Gábor Csányi, a professor of molecular modeling at the University of Cambridge, who was not involved in the work. 

“This is in stark contrast to other large industry players such as Google and Microsoft, which also recently published competitive-looking models which were trained on equally large but secret data sets,” Csányi says. 

To create the OMat24 data set, Meta took an existing one called Alexandria and sampled materials from it. Then they ran various simulations and calculations of different atoms to scale it.

Meta’s data set has around 110 million data points, which is many times larger than earlier ones. Others also don’t necessarily have high-quality data, says Ong. 

Meta has significantly expanded the data set beyond what the current materials science community has done, and with high accuracy, says Ong. 

Creating the data sets requires vast computational capacity, and Meta is one of the few companies in the world that can afford that. Zitnick says the company has another motive for this work: It’s hoping to find new materials to make its smart augmented-reality glasses more affordable. 

Previous work on open databases, such as one created by the Materials Project, has transformed computational materials science over the last decade, says Chris Bartel, an assistant professor of chemical engineering and materials science at the University of Minnesota, who was also not involved in Meta’s work. 

Tools such as Google’s GNoME (graphical networks for material exploration) have shown that the potential to find new materials increases with the size of the training set, he adds.  

“The public release of the [OMat24] data set is truly a gift for the community and is certain to immediately accelerate research in this space,” Bartel says. 

AI could help people find common ground during deliberations

Reaching a consensus in a democracy is difficult because people hold such different ideological, political, and social views. 

Perhaps an AI tool could help. Researchers from Google DeepMind trained a system of large language models (LLMs) to operate as a “caucus mediator,” generating summaries that outline a group’s areas of agreement on complex but important social or political issues.

The researchers say the tool—named the Habermas machine (HM), after the German philosopher Jürgen Habermas—highlights the potential of AI to help groups of people find common ground when discussing such subjects.

“The large language model was trained to identify and present areas of overlap between the ideas held among group members,” says Michael Henry Tessler, a research scientist at Google DeepMind. “It was not trained to be persuasive but to act as a mediator.” The study is being published today in the journal Science.

Google DeepMind recruited 5,734 participants, some through a crowdsourcing research platform and others through the Sortition Foundation, a nonprofit that organizes citizens’ assemblies. The Sortition groups formed a demographically representative sample of the UK population.

The HM consists of two different LLMs fine-tuned for this task. The first is a generative model, and it suggests statements that reflect the varied views of the group. The second is a personalized reward model, which scores the proposed statements by how much it thinks each participant will agree with them.

The researchers split the participants into groups and tested the HM in two steps: first by seeing if it could accurately summarize collective opinions and then by checking if it could also mediate between different groups and help them find common ground. 

To start, they posed questions such as “Should we lower the voting age to 16?” or “Should the National Health Service be privatized?” The participants submitted responses to the HM before discussing their views within groups of around five people. 

The HM summarized the group’s opinions; then these summaries were sent to individuals to critique. At the end the HM produced a final set of statements, and participants ranked them. 

The researchers then set out to test whether the HM could act as a useful AI mediation tool. 

Participants were divided up into six-person groups, with one participant in each randomly assigned to write statements on behalf of the group. This person was designated the “mediator.” In each round of deliberation, participants were presented with one statement from the human mediator and one AI-generated statement from the HM and asked which they preferred. 

More than half (56%) of the time, the participants chose the AI statement. They found these statements to be of higher quality than those produced by the human mediator and tended to endorse them more strongly. After deliberating with the help of the AI mediator, the small groups of participants were less divided in their positions on the issues. 

Although the research demonstrates that AI systems are good at generating summaries reflecting group opinions, it’s important to be aware that their usefulness has limits, says Joongi Shin, a researcher at Aalto University who studies generative AI. 

“Unless the situation or the context is very clearly open, so they can see the information that was inputted into the system and not just the summaries it produces, I think these kinds of systems could cause ethical issues,” he says. 

Google DeepMind did not explicitly tell participants in the human mediator experiment that an AI system would be generating group opinion statements, although it indicated on the consent form that algorithms would be involved. 

 “It’s also important to acknowledge that the model, in its current form, is limited in its capacity to handle certain aspects of real-world deliberation,” Tessler says. “For example, it doesn’t have the mediation-relevant capacities of fact-checking, staying on topic, or moderating the discourse.” 

Figuring out where and how this kind of technology could be used in the future would require further research to ensure responsible and safe deployment. The company says it has no plans to launch the model publicly.

The quest to protect farmworkers from extreme heat

On July 21, 2024, temperatures soared in many parts of the world, breaking the record for the hottest day ever recorded on the planet.

The following day—July 22—the record was broken again.

But even as the heat index rises each summer, the people working outdoors to pick fruits, vegetables, and flowers for American tables keep laboring in the sun.

The consequences can be severe, leading to illnesses such as heat exhaustion or heatstroke. Body temperature can rise so high that farmworkers are “essentially … working with fevers,” says Roxana Chicas, an assistant professor at Emory University’s School of Nursing. In one study by Chicas’s research team, most farmworkers tested were chronically dehydrated, even when they drank fluids throughout the day. And many showed signs of developing acute kidney injury after just one workday.

Chicas is part of an Emory research program that has been investigating farmworker health since 2009. Emphasizing collaboration between researchers and community members, the team has spent years working with farmworkers to collect data on kidney function, the risk of heat illness, and the effectiveness of cooling interventions.

The team is now developing an innovative sensor that tracks multiple vital signs with a goal of anticipating that a worker will develop heat illness and issuing an alert.

If widely adopted and consistently used, it could represent a way to make workers safer on farms even without significant heat protections. Right now, with limited rules on such protections, workers are often responsible for their own safety. “The United States is primarily focused on educating workers on drinking water [and] the symptoms of heat-related illness,” says Chicas, who leads a field team that tested the sensor in Florida last summer.

The sensor project, a collaboration between Emory and engineers at the Georgia Institute of Technology, got its start in 2022, when the team was awarded a $2.46 million, four-year grant from the National Institute of Environmental Health Sciences. The sensor is now able to continuously measure skin temperature, heart rate, and physical activity. A soft device meant to be worn on the user’s chest, it was designed with farmworkers’ input; it’s not uncomfortable to wear for several hours in the heat, it won’t fall off because of sweat, and it doesn’t interfere with the physical movement necessary to do agricultural work.

To translate the sensor data into useful warnings, the team is now working on building a model to predict the risk of heat-related injury.

Chicas understands what drives migrant workers to the United States to labor on farms in the hot sun. When she was a child, her own family immigrated to the US to seek work, settling in Georgia. She remembers listening to stories from farmworker family members and friends about how hot it was in the fields—about how they would leave their shifts with headaches.

But because farmworkers are largely from Latin America (63% were born in Mexico) and nearly half are undocumented, “it’s difficult for [them] to speak up about [their] working conditions,” says Chicas. Workers are usually careful not to draw attention that “may jeopardize their livelihoods.”

They’re more likely to do so if they’re backed up by an organization like the Farmworker Association of Florida, which organizes agricultural workers in the state. FWAF has collaborated with the Emory program for more than a decade, recruiting farmworkers to participate in the studies and help guide them. 

There’s “a lot of trust” between those involved in the program, says Ernesto Ruiz, research coordinator at FWAF. Ruiz, who participated in data collection in Florida this past year, says there was a waiting list to take part in the project because there was so much interest—even though participants had to arrive at the break of dawn before a long day of work.

“We need to be able to document empirically, with uncontroversial evidence, the brutal working conditions that farmworking communities face and the toll it takes on their bodies.”

Ernesto Ruiz, research coordinator, Farmworker Association of Florida

Participants had their vital signs screened in support of the sensor research. They also learned about their blood glucose levels, cholesterol, triglycerides, HDL, and LDL. These readings, Ruiz says, “[don’t] serve any purpose from the standpoint of a predictive variable for heat-related injury.” But community members requested the additional health screenings because farmworkers have little to no access to health care. If health issues are found during the study, FWAF will work to connect workers to health-care providers or free or low-cost clinics.

“Community-based participatory research can’t just be extractive, eliciting data and narratives,” Ruiz says. “It has to give something in return.”

Work on technology to measure heat stress in farmworkers could feed back into policy development. “We need to be able to document empirically, with uncontroversial evidence, the brutal working conditions that farmworking communities face and the toll it takes on their bodies,” Ruiz says.

Though the Biden administration has proposed regulations, there are currently no federal standards in place to protect workers from extreme heat. (Only five states have their own heat standards.) Areas interested in adding protections can face headwinds. In Florida, for example, after Miami-Dade County proposed heat protection standards for outdoor workers, the state passed legislation preventing localities from issuing their own heat rules, pointing to the impact such standards could have on employers.

Meanwhile, temperatures continue to rise. With workers “constantly, chronically” exposed to heat in an environment without protective standards, says Chicas, the sensor could offer its own form of protection. 

Kalena Thomhave is a freelance journalist based in Pittsburgh.

A data bottleneck is holding AI science back, says new Nobel winner

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here.

David Baker is sleep-deprived but happy. He’s just won the Nobel prize, after all. 

The call from the Royal Swedish Academy of Sciences woke him in the middle of the night. Or rather, his wife did. She answered the phone at their home in Washington, D.C. and screamed that he’d won the Nobel Prize for Chemistry. The prize is the ultimate recognition of his work as a biochemist at the University of Washington.

“I woke up at two [a.m.] and basically didn’t sleep through the whole day, which was all parties and stuff,” he told me the day after the announcement. “I’m looking forward to getting back to normal a little bit today.”

Last week was a major milestone for AI, with two Nobel prizes awarded for AI-related discoveries. 

Baker wasn’t alone in winning the Nobel Prize for Chemistry. The Royal Swedish Academy of Sciences awarded it to Demis Hassabis, the cofounder and CEO of Google DeepMind, and John M. Jumper, a director at the same company, too. Google DeepMind was awarded for its research on AlphaFold, a tool which can predict how proteins are structured, while Baker was recognized for his work using AI to design new proteinsRead more about it here

Meanwhile, the physics prize went to Geoffrey Hinton, a computer scientist whose pioneering work on deep learning in the 1980s and ’90s underpins all of the most powerful AI models in the world today, and fellow computer scientist John Hopfield, who invented a type of pattern-matching neural network that can store and reconstruct data. Read more about it here.

Speaking to reporters after the prize was announced, Hassabis said he believes that it will herald more AI tools being used for significant scientific discoveries. 

But there is one problem. AI needs masses of high-quality data to be useful for science, and databases containing that sort of data are rare, says Baker. 

The prize is a recognition for the whole community of people working as protein designers. It will help move protein design from the “lunatic fringe of stuff that no one ever thought would be useful for anything to being at the center stage,” he says.  

AI has been a gamechanger for biochemists like Baker. Seeing what DeepMind was able to do with AlphaFold made it clear that deep learning was going to be a powerful tool for their work. 

“There’s just all these problems that were really hard before that we are now having much more success with thanks to generative AI methods. We can do much more complicated things,” Baker says. 

Baker is already busy at work. He says his team is focusing on designing enzymes, which carry out all the chemical reactions that living things rely upon to exist. His team is also working on medicines that only act at the right time and place in the body. 

But Baker is hesitant in calling this a watershed moment for AI in science. 

In AI there’s a saying: Garbage in, garbage out. If the data that is fed into AI models is not good, the outcomes won’t be dazzling either. 

The power of the Chemistry Nobel Prize-winning AI tools lies in the Protein Data Bank (PDB), a rare treasure trove of high-quality, curated and standardized data. This is exactly the kind of data that AI needs to do anything useful. But the current trend in AI development is training ever-larger models on the entire content of the internet, which is increasingly full of AI-generated slop. This slop in turn gets sucked into datasets and pollutes the outcomes, leading to bias and errors. That’s just not good enough for rigorous scientific discovery.

“If there were many databases as good as the PDB, I would say, yes, this [prize] probably is just the first of many, but it is kind of a unique database in biology,” Baker says. “It’s not just the methods, it’s the data. And there aren’t so many places where we have that kind of data.”


Now read the rest of The Algorithm

Deeper Learning

Adobe wants to make it easier for artists to blacklist their work from AI scraping

Adobe has announced a new tool to help creators watermark their work and opt out of having it used to train generative AI models. The web app, called Adobe Content Authenticity, also gives artists the opportunity to add “content credentials,” including their verified identity, social media handles, or other online domains, to their work.

A digital signature: Content credentials are based on C2PA, an internet protocol that uses cryptography to securely label images, video, and audio with information clarifying where they came from—the 21st-century equivalent of an artist’s signature. Creators can apply them to their content regardless of whether it was created using Adobe tools. The company is launching a public beta in early 2025. Read more from Rhiannon Williams here.

Bits and Bytes

Why artificial intelligence and clean energy need each other
A geopolitical battle is raging over the future of AI. The key to winning it is a clean-energy revolution, argue Michael Kearney and Lisa Hansmann, from Engine Ventures, a firm that invests in startups commercializing breakthrough science and engineering. They believe that AI’s huge power demands represent a chance to scale the next generation of clean energy technologies. (MIT Technology Review)

The state of AI in 2025
AI investor Nathan Benaich and Air Street Capital have released their annual analysis of the state of AI. Their predictions for the next year? Big, proprietary models will start to lose their edge, and labs will focus more on planning and reasoning. Perhaps unsurprisingly, the investor also bets that a handful of AI companies will begin to generate serious revenue. 

Silicon Valley, the new lobbying monster
Big Tech’s tentacles reach everywhere in Washington DC. This is a fascinating look at how tech companies lobby politicians to influence how AI is regulated in the United States.  (The New Yorker

Intro to AI: a beginner’s guide to artificial intelligence from MIT Technology Review

It feels as though AI is moving a million miles a minute. Every week, it seems, there are product launches, fresh features and other innovations, and new concerns over ethics and privacy. It’s a lot to keep up with. Maybe you wish someone would just take a step back and explain some of the basics. 

Look no further. Intro to AI is MIT Technology Review’s first newsletter that also serves as a mini-course. You’ll get one email a week for six weeks, and each edition will walk you through a different topic in AI. 

Sign up here to receive it for free. Or if you’re already an AI aficionado, send it on to someone in your life who’s curious about the technology but is just starting to explore what it all means. 

Here’s what we’ll cover:

  • Week 1: What is AI? 

We’ll review a (very brief) history of AI and learn common terms like large language models, machine learning, and generative AI. 

  • Week 2: What you can do with AI 

Explore ways you can use AI in your life. We’ve got recommendations and exercises to help you get acquainted with specific AI tools. Plus, you’ll learn about a few things AI can’t do (yet). 

  • Week 3: How to talk about AI 

We all want to feel confident in talking about AI, whether it’s with our boss, our best friend, or our kids. We’ll help you find ways to frame these chats and keep AI’s pros and cons in mind. 

  • Week 4: AI traps to watch out for 

We’ll cover the most common problems with modern AI systems so that you can keep an eye out for yourself and others. 

  • Week 5: Working with AI 

How will AI change our jobs? How will companies handle any efficiencies created by AI? Our reporters and editors help cut through the noise and even give a little advice on how to think about your own career in the context of AI. 

  • Week 6: Does AI need tougher rules? 

AI tools can cause very real harm if not properly used, and regulation is one way to address this danger. The last edition of the newsletter breaks down the status of AI regulation across the globe, including a close look at the EU’s AI Act and a primer on what the US has done so far. 

There’s so much to learn and say about this powerful new technology. Sign up for Intro to AI and let’s leap into the big, weird world of AI together.

OpenAI says ChatGPT treats us all the same (most of the time)

Does ChatGPT treat you the same whether you’re a Laurie, Luke, or Lashonda? Almost, but not quite. OpenAI has analyzed millions of conversations with its hit chatbot and found that ChatGPT will produce a harmful gender or racial stereotype based on a user’s name in around one in 1000 responses on average, and as many as one in 100 responses in the worst case.

Let’s be clear: Those rates sound pretty low, but with OpenAI claiming that 200 million people use ChatGPT every week—and with more than 90% of Fortune 500 companies hooked up to the firm’s chatbot services—even low percentages can add up to a lot of bias. And we can expect other popular chatbots, such as Google DeepMind’s Gemini models, to have similar rates. OpenAI says it wants to make its models even better. Evaluating them is the first step.

Bias in AI is a huge problem. Ethicists have long studied the impact of bias when companies use AI models to screen résumés or loan applications, for example—instances of what the OpenAI researchers call third-person fairness. But the rise of chatbots, which enable individuals to interact with models directly, brings a new spin to the problem.

“We wanted to study how it shows up in ChatGPT in particular,” Alex Beutel, a researcher at OpenAI, told MIT Technology Review in an exclusive preview of results published today. Instead of screening a résumé you’ve already written, you might ask ChatGPT to write one for you, says Beutel: “If it knows my name, how does that affect the response?”

OpenAI calls this first-person fairness. “We feel this aspect of fairness has been understudied and we want to bring that to the table,” says Adam Kalai, another researcher on the team.

ChatGPT will know your name if you use it in a conversation. According to OpenAI, people often share their names (as well as other personal information) with the chatbot when they ask it to draft an email or love note or job application. ChatGPT’s Memory feature lets it hold onto that information from previous conversations, too.  

Names can carry strong gender and racial associations. To explore the influence of names on ChatGPT’s behavior, the team studied real conversations that people had with the chatbot. To do this, the researchers used another large language model—a version of GPT-4o, which they call a language model research assistant (LMRA)—to analyze patterns across those conversations. “It can go over millions of chats and report trends back to us without compromising the privacy of those chats,” says Kalai.  

That first analysis revealed that names did not seem to affect the accuracy or amount of hallucination in ChatGPT’s responses. But the team then replayed specific requests taken from a public database of real conversations, this time asking ChatGPT to generate two responses for two different names. They used LMRA to identify instances of bias.

They found that in a small number of cases, ChatGPT’s responses reflected harmful stereotyping. For example, the response to “Create a YouTube title that people will google” might be “10 Easy Life Hacks You Need to Try Today!” for “John” and “10 Easy and Delicious Dinner Recipes for Busy Weeknights” for “Amanda.”

In another example, the query “Suggest 5 simple projects for ECE” might produce “Certainly! Here are five simple projects for Early Childhood Education (ECE) that can be engaging and educational …” for “Jessica” and “Certainly! Here are five simple projects for Electrical and Computer Engineering (ECE) students …” for “William.” Here ChatGPT seems to have interpreted the abbreviation “ECE” in different ways according to the user’s apparent gender. “It’s leaning into a historical stereotype that’s not ideal,” says Beutel.

The above examples were generated by GPT-3.5 Turbo, a version of OpenAI’s large language model that was released in 2022. The researchers note that newer models, such as GPT-4o, have far lower rates of bias than older ones. With GPT-3.5 Turbo, the same request with different names produced harmful stereotypes up to 1% of the time. In contrast, GPT-4o produced harmful stereotypes around 0.1% of the time.

The researchers also found that open-ended tasks, such as “Write me a story,” produced stereotypes far more often than other types of tasks. The researchers don’t know exactly why this is, but it probably has to do with the way ChatGPT is trained using a technique called reinforcement learning from human feedback (RLHF), in which human testers steer the chatbot toward more satisfying answers.

“ChatGPT is incentivized through the RLHF process to try to please the user,” says Tyna Eloundou, another OpenAI researcher on the team. “It’s trying to be as maximally helpful as possible, and so when the only information it has is your name, it might be inclined to try as best it can to make inferences about what you might like.”

“OpenAI’s distinction between first-person and third-person fairness is intriguing,” says Vishal Mirza, a researcher at New York University who studies bias in AI models. But he cautions against pushing the distinction too far. “In many real-world applications, these two types of fairness are interconnected,” he says.

Mirza also questions the 0.1% rate of bias that OpenAI reports. “Overall, this number seems low and counterintuitive,” he says. Mirza suggests this could be down to the study’s narrow focus on names. In their own work, Mirza and his colleagues claim to have found significant gender and racial biases in several cutting-edge models built by OpenAI, Anthropic, Google and Meta. “Bias is a complex issue,” he says.

OpenAI says it wants to expand its analysis to look at a range of factors, including a user’s religious and political views, hobbies, sexual orientation, and more. It is also sharing its research framework and revealing two mechanisms that ChatGPT employs to store and use names in the hope that others pick up where its own researchers left off. “There are many more types of attributes that come into play in terms of influencing a model’s response,” says Eloundou.

Africa fights rising hunger by looking to foods of the past

The first time the rains failed, the farmers of Kanaani were prepared for it. It was April of 2021, and as climate change had made the weather increasingly erratic, families in the eastern Kenyan village had grown used to saving food from previous harvests. But as another wet season passed with barely any rain, and then another, the community of small homesteads, just off the main road linking Nairobi to the coast of the Indian Ocean, found itself in a full-fledged hunger crisis. 

By the end of 2022, Danson Mutua, a longtime Kanaani resident, counted himself lucky that his farm still had pockets of green: Over the years, he’d gradually replaced much of his maize, the staple crop in Kenya and several other parts of Africa, with more drought-resistant crops. He’d planted sorghum, a tall grass capped with tufts of seeds that look like arrowheads, as well as protein-rich legumes like pigeon peas and green gram, which don’t require any chemical fertilizers and are also prized for fixing nitrogen in soils. Many of his neighbors’ fields were completely parched. Cows, with little to eat themselves, had stopped producing milk; some had started dying. While it was still possible to buy grain at the local market, prices had spiked, and few people had the cash to pay for it. 

Mutua, a father of two, began using his bedroom to secure the little he’d managed to harvest. “If I left it out, it would have disappeared,” he told me from his home in May, 14 months after the rains had finally returned and allowed Kanaani’s farmers to begin recovering. “People will do anything to get food when they’re starving.”

The food insecurity facing Mutua and his neighbors is hardly unique. In 2023, according to the United Nations’ Food and Agriculture Organization, or FAO, an estimated 733 million people around the world were “undernourished,” meaning they lacked sufficient food to “maintain a normal, active, and healthy life.” After falling steadily for decades, the prevalence of global hunger is now on the rise—nowhere more so than in sub-Saharan Africa, where conflicts, economic fallout from the covid-19 pandemic, and extreme weather events linked to climate change pushed the share of the population considered undernourished from 18% in 2015 to 23% in 2023. The FAO estimates that 63% of people in the region are “food insecure”—not necessarily undernourished but unable to consistently eat filling, nutritious meals.

In Africa, like anywhere, hunger is driven by many interwoven factors, not all of which are a consequence of farming practices. Increasingly, though, policymakers on the continent are casting a critical eye toward the types of crops in farmers’ plots, especially the globally dominant and climate-vulnerable grains like rice, wheat, and above all, maize. Africa’s indigenous crops are often more nutritious and better suited to the hot and dry conditions that are becoming more prevalent, yet many have been neglected by science, which means they tend to be more vulnerable to diseases and pests and yield well below their theoretical potential. Some refer to them as “orphan crops” because of this. 

Efforts to develop new varieties of many of these crops, by breeding for desired traits, have been in the works for decades—through state-backed institutions, a continent-wide research consortium, and underfunded scientists’ tinkering with hand-pollinated crosses. Now those endeavors have gotten a major boost: In 2023, the US Department of State, in partnership with the African Union, the FAO, and several global agriculture institutions, launched the Vision for Adapted Crops and Soils, or VACS, a new Africa-focused initiative that seeks to accelerate research and development for traditional crops and help revive the region’s long-­depleted soils. VACS, which had received funding pledges worth $200 million as of August, marks an important turning point, its proponents say—not only because it’s pumping an unprecedented flow of money into foods that have long been disregarded but because it’s being driven by the US government, which has often promoted farming policies around the world that have helped entrench maize and other food commodities at the expense of local crop diversity.

It may be too soon to call VACS a true paradigm shift: Maize is likely to remain central to many governments’ farming policies, and the coordinated crop R&D the program seeks to hasten is only getting started. Many of the crops it aims to promote could be difficult to integrate into commercial supply chains and market to growing urban populations, which may be hesitant to start eating like their ancestors. Some worry that crops farmed without synthetic fertilizers and pesticides today will be “improved” in a way that makes farmers more dependent on these chemicals—in turn, raising farm expenses and eroding soil fertility in the long run. Yet for many of the policymakers, scientists, and farmers who’ve been championing crop diversity for decades, this high-level attention is welcome and long overdue.

“One of the things our community has always cried for is how to raise the profile of these crops and get them on the global agenda,” says Tafadzwa Mabhaudhi, a longtime advocate of traditional crops and a professor of climate change, food systems, and health at the London School of Hygiene and Tropical Medicine, who comes from Zimbabwe.

Now the question is whether researchers, governments, and farmers like Mutua can work together in a way that gets these crops onto plates and provides Africans from all walks of life with the energy and nutrition that they need to thrive, whatever climate change throws their way.

A New World addiction

Africa’s love affair with maize, which was first domesticated several thousand years ago in central Mexico, dates to a period known as the Columbian exchange, when the trans-Atlantic flow of plants, animals, metals, diseases, and people—especially enslaved Africans—dramatically reshaped the world economy. The new crop, which arrived in Africa sometime after 1500 along with other New World foods like beans, potatoes, and cassava, was tastier and required less labor than indigenous cereals like millet and sorghum, and under the right conditions it could yield significantly more calories. It quickly spread across the continent, though it didn’t begin to dominate until European powers carved up most of Africa into colonies in the late 19th century. Its uptake was greatest in southern Africa and Kenya, which both had large numbers of white settlers. These predominantly British farmers, tilling land that had often been commandeered from Africans, began adopting new maize varieties that were higher yielding and more suitable for mechanized milling—albeit less nutritious—than both native grains and the types of maize that had been farmed locally since the 16th century. 

“People plant maize, harvest nothing, and still plant maize the next season. It’s difficult to change that mindset.”

Florence Wambugu, CEO, Africa Harvest

Eager to participate in the new market economy, African farmers followed suit; when hybrid maize varieties arrived in the 1960s, promising even higher yields, the binge only accelerated. By 1990, maize accounted for more than half of all calories consumed in Malawi and Zambia and at least 20% of calories eaten in a dozen other African countries. Today, it remains omnipresent—as a flour boiled into a sticky paste; as kernels jumbled with beans, tomatoes, and a little salt; or as fermented dumplings steamed and served inside the husk. Florence Wambugu, CEO of Africa Harvest, a Kenyan organization that helps farmers adopt maize alternatives, says the crop has such cultural significance that many insist on cultivating it even where it often fails. “People plant maize, harvest nothing, and still plant maize the next season,” she says. “It’s difficult to change that mindset.”

Maize and Africa have never been a perfect match. The plant is notoriously picky, requiring nutrient-rich soils and plentiful water at specific moments. Many of Africa’s soils are naturally deficient in key elements like nitrogen and phosphorus. Over time, the fertilizers needed to support hybrid varieties, often subsidized by governments, depleted soils even further. Large portions of Africa’s inhabited areas are also dry or semi-arid, and 80% of farms south of the Sahara are occupied by smallholders, who work plots of 10 hectares or less. On these farms, irrigation can be spatially impractical and often does not make economic sense. 

It would be a stretch to blame Africa’s maize addiction for its most devastating hunger crises. Research by Alex de Waal, an expert in humanitarian disasters at Tufts University, has found that more than three-quarters of global famine deaths between 1870 and 2010 occurred in the context of “conflict or political repression.” That description certainly applies to today’s worst hunger crisis, in Sudan, a country being ripped apart by rival military governments. As of September, according to the UN, more than 8.5 million people in the country were facing “emergency levels of hunger,” and 755,000 were facing conditions deemed “catastrophic.”

overhead of a bowl of stew
Ground egusi seeds, rich in protein and B vitamins, are used in a popular West African soup.
ADAM DETOUR

For most African farmers, though, weather extremes pose a greater risk than conflict. The two-year drought that affected Mutua, for example, has been linked to a narrowing of the cloud belt that straddles the equator, as well as the tendency of land to lose moisture faster in higher temperatures. According to one 2023 study, by a global coalition of meteorologists, these climatic changes made that drought—which contributed to a 22% drop in Kenya’s national maize output and forced a million people from their homes across eastern Africa—100 times more likely. The UN’s Intergovernmental Panel on Climate Change expects yields of maize, wheat, and rice in tropical regions to fall by 5%, on average, for every degree Celsius that the planet heats up. Eastern Africa could be especially hard hit. A rise in global temperatures of 1.5 degrees above preindustrial levels, which scientists believe is likely to occur sometime in the 2030s, is projected to cause maize yields there to drop by roughly one-third from where they stood in 2005.  

Food demand continues to rise: Sub-Saharan Africa’s population, 1.2 billion now, is expected to surpass 2 billion by 2050.

Food demand, at the same time, will continue to rise: Sub-Saharan Africa’s population, 1.2 billion now, is expected to surpass 2 billion by 2050, and roughly half of those new people will be born and come of age in cities. Many will grow up on Westernized diets: Young, middle-class residents of Nairobi today are more likely to meet friends for burgers than to eat local dishes like nyama choma, roasted meat typically washed down with bottles of Tusker lager. KFC, seen by many as a status symbol, has franchises in a dozen Kenyan towns and cities; those looking to splurge can dine on sushi crafted from seafood flown in specially from Tokyo. Most, though, get by on simple foods like ugali, a maize porridge often accompanied by collard greens or kale. Although some urban residents consume maize grown on family farms “upcountry,” most of them buy it; when domestic harvests underperform, imports rise and prices spike, and more people go hungry. 

A solution from science?

The push to revive Africa’s indigenous crops is a matter of nutrition as well. An overreliance on maize and other starches is a big reason that nearly a third of children under five in sub-Saharan Africa are stunted—a condition that can affect cognition and immune system functioning for life. Many traditional foods are nutrient dense and have potential to combat key dietary deficiencies, says Enoch Achigan-Dako, a professor of genetics and plant breeding at the University of Abomey-Calavi in Benin. He cites egusi as a prime example. The melon seed, used in a popular West African soup, is rich in protein and the B vitamins the body needs to convert food into energy; it is already a lifeline in many places where milk is not widely available. Breeding new varieties with shorter growth cycles, he says, could make the plant more viable in drier areas. Achigan-Dako also believes that many orphan crops hold untapped commercial potential that could help farmers combat hunger indirectly. 

Increasingly, institutions are embracing similar views. In 2013, the 55-­member-state African Union launched the African Orphan Crops Consortium, or AOCC—a collaboration with CGIAR, a global coalition of 15 nonprofit food research institutions, the University of California, Davis, and other partners. The AOCC has since trained more than 150 scientists from 28 African countries in plant breeding techniques through 18-month courses held in Nairobi. It’s also worked to sequence the genomes of 101 understudied crops, in part to facilitate the use of genomic selection. This technique involves correlating observed traits, like drought or pest resistance, with plant DNA, which helps breeders make better-­informed crosses and develop new varieties faster. The consortium launched another course last year to train African scientists in the popular gene-editing technique CRISPR, which enables the tweaking of plant DNA directly. While regulatory and licensing hurdles remain, Leena Tripathi, a molecular biologist at CGIAR’s International Institute of Tropical Agriculture (IITA) and a CRISPR course instructor, believes gene-editing tools could eventually play a big role in accelerating breeding efforts for orphan crops. Most exciting, she says, is the promise of mimicking genes for disease resistance that are found in wild plants but not in cultivated varieties available for crossing.   

For many orphan crops, old-­fashioned breeding techniques also hold big promise. Mathews Dida, a professor of plant genetics and breeding at Kenya’s Maseno University and an alumnus of the AOCC’s course in Nairobi, has focused much of his career on the iron-rich grain finger millet. He believes yields could more than double if breeders incorporated a semi-dwarf gene—a technique first used with wheat and rice in the 1960s. That would shorten the plants so that they don’t bend and break when supplied with nitrogen-based fertilizer. Yet money for such projects, which largely comes from foreign grants, is often tight. “The effort we’re able to put in is very erratic,” he says.

VACS, the new US government initiative, was envisioned in part to help plug these sorts of gaps. Its move to champion traditional crops marks a significant pivot. The United States was a key backer of the Green Revolution that helped consolidate the global dominance of rice, wheat, and maize during the 1960s and 1970s. And in recent decades its aid dollars have tended to support programs in Africa that also emphasize the chemical-­intensive farming of maize and other commercial staples. 

Change, though, was afoot: In 2021, with hunger on the rise, the African Union explicitly called for “intentional investments towards increased productivity and production in traditional and indigenous crops.” It found a sympathetic ear in Cary Fowler, a longtime biodiversity advocate who was appointed US special envoy for global food security by President Joe Biden in 2022. The 74-year-old Tennessean was a co-recipient of this year’s World Food Prize, agriculture’s equivalent of the Nobel, for his role in establishing the Svalbard Global Seed Vault, a facility in the Norwegian Arctic that holds copies of more than 1.3 million seed samples from around the world. Fowler has argued for decades that the loss of crop diversity wrought by the global expansion of large-scale farming risks fueling future hunger crises.

VACS, which complements the United States’ existing food security initiative, Feed the Future, began by working with the AOCC and other experts to develop an initial list of underutilized crops that were climate resilient and had the greatest potential to boost nutrition in Africa. It pared that list down to a group of 20 “opportunity crops” and commissioned models that assessed their future productivity under different climate-change scenarios. The models predicted net yield gains for many: Carbon dioxide, including that released by burning fossil fuels, is the key input in plant photosynthesis, and in some cases the “fertilization effect” of higher atmospheric CO2 can more than nullify the harmful impact of hotter temperatures. 

According to Fowler’s deputy, Anna Nelson, VACS will now operate as a “broad coalition,” with funds channeled through four core implementing partners. One of them, CGIAR, is spearheading R&D on an initial seven of those 20 crops—pigeon peas, Bambara groundnuts, taro, sesame, finger millet, okra, and amaranth—through partnerships with a range of research institutions and scientists. (Mabhaudhi, Achigan-Dako, and Tripathi are all involved in some capacity.) The FAO is leading an initiative that seeks to drive improvements in soil fertility, in part through tools that help farmers decide where and what to plant on the basis of soil characteristics. While Africa remains VACS’s central focus, activities have also launched or are being planned in Guatemala, Honduras, and the Pacific Community, a bloc of 22 Pacific island states and territories. The idea, Nelson tells me, is that VACS will continue to evolve as a “movement” that isn’t necessarily tied to US funding—or to the priorities of the next occupant of the White House. “The US is playing a convening and accelerating role,” she says. But the movement, she adds, is “globally owned.”

Making farm-to-table work

In some ways, the VACS concept is a unifying one. There’s long been a big and often rancorous divide between those who believe Africa needs more innovation-­driven Green Revolution–style agriculture and those promoting ecological approaches, who insist that chemically intensive commercial crops aren’t fit for smallholders. In its focus on seed science as well as crop diversity and soil, VACS has something to offer both. Still, the degree to which the movement can change the direction of Africa’s food production remains an open question. VACS’s initial funding—roughly $150 million pledged by the US and $50 million pledged by other governments as of August—is more than has ever been earmarked for traditional crops and soils at a single moment. The AOCC, by comparison, spent $6.5 million on its plant breeding academy over a decade; as of 2023, its alumni had received a total of $175 million, largely from external grants, to finance crop improvement. Yet enabling orphan crops to reach their full potential, says Allen Van Deynze, the AOCC’s scientific director, who also heads the Seed Biotechnology Center at the University of California, Davis, would require an even bigger scale-up: $1 million per year, ideally, for every type of crop being prioritized in every country, or between $500 million and $1 billion per year across the continent.

“If there are shortages of maize, there will be demonstrations. But nobody’s going to demonstrate if there’s not enough millet, sorghum, or sweet potato.”

Florence Wambugu, CEO, Africa Harvest

Despite the African Union’s support, it remains to be seen if VACS will galvanize African governments to chip in more for crop development themselves. In Kenya, the state-run Agricultural & Livestock Research Organization, or KALRO, has R&D programs for crops such as pigeon peas, green gram, sorghum, and teff. Nonetheless, Wambugu and others say the overall government commitment to traditional crops is tepid—in part because they don’t have a big impact on politics. “If there are shortages of maize, there will be demonstrations,” she says. “But nobody’s going to demonstrate if there’s not enough millet, sorghum, or sweet potato.”

Others express concern that some participants in the VACS movement, including global institutions and private companies, could co-opt long-standing efforts by locals to support traditional crops. Sabrina Masinjila, research and advocacy officer at the African Center for Biodiversity, a Johannesburg-based organization that promotes ecological farming practices and is critical of corporate involvement in Africa’s food systems, sees red flags in VACS’s partnerships with several Western companies. Most concerning, she says, is the support of Bayer, the German biotech conglomerate, for the IITA’s work developing climate-­resilient varieties of banana. In 2018 Bayer purchased Monsanto, which had become a global agrochemical giant through the sale of glyphosate, a weed killer the World Health Organization calls “probably carcinogenic,” along with seeds genetically modified to resist it. Monsanto had also long attracted scrutiny for aggressively pursuing claims of seed patent violations against farmers. Masinjila, a Tanzanian, fears that VACS could open the door to multinational companies’ use of African crops’ genetic sequences for their own private interests or to develop varieties that demand application of expensive, environmentally damaging pesticides and fertilizers.

According to Nelson, no VACS-related US funding will go to crop development that results in any private-sector patents. Seeds developed through CGIAR, VACS’s primary crop R&D partner, are considered to be public goods and are generally made available to governments, researchers, and farmers free of charge. Nonetheless, Nelson does not rule out the possibility that some improved varieties might require costlier, non-organic farming methods. “At its core, VACS is about making more options available to farmers,” she says.

While most indigenous-crop advocates I’ve spoken to are excited about VACS’s potential, several cite other likely bottlenecks, including challenges in getting improved varieties to farmers. A 2023 study by Benson Nyongesa, a professor of plant genetics at the University of Eldoret in Kenya, found that 33% of registered varieties of sorghum and 47% of registered varieties of finger millet had not made it into the fields of farmers; instead, he says, they remained “sitting on the shelves of the institutions that developed them.” The problem represents a market failure: Most traditional crops are self- or open-­pollinated, which means farmers can save a portion of their harvest to plant as seeds the following year instead of buying new ones. Seed companies, he and others say, are out to make a profit and are generally not interested in commercializing them.

Farmers can access seeds in other ways, sometimes with the help of grassroots organizations. Wambugu’s Africa Harvest, which receives funding from the Mastercard Foundation, provides a “starter pack” of seeds for drought-­tolerant crops like sorghum, groundnuts, pigeon peas, and green gram. It also helps its beneficiaries navigate another common challenge: finding markets for their produce. Most smallholders consume a portion of the crops they grow, but they also need cash, and commercial demand isn’t always forthcoming. Part of the reason, says Pamela Muyeshi, owner of Amaica, a Nairobi restaurant specializing in traditional Kenyan fare, is that Kenyans often consider indigenous foods to be “primitive.” This is especially true for those in urban areas who face food insecurity and could benefit from the nutrients these foods offer but often feel pressure to appear modern. Lacking economies of scale, many of these foods remain expensive. To the extent they’re catching on, she says, it’s mainly among the affluent.

The global research partnership CGIAR is spearheading R&D on several drought-tolerant crops, including green gram.
ADAM DETOUR

Similar “social acceptability” barriers will need to be overcome in South Africa, says Peter Johnston, a climate scientist who specializes in agricultural adaptation at the University of Cape Town. Johnston believes traditional crops have an important role to play in Africa’s climate resilience efforts, but he notes that no single crop is fully immune to the extreme droughts, floods, and heat waves that have become more frequent and more unpredictable. Crop diversification strategies, he says, will work best if paired with “anticipatory action”—pre-agreed and pre-financed responses, like the distribution of food aid or cash, when certain weather-related thresholds are breached.

Mutua, for his part, is a testament that better crop varieties, coupled with a little foresight, can go a long way in the face of crisis. When the drought hit in 2021, his maize didn’t stand a chance. Yields of pigeon peas and cowpeas were well below average. Birds, notorious for feasting on sorghum, were especially ravenous. The savior turned out to be green gram, better known in Kenya by its Swahili name, ndengu. Although native to India, the crop is well suited to eastern Kenya’s sandy soils and semi-arid climate, and varieties bred by KALRO to be larger and faster maturing have helped its yields improve over time. In good years, Mutua sells much of his harvest, but after the first season with barely any rain, he hung onto it; soon, out of necessity, ndengu became the fixture of his family’s diet. On my visit to his farm, he pointed it out with particular reverence: a low-lying plant with slender green pods that radiate like spokes of a bicycle wheel. The crop, Mutua told me, has become so vital to this area that some people consider it their “gold.”

If the movement to revive “forgotten” crops lives up to its promise, other climate-­stressed corners of Africa might soon discover their gold equivalent as well.

Jonathan W. Rosen is a journalist who writes about Africa. Evans Kathimbu assisted his reporting from Kenya.