This baby with a head camera helped teach an AI how kids learn language

Human babies are far better at learning than even the very best large language models. To be able to write in passable English, ChatGPT had to be trained on massive data sets that contain millions upon millions of words. Children, on the other hand, have access to only a tiny fraction of that data, yet by age three they’re communicating in quite sophisticated ways.

A team of researchers at New York University wondered if AI could learn like a baby. What could an AI model do when given a far smaller data set—the sights and sounds experienced by a single child learning to talk?

A lot, it turns out. This work, published in Science, not only provides insights into how babies learn but could also lead to better AI models. Read the full story.

—Cassandra Willyard

The next generation of mRNA vaccines is on its way

Japan recently approved a new mRNA vaccine for covid, and this one is pretty exciting. Just like the mRNA vaccines you know and love, it delivers the instructions for making the virus’s spike protein. But here’s what makes it novel: it also tells the body how to make more mRNA. Essentially, it’s self-amplifying.

These kinds of vaccines offer a couple of important advantages over conventional mRNA vaccines, at least in theory. The dose can be much lower, and it’s possible that they will induce a more durable immune response.

The company that makes the Japan-approved vaccine has already filed for approval in Europe. It’s also working on a self-amplifying mRNA vaccine for flu, both seasonal and pandemic. Read the full story.

—Cassandra Willyard

This story is from The Checkup, our weekly newsletter covering health and biotech. Sign up to receive it in your inbox every Thursday.

The must-reads

I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.

1 The world’s largest music label has yanked its artists’ music off TikTok
Universal Music Group claims TikTok is unwilling to compensate musicians appropriately. (The Guardian)
+ Taylor Swift fans are kicking off. (Wired $)
+ Indie record labels don’t like the sound of Apple’s pay plans either. (FT $)

2 Los Angeles has made digital discrimination a crime
Internet providers can no longer withhold deals from marginalized neighborhoods. (The Markup)
+ How climate vulnerability and the digital divide are linked. (MIT Technology Review)

3 Big Tech’s cloud services are raking in serious money
But we still need some concrete use cases for all those AI models. (WSJ $)
+ Unsurprisingly, AI chipmaker Nvidia’s market value is skyrocketing too. (Reuters)
+ An AI data center startup plans to go public off the back of the AI boom. (The Information $)

4 Amazon has released an AI shopping bot called Rufus
To helpfully steer you towards buying more stuff from Amazon. (The Verge)
+ The company’s racing to catch up with its generative AI-embracing rivals. (NYT $)

5 A new kind of plastic can be molded to shift shape
Heating and cooling cause it to take on new properties, meaning a single sheet could become multiple different objects. (NYT $)
+ Google DeepMind’s new AI tool helped create more than 700 new materials. (MIT Technology Review)

6 A prestigious art competition is allowing AI-generated entries
But artists need to own full copyright over their creation. (The Guardian)
+ This artist is dominating AI-generated art. And he’s not happy about it. (MIT Technology Review)

7 What we lost in the switch from analog to digital
A sense of ownership helps us to forge connections with our devices. (FT $)
+ The first musical synthesizers are much older than you may realize. (IEEE Spectrum)

8 Artists are burning out promoting themselves online
It’s a whole extra job on top of their day job. (Vox)

9 Free yourself from the Apple Watch’s tyranny
It’s good to be active, but it’s better to be free. (The Atlantic $)

10 A Tesla that crashed into a Norwegian harbor was rescued by a floating sauna
Insert joke about getting into hot water here. (WP $)

Quote of the day

“We have to move to (Instagram) Reels I fear.” 

—An anonymous TikTok user jokes about abandoning the platform in favor of Meta’s less cool competitor, after the music label Universal pulled its music from TikTok due to a row about compensation, Reuters reports.

The big story

The iPad was meant to revolutionize accessibility. What happened?

June 2023

On April 3, 2010, Steve Jobs debuted the iPad. What for most people was basically a more convenient form factor was something far more consequential for non-speakers: a life-­changing revolution in access to a portable, powerful communication device for just a few hundred dollars.

But a piece of hardware, however impressively designed and engineered, is only as valuable as what a person can do with it. After the iPad’s release, the flood of new, easy-to-use augmentative and alternative communication apps that users were in desperate need of never came.

Today, there are only around half a dozen apps, each retailing for $200 to $300, that ask users to select from menus of crudely drawn icons to produce text and synthesized speech. It’s a depressingly slow pace of development for such an essential human function. Read the full story.

—Julie Kim

We can still have nice things

A place for comfort, fun and distraction in these weird times. (Got any ideas? Drop me a line or tweet ’em at me.)

+ Let’s hear it for the unseen cameramen of extreme sports! (Including the one who shot the video of this guy.)
+ I can’t believe Joni Mitchell’s Court and Spark is 50 years old.
+ This innocent pigeon has finally been released after being detained as a spy.
+ Pst—can you keep a secret?
+ Justice for British food! (It’s actually delicious.)

Welcome back to The Checkup! Today I want to talk about … mRNA vaccines.

I can hear the collective groan from here, but wait—hear me out! I know you’ve heard a lot about mRNA vaccines, but Japan recently approved a new one for covid. And this one is pretty exciting. Just like the mRNA vaccines you know and love, it delivers the instructions for making the virus’s spike protein. But here’s what makes it novel: it also tells the body how to make more mRNA. Essentially, it provides instructions for making more instructions. It’s self-amplifying.

I’ll wait while your head explodes.

Self-amplifying RNA vaccines (saRNA) offer a couple of important advantages over conventional mRNA vaccines, at least in theory. Because saRNA vaccines come with a built-in photocopier, the dose can be much lower. One team of researchers tested both an mRNA vaccine and an saRNA vaccine in mice and found that they could achieve equivalent levels of protection against influenza with just 1/64th the dose. Second, it’s possible that saRNA vaccines will induce a more durable immune response because the RNA keeps copying itself and  sticks around longer. While mRNA might last a day or two, self-amplifying RNA can persist for a month.

Lest you think that this is just a tweaked version of conventional mRNA, It’s not. “saRNA is a totally different beast,” Anna Blakney, a bioengineer at the University of British Columbia, told Nature. (Blakney was one of our 35 Innovators Under 35 in 2023.)

What makes it a different beast? Conventional mRNA vaccines consist of messenger RNA that carries the genetic code for covid’s spike protein. Once that mRNA enters the body, it gets translated into proteins by the same cellular machinery that translates our own messenger RNA. 

Self-amplifying mRNA vaccines contain a gene that encodes the spike protein as well as viral genes that code for replicase, the enzyme that serves as a photocopier. So one self-amplifying mRNA molecule can produce many more. The idea of a vaccine that copies itself in the body might sound a little, well, unnerving. But there are a few things I should make clear. Although the genes that give these vaccines the ability to self-amplify come from viruses, they don’t encode the information needed to make the virus itself. So saRNA vaccines can’t produce new viruses. And just like mRNA, saRNA degrades quickly in the body. It lasts longer than mRNA, but it doesn’t amplify forever. 

Japan approved the new vaccine, called LUNAR-COV19, in late November on the basis of results from a 16,000-person trial in Vietnam. Last month researchers published results of a head-to-head comparison between LUNAR-COV19 and Comirnaty, the mRNA vaccine from Pfizer-BioNTech. In that 800-person study, vaccinated participants received either five  micrograms of LUNAR-COV19 or 30 micrograms of Comirnaty as a fourth dose booster. Reactions to both shots tended to be mild and resolve quickly. But the self-amplifying mRNA shot did elicit antibodies in a greater percentage of people than Comirnaty. And a month out, antibody levels against Omicron BA.4/5 were higher in people who received LUNAR-COV19. That could be a signal of increased durability.

The company has already filed for approval in Europe. It’s also working on a self-amplifying mRNA vaccine for flu, both seasonal and pandemic. Other companies are exploring the possibility that self-amplifying mRNA might be useful in rare genetic conditions to replace missing proteins. Arcturus, the company that co-developed LUNAR-COV19 with the global biotech CSL, is also developing self-amplifying messenger RNA to treat ornithine transcarbamylase deficiency, a rare and life-threatening genetic disease. It’s an mRNA bonanza that will hopefully lead to better vaccines and new therapies. 

Another thing

Babies and AI learn language in very different ways. The former rely on a relatively small set of experiences. The latter relies on data sets that encompass a trillion words. But this week I wrote about a new study that shows AI can learn language like a baby—at least some aspects of language. The researchers found that a neural network trained on things a single child saw and heard over the course of a year and a half could learn to match words to the objects they represent. Here’s the story. 

Read more from MIT Technology Review’s archive

mRNA vaccines helped tackle covid, but they can help with so much more—malaria, HIV, TB, Zika, even cancer. Jessica Hamzelou wrote about their potential in January, and I followed up with a story after two mRNA researchers won a Nobel Prize. 

Using self-amplifying RNA isn’t the only way to make mRNA vaccines more powerful. Researchers are tweaking them in other ways that might help boost the immune response, writes Anne Trafton. 

From around the web

Elon Musk says his company Neuralink has implanted a brain chip in a person for the first time. The device is designed to allow people to control external devices like smartphones and computers with their thoughts. (Washington Post)

In August I  wrote about Vertex’s quest to develop a non-opioid pain pill. This week the company announced positive results from phase 3 trials. The company expects to seek regulatory approval in the coming months, and if approved, the drug is likely to become a blockbuster. (Stat)

In some rare cases, it appears that Alzheimer’s can be transmitted from one person to another. That’s the conclusion of a new study: it found that eight people who received growth hormone from the brains of cadavers before the 1980s had sticky beta-amyloid plaques in their brains, a hallmark of the disease. The growth hormone they received also contained these proteins. And when researchers injected these proteins into mice, the mice also developed amyloid plaques. (Science)

A team of researchers at New York University wondered if AI could learn like a baby. What could an AI model do when given a far smaller data set—the sights and sounds experienced by a single child learning to talk?

A lot, it turns out.  The AI model managed to match words to the objects they represent.  “There’s enough data even in this blip of the child’s experience that it can do genuine word learning,” says Brenden Lake, a computational cognitive scientist at New York University and an author of the study. This work, published in Science today, not only provides insights into how babies learn but could also lead to better AI models.

For this experiment, the researchers relied on 61 hours of video from a helmet camera worn by a child who lives near Adelaide, Australia. That child, Sam, wore the camera off and on for one and a half years, from the time he was six months old until a little after his second birthday. The camera captured the things Sam looked at and paid attention to during about 1% of his waking hours. It recorded Sam’s two cats, his parents, his crib and toys, his house, his meals, and much more. “This data set was totally unique,” Lake says. “It’s the best window we’ve ever had into what a single child has access to.” 

To train the model, Lake and his colleagues used 600,000 video frames paired with the phrases that were spoken by Sam’s parents or other people in the room when the image was captured—37,500 “utterances” in all. Sometimes the words and objects matched. Sometimes they didn’t. For example, in one still, Sam looks at a shape sorter and a parent says, “You like the string.” In another, an adult hand covers some blocks and a parent says, “You want the blocks too.” 

The team gave the model two cues. When objects and words occur together, that’s a sign that they might be linked. But when an object and a word don’t occur together, that’s a sign they likely aren’t a match. “So we have this sort of pulling together and pushing apart that occurs within the model,” says Wai Keen Vong, a computational cognitive scientist at New York University and an author of the study. “Then the hope is that there are enough instances in the data where when the parent is saying the word ‘ball,’ the kid is seeing a ball,” he says.

Matching words to the objects they represent may seem like a simple task, but it’s not. To give you a sense of the scope of the problem, imagine the living room of a family with young children. It has all the normal living room furniture, but also kid clutter. The floor is littered with toys. Crayons are scattered across the coffee table. There’s a snack cup on the windowsill and laundry on a chair. If a toddler hears the word “ball,” it could refer to a ball. But it could also refer to any other toy, or the couch, or a pair of pants, or the shape of an object, or its color, or the time of day. “There’s an infinite number of possible meanings for any word,” Lake says.

The problem is so intractable that some developmental psychologists have argued that children must be born with an innate understanding of how language works to be able to learn it so quickly.  But the study suggests that some parts of language are learnable from a really small set of experiences even without that innate ability, says Jess Sullivan, a developmental psychologist at Skidmore University, who was part of the team that collected Sam’s helmet camera data but was not involved in the new study. “That, for me, really does shake up my worldview.” 

But Sullivan points out that being able to match words to the objects they represent, though a hard learning problem, is just part of what makes up language. There are also rules that govern how words get strung together. Your dog might know the words “ball” or “walk,” but that doesn’t mean he can understand English. And it could be that whatever innate capacity for language babies possess goes beyond vocabulary. It might influence how they move through the world, or what they pay attention to, or how they respond to language. “I don’t think the study would have worked if babies hadn’t created the data set that the neural net was learning from,” she says. 

BRENDEN LAKE

The next step for Lake and his colleagues is to try to figure out what they need to make the model’s learning more closely replicate early language learning in children. “There’s more work to be done to try to get a model with fully two-year-old-like abilities,” he says. That might mean providing more data. Lake’s child, who is now 18 months old, is part of the next cohort of kids who are providing that data. She  wears a helmet camera for a few hours a week. Or perhaps the model needs to pay attention to the parents’ gaze, or to have some sense of the solidity of objects—something children intuitively grasp. Creating models that can learn more like children will help the researchers better understand human learning and development. 

AI models that can pick up some of the ways in which humans learn language might be far more efficient at learning; they might act more like humans and less like “a lumbering statistical engine for pattern matching,” as the linguist Noam Chomsky and his colleagues once described large language models like ChatGPT. “AI systems are still brittle and lack common sense,” says Howard Shrobe, who manages the program at the US government’s Defense Advanced Research Projects Agency that helped fund Lake’s team. But AI that could learn like a child might be capable of understanding meaning, responding to new situations, and learning from new experiences. The goal is to bring AI one step closer to human intelligence.

Why recycling alone can’t power climate tech

The potential to use old, discarded products to make something new sounds a little bit like magic. This is why, in some cases at least, recycling is going to be a crucial tool for climate technology.

But there are massive challenges ahead in material demand for climate technologies, and unfortunately, recycling alone won’t be enough to address them. Our climate reporter Casey Crownhart has taken a look at why recycling isn’t always the answer—and what else might help instead. Read the full story. 

This story is from The Spark, our weekly newsletter covering climate technology. Sign up to receive it in your inbox every Wednesday.

This robot can tidy a room without any help

The news: A new robotic system called OK-Robot could train robots to pick up and move objects in settings they haven’t encountered before. A robot running the system, which uses publicly-available AI models, was able to pick up and move objects around a room with an 58.5% success rate—rising to 82% in rooms that were less cluttered.

Why it matters: While robots may easily complete tasks like these in a laboratory, getting them to work in an unfamiliar environment is a real challenge. OK-Robot could help to plug the gap between rapidly improving AI models and actual robot capabilities, as it doesn’t require any additional training. Read the full story.

—Rhiannon Williams

If you’re interesting in reading more about home robots, why not check out:

+ Here’s why everyone’s getting excited about household robots again.

+ This new system can teach a robot a simple household task within 20 minutes. The Dobb-E domestic robotics system was trained in real people’s homes and could help solve the field’s data problem. Read the full story.

+ Watch this robot cook shrimp and clean autonomously. Even cheap hardware can perform complex tasks, and AI is helping robots get smarter still. Read the full story.

Check out the new MIT Technology Review app

MIT Technology Review’s new app brings our trusted journalism and expert insights right to your fingertips, ensuring you’re always at the forefront of breakthrough innovations.

Our redesigned app provides a user-friendly interface with the ability to save favorite articles and search trending topics—and explore all our exclusive technology coverage along with our trending daily newsletters. Check it out!

The must-reads

I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.

1 Mark Zuckerberg apologized to families harmed by social media
Under some duress, during a Senate grilling. (WP $)
+ X boss Linda Yaccarino acknowledges that the platform needs more moderators. (Wired $)
+ TikTok’s Singaporean CEO was grilled at length, but not about child safety. (Motherboard)
+ Dramatic moments were in plentiful supply during the hearing. (NY Mag $)
+ Child online safety laws will actually hurt kids, critics say. (MIT Technology Review)

2 Autonomous car firm Cruise is hemorrhaging money
To the tune of almost $3.5 billion last year alone. (WP $)
+ What’s next for robotaxis in 2024. (MIT Technology Review)

3 Hackers spied on critics of Jordan using NSO Group’s spyware 
In a concerted campaign targeting journalists, activists, and lawyers. (Bloomberg $)

4 AI startups are putting themselves up for sale
They’re incredibly expensive to run, and increased competition has small businesses clamoring to be acquired. (The Information $)
+ One growing startup has ambitions of challenging Google’s search dominance. (NYT $)
+ Make no mistake—AI is owned by Big Tech. (MIT Technology Review)

5 Amazon could soon be liable for the safety of third-party products
A proposed US government order could reclassify it from a retailer to a distributor. (WSJ $)

6 Ovaries may offer clues on how to extend our lifespans
Listen up, longevity bros. (FT $)
+ Inside the billion-dollar meeting for the mega-rich who want to live forever. (MIT Technology Review)

7 Product placement is getting an AI-powered makeover
It’s a sneaky new way of getting products in front of eyeballs outside of traditional ads. (NYT $)
+ TikTok could be about to lose a major chunk of its music catalog. (Wired $)

8 Life is imitating art for Ukraine’s video game makers
Apocalyptic survival game Stalker 2: Heart of Chernobyl has a new resonance for its creators. (The Guardian)

9 The Sun’s magnetic poles are about to reverse
And communication satellites and our energy grid back on Earth is at risk of disruption. (Vox)
+ LED lights are contributing to a whole lot of light pollution. (The Atlantic $)

10 How to tell if we’re living in a simulation or not
Simulated and fake are not the same thing. (Ars Technica)

Quote of the day

“Your product is killing people.”

—US Senator Josh Hawley rounds on Facebook boss Mark Zuckerberg during a Senate hearing on child safety, the New York Times reports.

The big story

Government technology is famously bad. It doesn’t have to be.

October 2023

It’s a reality of politics that is often overlooked: once a law is passed, it needs to evolve from an idea into a plan with a budget and staff, and from there it needs to actually reach the lives of millions of people. Moving from policy to implementation has always been a hard part of governing. Today it’s easy to assume technology can make it easier. But is that actually true?

New York City is something of a test lab for strategies to confront some big problems that plague the modern state. And it is finding that sometimes the solutions are surprisingly low-tech. Read the full story.

—Tate Ryan-Mosley

We can still have nice things

A place for comfort, fun and distraction in these weird times. (Got any ideas? Drop me a line or tweet ’em at me.)

+ Crime Pays But Botany Doesn’t is a fantastic name for a YouTube channel.
+ It’s called the world’s greatest stretch for a reason.
+ Road tripping in Saudi Arabia looks suitably surreal.
+ Stop putting it off: here’s how to declutter your home, room by room.
+ What can’t The Bear actor Jeremy Allen White live without? Temporary tattoos, apparently.

The potential to use old, discarded products to make something new sounds a little bit like magic. I absolutely understand the draw, and in some cases, recycling is going to be a crucial tool for climate technology. I’ve written about recycling for basically any climate technology you can think of, including solar panels, wind turbines, and batteries. (I’ve also covered efforts to recycle plastic waste.)

For my most recent story, I was researching the materials used for the magnets that power EVs and wind turbines. (Read the result here!) And once again, I was struck by a stark reality: there are massive challenges ahead in material demand for climate technologies, and unfortunately, recycling alone won’t be enough to address them. Let’s take a look at why recycling isn’t always the answer, and what else might help. 

Mind the gap

We’re building a whole lot more climate technologies than we used to, which means there aren’t enough old, discarded technologies sitting around, waiting to be mined for materials. Obviously the growth in clean-energy technologies is a great thing for climate action. But it presents a problem for recycling. 

Take solar panels, for instance. They tend to last at least 25, maybe 30 years before they start to lose the ability to efficiently harness energy from the sun and transform it into electricity. So the panels available for recycling today are those that were installed over two decades ago (a relatively small fraction are ones that have been broken or need to be taken down early). 

In 2000, there was a little over one gigawatt of solar power installed globally. (Yes, 2000 was nearly 25 years ago—sorry!) So today’s recycling companies are competing with each other for that relatively small amount of material. If they can hang in there, there will eventually be plenty of solar panels to go around. Over 300 gigawatts of solar power were added in 2023.  

This gap is a common challenge in recycling for other technologies, too. In fact, one of the problems facing the growing number of battery recycling companies is a looming shortage of materials to recycle.

It’s important to start building infrastructure now, so we’re ready for the inevitable wave of solar panels and batteries that will eventually be ready for recycling. In the meantime, recyclers can get creative in where they’re sourcing materials. Battery recyclers today will rely on a lot of manufacturing scrap. Looking to other products can help as well—rare earth metals for EV motors and wind turbines could be partially sourced from old iPhones and laptops.

Closing the loop

Even if we weren’t seeing explosive growth for new technologies, there would be another problem: no recycling process is perfect. 

The issues start at the stage of collecting old materials (think of the iPods and flip phones in your junk drawer, gathering dust), but even once material makes it to a recycling center, some will wind up in the waste because it breaks down in the process or just can’t be economically recovered. 

Exactly how much material can be recovered depends on the material, the recycling process, and the economics at play. Some metals, like the silver in solar cells, might be able to reach 99% recovery or higher. Others can pose harder challenges, including the lithium in batteries—one recycler, Redwood Materials, told me last year its process can recover around 80% of the lithium from used batteries and manufacturing scrap. The rest will be lost.

I don’t mean to be a Debbie Downer. Even with imperfect recovery, recycling could help meet demand for materials in many energy technologies in the future. Recycling rare earth metals could cut mining for metals like neodymium in half, or more, by 2050.

But a robust supply of recycled materials for many climate technologies is still decades away. In the meantime, many companies are working to build options that use more widely available, cheaper alternatives. Check out my story on one startup, Niron Magnetics, which is working to build permanent magnets without rare earth metals, to see how new materials can help accelerate climate action and close the gap that recycling leaves. 

Related reading

See how old batteries could help power tomorrow’s EVs in my feature story on Redwood Materials.

For more on where battery recycling might be going, check out this accompanying interview with former Tesla exec and Redwood founder JB Straubel. 

Some companies are working out ways to recycle the valuable materials in solar panels.

Scientists are still trying to determine how we can best recycle wind turbine blades.

COSTFOTO/NURPHOTO VIA AP

Two more things

The world’s largest EV maker is getting into the shipping business. BYD is amassing a fleet of ships to export its vehicles from China to the rest of the world. Read more about why the automaker is getting creative and what comes next in this fascinating story from my colleague Zeyi Yang. 

Also, be sure to read the second part of James Temple’s blockbuster series on critical minerals. This one is a fascinating analysis that digs into how one Minnesota mine could unlock billions of dollars for EVs and batteries in the US. If you missed part one detailing what’s going on with the mine and the local community, that’s here, and you can check out my interview with James about his reporting in last week’s newsletter here.

Keeping up with climate  

The world’s largest cruise ship departed on its maiden voyage last week. The whole thing is a bit of a climate fiasco. Taking a cruise can be about twice as emissions intensive as flying and staying in a hotel. (Bloomberg)

A new refinery in Georgia will churn out millions of tons of jet fuel made from plants instead of petroleum. The new facility marks a milestone for alternative jet fuels. (Canary Media)

→ While alternatives are often called “sustainable aviation fuels” or SAFs, some varieties are anything but sustainable. Here’s what you need to know about all these newfangled jet fuels. (MIT Technology Review)

China nearly quadrupled its new energy storage capacity last year. It’s a massive jump for the growing industry, which is key to balancing the growing fraction of renewables on the grid. (Bloomberg)

Huge charging depots for electric trucks are coming to California. Big batteries in big vehicles require big chargers, and new funding from the US government could be crucial in building them. (Canary Media)

→ The three biggest truck makers are calling for better charging infrastructure for heavy-duty vehicles (New York Times)

EV charging can get a bit tricky for those of us who don’t live in single-family homes with a garage to charge in. Here are some solutions. (Washington Post)

The US is the world’s largest exporter of liquefied natural gas, but new exports are on pause. The Department of Energy says it’s trying to work out how to regulate them, and what the climate impact of cutting gas exports might be. (Grist)

Now, a new system called OK-Robot could train robots to pick up and move objects in settings they haven’t encountered before. It’s an approach that might be able to plug the gap between rapidly improving AI models and actual robot capabilities, as it doesn’t require any additional costly, complex training.

To develop the system, researchers from New York University and Meta tested Stretch, a commercially available robot made by Hello Robot that consists of a wheeled unit, a tall pole, and a retractable arm, in a total of 10 rooms in five homes. 

While in a room with the robot, a researcher would scan their surroundings using Record3D, an iPhone app that uses the phone’s lidar system to take a 3D video to share with the robot. 

The OK-Robot system then ran an open-source AI object detection model over the video’s frames. This, in combination with other open-source models, helped the robot identify objects in that room like a toy dragon, a tube of toothpaste, and a pack of playing cards, as well as locations around the room including a chair, a table, and a trash can.

The team then instructed the robot to pick up a specific item and move it to a new location. The robot’s pincer arm did this successfully in 58.5% of cases; the success rate rose to 82% in rooms that were less cluttered. (Their research has not yet been peer reviewed.)

The recent AI boom has led to enormous leaps in language and computer vision capabilities, allowing robotics researchers access to open-source AI models and tools that didn’t exist even three years ago, says Matthias Minderer, a senior computer vision research scientist at Google DeepMind, who was not involved in the project.

“I would say it’s quite unusual to be completely reliant on off-the-shelf models, and that it’s quite impressive to make them work,” he says.

“We’ve seen a revolution in machine learning that has made it possible to create models that work not just in laboratories, but in the open world,” he adds. “Seeing that this actually works in a real physical environment is very useful information.”

Because the researchers’ system used models that weren’t fine-tuned to this particular project, when the robot couldn’t find the object it was instructed to look for it simply stopped in its tracks instead of trying to work out a solution. That significant limitation is one reason the robot was more likely to succeed in tidier environments—fewer objects meant fewer chances for confusion, and a clearer space for navigation.

Using ready-made open-source models was both a blessing and a curse, says Lerrel Pinto, an assistant professor of computer science at New York University, who co-led the project. 

“On the positive side, you don’t have to give the robot any additional training data in the environment, it just works,” he says. “On the con side, it can only pick an object up and drop it somewhere else. You can’t ask it to open a drawer, because it only knows how to do those two things.” 

Combining OK-Robot with voice recognition models could allow researchers to deliver instructions simply by speaking to the robot, making it easier for them to experiment with readily available datasets, says Mahi Shafiullah, a PhD student at New York University who co-led the research.

“There is a very pervasive feeling in the [robotics] community that homes are difficult, robots are difficult, and combining homes and robots is just completely impossible,” he says. “I think once people start believing home robots are possible, a lot more work will start happening in this space.”

Why the world’s biggest EV maker is getting into shipping

Earlier this month, a massive ship picked up over 5,000 electric cars from two ports in northern and southern China. Five days later, it passed through Singapore, and it is now headed for India. However, its final destination is in Europe, where most of the cars will be sold. 

The ship’s name is BYD Explorer No.1. As the first of a massive fleet that BYD is building, it reflects the Chinese company’s ambition to establish a seafaring business that supports its new role in the global car trade.

There’s rising appetite for BYD’s cars overseas. In 2023, BYD exported over 240,000 vehicles, up from 55,000 in 2022. But it’s run into a snag: to get the most financial benefit from its exploding popularity abroad, it’s having to expand beyond the car trade into the shipping business. Read the full story.

—Zeyi Yang

To read more about BYD’s shipping ambitions, check out the latest edition of China Report, our weekly newsletter giving you the inside track on all things happening in China. Sign up to receive it in your inbox every Tuesday.

How new magnets could accelerate climate action

The motor in your vacuum cleaner and the one in your electric vehicle likely have at least one thing in common: they both rely on powerful permanent magnets to function. But the materials for those magnets could soon be in short supply. 

The permanent magnets used in high-end motors today are built using a class of materials called rare earth metals. Demand for these materials is expected to skyrocket in the coming decades, fueled in particular by the growth of electric vehicles and wind turbines. But there’s good news: work is underway to address this looming shortage. Read the full story.

—Casey Crownhart

The must-reads

I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.

1 Big Tech is preparing to face a grilling in the US Senate over child safety
X boss Linda Yaccarino will testify alongside fellow leaders today for the first time. (Insider $)
+ The Senate and the House can’t agree on what should be done. (FT $)
+ Here’s a handy who’s who of who’s testifying today. (WP $)
+ Why child safety bills are popping up all over the US. (MIT Technology Review)

2 Wuhan is becoming a major autonomous vehicle test bed
So much so that China poses a mounting challenge to the west’s driverless dominance. (FT $)
+ General Motors is slashing its spending on Cruise robotaxis. (Reuters)
+ A race for autopilot dominance is giving China the edge in autonomous driving. (MIT Technology Review)

3 Fake AI-generated images of celebrities are spreading across Google
All fingers point to an aspiring movie star. (Motherboard)
+ These six questions will dictate the future of generative AI. (MIT Technology Review)

4 How a missing chip shipment sparked a US probe
Authorities stepped in to investigate suspicions that it was ultimately bound for China. (WSJ $)
+ New kinds of chips are on the way. (IEEE Spectrum)
+ The next big thing? Origami computers, obviously. (Quanta Magazine)
+ China’s big volley in the semiconductor exports war. (MIT Technology Review)

5 Apple’s Vision Pro headset gathers reams of data
And it could end up revealing much more than anticipated. (WP $)
+ Apple is relying on its cool cachet to get people to wear them in the first place. (Bloomberg $)
+ Meanwhile, Meta’s AR glasses division looks like it’s struggling. (The Information $)

6 A single pig butchering scam made fraudsters $40 million
But the true amount is likely to be even higher. (404 Media)
+ The involuntary criminals behind pig-butchering scams. (MIT Technology Review)

7 Is AI drug discovery really better than humans?
Just because it’s faster doesn’t necessarily mean it’ll be more effective. (Bloomberg $)
+ AI is dreaming up drugs that no one has ever seen. Now we’ve got to see if they work.  (MIT Technology Review)

8 A new theory of how everything came to be is emerging
It’s a riff on quantum theory using the laws of thermodynamics. (New Scientist $)
+ Why is the universe so complex and beautiful? (MIT Technology Review)

9 The JWST has captured some beautiful spiral galaxies
Prepare to be dazzled. (The Atlantic $)
+ Contrary to what you may have heard, aliens have not visited Earth. (NY Mag $)

10 TikTok’s bite-sized soap operas are gripping fans
The minute-long clips are formulaic, but extremely compulsive viewing. (NYT $)

Quote of the day

“We’ve been better, Elmo.”

—An X user responds to Elmo the lovable muppet’s mistake of asking how everyone was doing on the social network, the Washington Post reports.

The big story

Inside the cozy but creepy world of VR sleep rooms

March 2023 

People are gathering in virtual spaces to relax, and even sleep, with their headsets on. VR sleep rooms are becoming popular among people who suffer from insomnia or loneliness, offering cozy enclaves where strangers can safely find relaxation and company—most of the time.

These rooms are created to induce calm. Some imitate beaches and campsites with bonfires, while others mimic hotel rooms or cabins. 

The opportunity to sleep in groups can be particularly appealing to isolated or lonely people who want to feel less alone, and safe enough to fall asleep. The trouble is, what if the experience doesn’t make you feel that way? Read the full story.

—Tanya Basu

We can still have nice things

A place for comfort, fun and distraction in these weird times. (Got any ideas? Drop me a line or tweet ’em at me.)

+ Omg it’s an actual baby shark! (doo doo doo doo).
+ A potted history of beds, and why they’re so wonderful
+ Drop everything: this is how your precious cat experiences the world.
+ Some essential tips on what to do should you ever come across an escaped monkey (above all, don’t panic.)
+ Yes chef! Here are the brightest rising culinary stars you’ll want to keep an eye on.

For people who have been watching BYD for a long time, it won’t be surprising that the company has just ventured into a new field. 

The Chinese electric-vehicle maker has been particularly good at expanding into different, related businesses. Not only can it make high-performing and safe batteries for cars, but it also does almost everything in house, from designing car chips to mining lithium and other materials. The fact that it has subsidiaries in every step of the EV supply chain enables the company to keep its costs down and sell cars at more competitive prices.

Now, to pull that off once again, BYD is starting a sea freight business. As I just wrote in a story published today, the company is assembling a fleet of at least eight car-carrier ships that will transport BYD cars from factories in China to sell in Europe, South America, and other markets.

BYD has had a meteoric rise to become the Chinese EV sector’s poster child in recent years, and 2023 was particularly good for the company. It sold 3 million electric cars and plug-in hybrid models last year, up from 1.8 million in 2022. BYD managed to beat Tesla to become the world’s top-selling EV company in the fourth quarter of 2023. 

While the majority of those cars were sold in China, BYD’s export business has been expanding significantly. It exported over 240,000 cars in 2023, more than a fourfold increase from 55,000 cars in 2022; and the latter number was itself more than a fourfold increase from 13,000 in 2021.

But one thing has been getting in the way of these bonkers numbers: the lack of car-carrier ships internationally. A bust cycle in the international shipping industry since 2008, the technological challenge of making ships greener, and the fact that existing vessels are often already reserved by automakers in other countries—these factors have collectively resulted in ever-rising costs to hire a ship that can transport Chinese EVs abroad.

So Chinese companies like BYD and SAIC Motor are following in the footsteps of Japanese and Korean automakers: they’re building, chartering, and managing their own fleets of ships. This January, one boat operated by BYD and another operated by SAIC Motor set sail for the first time, between them carrying over 10,000 vehicles toward Europe. 

These two massive ships are a symbol of just how competitive and successful China’s EV industry has become. And that’s likely to continue for some time, as other countries and traditional car brands are belatedly playing catch-up.

This is not to say China’s EV industry has nothing to fear. As I’ve laid out in previous articles, there are still factors that could slow down or even derail the export of Chinese EVs. Geopolitics is a major one. For example, in Europe, where many of the new car-carrier ships are heading, there’s already an anti-subsidy investigation against Chinese cars going on, which could end up making it much more costly to sell there.

Chinese companies going into sea shipping should note at least one cautionary tale from recent history. Before BYD, there was another Chinese car company called Chery, which started exporting its cars in the 2000s. In 2007, it acquired a shipbuilding company for the exact same reason: it wanted to increase the capacity to ship cars abroad. But the financial crisis doomed Chery’s burgeoning export business, and it didn’t build its first ship until a decade later. 

Chery is still around today. It has made the pivot from gas to electric cars and is competing with BYD both domestically and in the export market. But its ill-fated shipbuilding attempt could be a lesson for other Chinese companies that are now making similar moves: even though the future looks bright, building and maintaining these massive ships is a risky, expensive business if their car sales don’t keep up.

Do you think it’s the right decision for companies like BYD and SAIC Motor to build their own car-carrier fleet? Tell me your thoughts at zeyi@technologyreview.com.

Catch up with China

1. The White House plans to cut off Chinese entities’ access to American cloud services to train AI models. (Reuters $)

2. Some legislators in the US want to reactivate the Justice Department’s China Initiative. (NBC News)

• The controversial program was built to protect national security. But it strayed from its focus and ended in 2022. (MIT Technology Review)

3. Another proposed bill in Congress seeks to ban Chinese biotech firms from federal contracts. (South China Morning Post $)

4. After an almost five-year import freeze on Boeing’s 737 MAX, Chinese airline companies have restarted purchasing the controversial jet model. (Reuters $)

5. The Chinese movie market used to be a cash machine for Hollywood blockbusters. Not anymore. (New York Times $)

6. The Taiwanese government is funding efforts to build its own Chinese AI model that’s free of China’s political influence. (Bloomberg $)

• Meanwhile, US spies want an AI model of their own to use against China without leaking national secrets. (Bloomberg $)

7. Elon Musk has praised Chinese electric vehicles, again. He says Chinese EV makers will “pretty much demolish” most competitors if there are no trade barriers. (CNBC)

Lost in translation

Another type of device is getting an AI transformation in China: student tablets. Commonly called “learning machines” (though they have no connection to machine learning), these are tablets specifically designed to tutor children in school subjects, supporting functions like electronic dictionaries and virtual classes. According to Chinese outlet IQ Tax Research Center, many of these sorts of products have embraced AI in the past year, including devices made by China’s leading AI companies like Baidu and iFlytek. 

However, some parents have found these “AI-powered devices” prone to errors and inaccuracies. For example, one user mentioned that a math problem was solved with different answers each time the AI explained it. Others felt the educational content recommended by the AI was not always suitable for their children’s needs. At the end of the day, these “learning machines” are often still inadequate, despite how they are marketed.

One more thing

Do you stick to reserving dinner at restaurants with 4.5+ stars on Google? In China, some young people have had too many disappointing experiences chasing after viral restaurants with inflated reviews. Instead, they are starting a trend of choosing restaurants with review scores around 3.5. Their justification? “If a restaurant can survive for decades with such a low review, there must be something really special about it,” one comment on social media reads. It’s also about rebelling against the ubiquitous digital platforms that dictate where everybody goes, reports China’s Lifeweek magazine.

Permanent magnets can maintain a magnetic field on their own without an electric charge. They’re commonly used in motors, making them spin when an electric field is applied. The permanent magnets used in high-end motors today are built using a class of materials called rare earth metals. Demand for these materials is expected to skyrocket in the coming decades, fueled in particular by the growth of electric vehicles and wind turbines. As mines and processing facilities struggle to keep up, supplies may stretch thin.

One Minnesota startup has been working to address this looming shortage. Niron Magnetics is building a large-scale manufacturing facility to produce iron nitride, a magnetic material derived from common elements, while also working to improve the material’s properties so that it can be used in stronger magnets to power more products. The results may help address yet another coming supply crunch that threatens to slow down action on climate change.

A growing gap

The permanent magnets you’re probably most familiar with are the cheap ones made from materials called ferrites that are holding up postcards and wedding announcements on your refrigerator.

But many of the devices sprinkled through our daily lives, like our vacuums and EVs, require much higher-powered magnets. Motors that generate motion using permanent magnets tend to be more powerful and efficient, so rare earth metals, such as neodymium and dysprosium, have become vital for a wide range of devices. In a wind turbine, for instance, magnets in the generator harness motion from the blades and turn it into electricity.  

Like many of the other materials needed for clean energy technologies, we can expect a meteoric rise in demand for rare earth metals used in magnets as the world rushes to address climate change.

In the case of neodymium and dysprosium, supply will need to increase sevenfold by 2050 just to meet demand for wind turbines, says Seaver Wang, co-director of the climate and energy team at the Breakthrough Institute, an environment and policy think tank.

In addition, rare earth metal demand for electric vehicles could increase 15-fold from today’s levels by 2040, according to an analysis from the International Energy Agency. And it’s not just clean energy technologies—increased access to electricity and cheap electronics means demand for rare earth metals will rise across other sectors, too. 

The world is unlikely to exhaust the geological reserves of rare earth metals anytime soon, Breakthrough’s Wang says—rare earth metals aren’t actually all that rare, at least when it comes to the entire planet’s supply. But they don’t tend to be very concentrated even in the places they are found, so scaling the supply of rare earth metals quickly and economically enough will be a major challenge.

In the near term, global demand for magnets made with neodymium could triple by 2035, while production will likely only double by then, given the long lead times required to build new mines, according to materials research firm Adamas Intelligence.

Given the growing demand, “the world needs a different solution and technology,” says Jonathan Rowntree, CEO of Niron Magnetics.

Few alternatives to permanent magnets exist today. Recycling can help reduce the need for future rare earth mining and processing, but there won’t be enough used material to meet the growing demand for decades.

Tesla announced in 2023 that it would move away from rare earth metals in its motors in the future, though the company hasn’t shared details about how it will do so. Some experts have speculated that it plans to use lower-powered ferrite materials, which would add bulk and weight to the motor. 

Rowntree and his colleagues see iron nitride as part of the solution to the anticipated problem of constraints in the supply of rare earth metals. Iron nitride magnets don’t use those metals, and they don’t require cobalt, another metal sometimes used in magnets (and in lithium-ion batteries) that’s under growing scrutiny because of the environmental and humanitarian issues often associated with its mining. And some experts say these iron-based materials might end up creating magnets just as strong as those that include rare earth metals. 

An attractive alternative

Though iron nitride (specifically, a phase called alpha double prime) was discovered in the 1950s, it wasn’t until the 1970s that researchers discovered its strong magnetic properties, says Jian-Ping Wang, a professor at the University of Minnesota and the technical founder and chief scientist at Niron Magnetics.

Even then, scientists couldn’t explain the physics underlying the material’s magnetic properties, and they struggled to recreate magnetic samples reliably through the 1990s. Intrigued by this problem, Wang began work on iron nitride materials at the university in 2002.

After making hundreds of samples and working for nearly a decade, Wang cracked the code to reliably make iron nitride materials in thin films. He presented his findings at a major conference in 2010, the same year geopolitical tensions between Japan and China sparked a huge increase in the price of rare earth metals.

Suddenly, there was a greater appetite for alternatives to rare earths that could be used to make strong permanent magnets. The US Department of Energy’s ARPA-E office sponsored grants to develop such materials, awarding one to Wang and the research that would eventually become Niron Magnetics.

Rare earth metals became ubiquitous across technologies because they represented “a huge jump” in the energy density of magnets when they were discovered in the 1960s, says Matthew Kramer, a senior scientist at Ames National Laboratory.

One of the primary gauges of a magnet’s properties is its energy density, measured in mega-gauss-oersteds (MGOe). While the ferrite magnets on your fridge likely have an MGOe of around 5, neodymium-based magnets are much stronger, reaching around 50 MGOe.

Rare earth metals like neodymium are currently a crucial ingredient in permanent magnets because they can wrangle other metals into an arrangement that helps generate a strong magnetic field.

Permanent magnets produce magnetic fields because of spinning electrons, small charged particles in atoms. Different elements have different numbers of free electrons that in some circumstances can be made to spin in the same direction, generating a magnetic field. The more electrons that are free and spinning in the same direction, the stronger the magnetic field.

Iron has a lot of free electrons, but without an overarching structure they tend to spin in different directions, canceling each other out. Adding in neodymium, dysprosium, and other rare earth metals can help arrange iron atoms in a way that allows their electrons to work together, resulting in powerful magnets.

Iron nitride does what few other materials can: it arranges iron into a structure that gets electrons spinning together in this way and keeps them aligned—no rare earth metals required.

“If you could get the nitrogen to spread these irons out in the appropriate way, you should be able to potentially get a really, really good permanent magnet,” Kramer says. That has proven to be a challenge though, he adds, because it’s difficult to make these materials in bulk and to harness the complex chemistry in a way that forces them to retain their magnetization. 

Idea to execution

After Wang was able to reliably create thin films of iron nitride, the next step was to figure out how to make it in bulk, grind it up, and squish it together to make magnets.

Finding a manufacturing process was a challenge in part because iron nitride degrades at high temperatures, which limits the options available in traditional magnet manufacturing, Wang explains. He developed several methods to make iron nitride in bulk, one of the most promising of which involves diffusing nitrogen through iron oxide (rust is a type of iron oxide) under very specific conditions.

In recent years, Niron has focused on perfecting and scaling up the manufacturing process, Rowntree says. A significant remaining challenge is determining how to help iron nitride reach its full potential.

NIRON MAGNETICS

In theory, iron nitride should be able to produce magnets that are even stronger than neodymium ones. But today, Niron’s magnets can only reach around 10 MGOe, Rowntree says. That’s sufficient for devices like speakers, which the company is exploring as an early product. It displayed small speakers made with Niron magnets at CES in January.

With higher magnet strength, iron nitride magnets will be more useful in devices like electric vehicles and wind turbines. In theory, the material should be able to reach 20 to 30 MGOe using Niron’s current manufacturing method, Wang says, though achieving that will require “a lot of optimization.” The theoretical ceiling is much higher, with iron nitride potentially being able to form magnets stronger than the neodymium ones used today.

Niron recently received over $30 million from investors, including GM Ventures and Stellantis Ventures, for a total of more than $100 million in funding. The company is working to scale up production capacity in its current pilot plant, with the aim of reaching 1,000 kilograms of production capacity by the end of 2024. 

Niron’s work, along with other alternatives and workarounds, could be crucial in loosening a major potential bottleneck for several critical climate technologies. 

“Increased magnets and increased magnet supply are critical to enabling the energy transition,” says Gregg Cremer, an advisor at ARPA-E. “Without more magnets, we’re just not going to be able to meet our objectives.”

The ship’s name is BYD Explorer No.1. As the first of a massive fleet that BYD is building, it reflects the Chinese company’s ambition to establish a seafaring business that supports its new role in the global car trade.

BYD, founded by a Chinese metallurgy researcher named Wang Chuanfu in 1995, started out making small batteries for mobile devices. It later expanded its business to automobiles and eventually combined the two to make electric vehicles. In two decades, it became China’s largest EV maker. By the fourth quarter of 2023, it was the world’s.

BYD offers a lot of options, ranging from affordable sedans to luxury SUVs, and there’s rising appetite for its cars overseas. In 2023, BYD exported over 240,000 vehicles, up from 55,000 in 2022. But it’s run into a snag: to get the most financial benefit from its exploding popularity abroad, it’s having to expand beyond the car trade into the shipping business. 

The shortage of car-carrier vessels

To understand why BYD has made this move, you need to learn a little about how cars are transported across the sea. Usually, the cargo industry uses roll-on/roll-off (RORO) ships. Unlike ships that use a crane to lift up cargo and place it aboard, RORO ships have ramps that allow vehicles to be driven directly on, making the whole process much easier.

But these ships have been in short supply for the past few years. While old vessels have been entering retirement, new ship orders were down because of the 2008 financial crisis and the industry-wide upgrade to more environmentally friendly fuels, leaving a shortfall. 

Also, most car companies have longstanding relationships with shipping firms or own their own fleets of boats. For example, Japanese car makers like Nissan and Toyota each have fleets of RORO ships that can carry tens of thousands of cars. But China’s domestic car-carrier vessels represent only 2.8% of global shipping capacity, leaving the Chinese brands few options for getting their cars across the seas.

As a result, their access to RORO ships has become prohibitively expensive. According to Clarksons Research, the intelligence arm of the world’s largest shipping services provider, the price to rent (or charter, as the industry says) a car-carrier ship for a day skyrocketed to $115,000 in 2023, a historical high and close to seven times the average pre-pandemic price, which was around $17,000 in 2019.

New demand to ship cars predominantly comes from China right now. The country is on the verge of becoming the world’s largest car exporter (in fact, it may already have gained that status in 2023, but we won’t know until the official numbers are finalized). It exports a mix of traditional gas cars, electric cars made by Chinese companies, and Tesla cars manufactured in the Giga Shanghai plant. 

The shortage of shipping capacity is what’s standing in its way.

Venturing out by themselves

That’s why Chinese auto companies, which have become such prominent exporters thanks to the rise of EVs, are starting to form their own shipping businesses. 

News that BYD was looking to buy or charter ships was first reported by the shipping outlet Lloyd’s List in late 2022. In December that year, the company changed its corporate registration to include the business of international cargo shipping and ship management. MIT Technology Review contacted BYD for comment, but it did not respond in time for publication. 

BYD Explorer No.1 was delivered at the beginning of this year. The RORO vessel, which can carry 7,000 cars at the same time, is officially registered under Zodiac Maritime, a UK company controlled by the Israeli shipping tycoon Eyal Ofer, but BYD has leased it for an undisclosed period of time. In a press release, BYD says it plans to add seven more vessels to the fleet in the next two years. It also plans to let other companies export their vehicles using BYD’s ships, it says.

For its maiden voyage, the ship is carrying over 5,000 BYD vehicles and heading toward the ports of Vlissingen in the Netherlands and Bremerhaven in Germany, according to Chinese state media outlet Xinhua.

BYD is not the only Chinese automaker making this move. SAIC Motor, a Chinese state-owned company, sold 1.2 million vehicles overseas in 2023, 24% of which were EVs. It formed a RORO shipping subsidiary in 2021, and its newest RORO vessel, the largest of its kind and able to carry 7,600 cars, also set sail for the first time in January. Like BYD Explorer No.1, it’s going to Europe.

While BYD has announced that it will add energy-storing battery tech to its vessels, the RORO ships it’s chartering today are not electric yet. Most of the newer ships can be powered by either traditional fuel or liquefied natural gas, which is a cleaner energy source.

From carrying wood pulp to cars

It will be some time before these Chinese companies finish assembling their sea freight empires, as these massive new ships take years to build. In the meantime, some have turned to creative fixes for the supply shortage: repurposing ships that were designed for other types of cargo.

Particularly, they have their eyes on gigantic ships typically used to import thousands of tons of wood pulp from South America to China, where it’s made into everyday products like tissue, paper, and books. These wood-pulp carriers often end up empty or barely loaded on the way back, because China doesn’t have similar products to export. 

However, in recent years Chinese car companies have started to sell their vehicles to the South American continent, and shipping companies saw an opportunity there. China Ocean Shipping Company (COSCO), one of the largest shipping companies in the world, designed a foldable rack that can load cars and stack them up into a wood-pulp carrier. In July last year, COSCO loaded such a ship with over 2,700 cars and sent them to Brazil.

With makeshift arrangements like this and new RORO vessels being built, shipping bottlenecks for Chinese automakers could be significantly reduced in the next couple of years. Having their own fleets or chartering ships from domestic shipping companies could also further drive down costs, making Chinese cars even more competitive abroad. 

And just as the auto industry in Japan and South Korea has pushed these two countries to become global leaders in shipping, EVs could make China a major player in the ocean too.