As AI models continue to grow in size and complexity, chipmakers are shifting their focus from individual processor performance to the ability to efficiently connect thousands of accelerators into a single computing system. 

These highly interconnected server architectures have become a key competitive area for AI infrastructure companies, which are racing to deliver the computing scale needed for models with trillions of parameters and the growing adoption of AI agents.

According to Ding Yunfan, vice president of AI framework architecture at Biren, improvements in the performance of a single graphics processing unit (GPU) are no longer enough to meet these demands. Instead, the industry’s biggest challenge is turning vast GPU clusters into unified, seamless computing platforms capable of operating as a single system.

Architecture designed to break GPU cluster limits

Biren believes the future of large-scale AI computing depends on replacing conventional electrical interconnects with optical technology. According to Ding, traditional copper-based connections are approaching their physical limits, effectively capping today’s server architectures at around 128 GPUs, the South China Morning Post reports.

To overcome that bottleneck, the company has developed a distributed, decoupled supernode architecture built around near-packaged optics (NPO), which positions optical fibres closer to the chips to increase bandwidth and improve data transmission. Biren said the approach could enable clusters with up to 1,024 AI accelerator cards, significantly expanding the scale of computing systems needed to support increasingly demanding AI workloads.

Ding said optical interconnects are becoming essential for overcoming the scalability limits of conventional server architectures. Shanghai-based Biren is among a growing number of Chinese companies tackling these infrastructure challenges. Rivals including MetaX, Enflame and Alibaba are also developing supernode architectures to compete with Nvidia’s NVL72 and next-generation NVL144 AI platforms.

Biren’s Shanghai-based rival MetaX has introduced its Xijing S600 AI supernode, featuring a single rack that connects 64 GPUs. The company said its design reduces signal loss by removing external cabling between compute nodes, switch nodes and interconnect systems, improving efficiency for large-scale AI workloads.

Looking beyond electrical links to build larger AI supernodes

Biren is also working on new approaches to expand the scale of its AI computing systems. The company’s orthogonal hardware architecture, co-developed with telecommunications giant ZTE, uses traditional electrical connections, while Biren is simultaneously testing a prototype near-packaged optics (NPO) interconnect system. 

The chipmaker said the optical approach could eventually support supernodes with more than 512 accelerator cards, helping address the growing demand for larger and more efficient AI infrastructure.

Despite growing interest in optical networking for AI infrastructure, Ding said the semiconductor industry is still in the early stages of developing and testing optical interconnect technologies. He noted that widespread adoption will require further advances before these systems can be deployed at scale.

Furthermore, Ding emphasized that the technology still requires real-world validation before it can be widely adopted. He projected that NPO optical systems could reach mass commercial deployment around 2028, as the industry continues refining the technology and addressing remaining engineering challenges.

💾

From the factories of Detroit to the postal warehouses in Shenzhen, a new kind of worker is clocking in — and the humans it’s replacing aren’t going down without a fight.

In late June, thousands of rank-and-file workers in the Korean Metal Workers’ Union voted to authorize a strike in response to Hyundai’s announcement it would deploy over 25,000 humanoid robots across its factories.

Weeks later, that strike is slowing production at a Hyundai factory down to a trickle. According to the Wall Street Journal, the workers began a partial strike after negotiations broke down last week. Starting on Monday, workers refused to work for four hours of each shift, leaving production lines stalled for half the work day.

Per the WSJ, the economic impacts of the strike are stark: the pullback could disrupt production of about 5,000 cars, and cut into Hyundai’s sales revenue by around $135 million. Should Hyundai choose to dig in their heels after the labor action, the union said it could escalate the strike, either by prolonging the length of the strike, upgrading it to a full strike, or both.

The mobilization appears to be the first of its kind over humanoid robotics, specifically, marking a pivotal moment in labor history. At issue is Hyundai’s threat to deploy thousands of “Atlas” model robots built by Boston Dynamics, bipedal humanoids that measure about 6’2″ in height and can lift a maximum of 110 pounds.

While the robots have yet to be deployed, union organizers explained that the strike is meant to give workers the power to renegotiate their contracts on their own terms, before the automation drive begins in earnest.

“We have to prepare to ensure there are safeguards in place,” secretary general of the KMWU and lead negotiator Byun Jun-hwan told the WSJ.

Specifically, the union’s demands include job security guarantees designed around AI and automation, larger bonuses that reflect Hyundai’s profit margin, a transition away from hourly wages to a fixed salary agreement, and a formal agreement allowing workers final say on any humanoid robotics deployment.

More on automation: AI Is Pushing Older Employees Straight Out of the Workforce, New Report Finds

The post Striking Workers Bring Car Factory to a Screeching Halt Over Humanoid Robots appeared first on Futurism.

Creating a realistic 3D model from a simple image remains surprisingly difficult for artificial intelligence. 

While AI can generate impressive pictures, producing the precise CAD code engineers use to design aircraft parts, cars, and consumer products is far more challenging. Even a small mistake can make a design unusable. 

Now, researchers from MIT, IBM, and Red Hat have developed a framework that significantly improves the ability of AI systems to turn 2D images into functional CAD programs, producing more accurate designs while reducing inference computation by about 80 percent.

This new research work tackles one of the biggest obstacles in AI-driven engineering, which is the shortage of high-quality CAD training data. Existing image-to-CAD systems need vast numbers of examples linking 2D images to working CAD programs, but creating those datasets is expensive and labor-intensive. 

Instead of relying on more human-created data, the new approach teaches AI using the mistakes it already makes. “We want to obtain data augmentation that is informed by the model itself,” lead study author and AI researcher Giorgio Giannone said.

Turning failed designs into training material

The team’s framework, called Geometric Inference Feedback Tuning (GIFT), works with vision-language models that convert a 2D image and text description into Python code that can be executed in CAD software to create a 3D object.

To understand where a model struggles, GIFT asks it to solve the same CAD problem multiple times in parallel. Some outputs are correct, some are close to the target design, and others are much further away. 

According to the researchers, these ‘near-misses’ are the most valuable because they reveal exactly what the model does not yet understand.

The system executes each generated CAD program and compares the resulting 3D geometry with the target using a standard engineering metric known as Intersection over Union (IoU), which measures how closely two 3D shapes match. 

Rather than discarding imperfect outputs, GIFT identifies those that are close to correct and uses them to generate new training examples.

Giannone explained, “If we sample the model 10 times and it generates 10 correct answers to the same problem, then there is not much for it to learn. We care about the in-between cases, where the model might only solve the problem 50 percent of the time.”

Teaching the model how to fix its own errors

GIFT improves the model through two complementary techniques. The first, called GIFT-REJECT, keeps alternative CAD programs that generate the correct geometry even if the code differs from the original reference solution. 

Traditional AI training often assumes there is only one correct CAD program for each object, even though engineers can create the same part in many different ways. By exposing the model to multiple valid solutions, GIFT broadens its understanding of how CAD designs can be constructed.

The second method, GIFT-FAIL, focuses on imperfect outputs. The system takes a nearly correct 3D model generated by the AI, converts it back into an image, and pairs that image with the correct CAD program. By doing so, it creates new examples that effectively teach the model how to recover from specific mistakes.

More importantly, this process requires no human intervention. The framework automatically identifies errors, creates new training data, and uses that information to improve the model.

The researchers also designed GIFT around a technique known as inference-time scaling. Instead of retraining an entire AI model, which can be extremely expensive, the system improves performance during generation. 

Rather than discarding the extra computation spent searching for better CAD programs, GIFT converts successful and near-successful attempts into new training examples that can later be used to improve the model through fine-tuning.

Moreover, users can decide how much computation they want to spend depending on their budget and time constraints.

Better CAD designs with a fraction of the computing power

When evaluated on the GenCAD-Code dataset, models enhanced with GIFT achieved a 12 percent improvement in geometric accuracy compared with a strong supervised-learning baseline. 

The framework also reduced the inference computation needed to achieve comparable performance by about 80%.

“With GIFT, we started with geometry because with engineering problems, if the geometry of a 3D shape is not correct, nothing else will be correct, but there are many other aspects to consider,” Giannone said. 

The system’s ability to learn from its own mistakes could help accelerate product development by making AI-generated CAD models more reliable. 

Engineers could move more quickly from sketches to testable 3D designs, reducing development costs, shortening design cycles, and exploring a wider range of ideas without creating large amounts of new training data.

The researchers say this is only the beginning. So far, GIFT focuses on generating correct geometry, but future versions will aim to teach AI how to design parts that are not only accurate in shape but also easier to manufacture and better suited for real-world performance. 

The study is published on arXiv.

Georgia Tech’s Learn to Teach framework trains AI models concurrently, helping a humanoid robot navigate sand, gravel, slopes, and slippery surfaces.

Agility is opening a new training center for its Digit robots in Fremont, California.

You might wince at the professional spectacle of two humans beating each other to a pulp. But when it’s two robots? We want maximum carnage.

Thankfully, the inaugural Ultimate Robot Knock-out Legend (URKL) competition in Shenzhen, China, has delivered on providing the most brutal-but-bloodless spectacle you can lay eyes on.

In a video of a cage match between two mechanical humanoids, one of the clanker combatants literally knocks its opponent’s block off by delivering a flying kick to the face.

The robot continues to fight with its head dangling from between its shoulders and throws a few punches, but not for long.

After being knocked to the ground again, it flails uncontrollably while trying to get to its feet. The violent jerking severs the cable connecting the hanging head, and the robot collapses onto the ground. The other robot, victorious, dances over its unmoving body.

Insane new humanoid battle tournament in Shenzhen
byu/Fit-Case1093 insingularity

According to the state-run tabloid Global Times, the URKL competition featured 32 teams from around the world who all used the T800 humanoid robot from EngineAI, a Chinese robotics company.

The point of the competition, according to EngineAI CEO Zhao Tongyang, was twofold: create a global commercial robot fighting brand and promote research and industrialization.

“Let the event feed back into technology, and let technology drive the industry,” Tongyang told Global Times.

You can’t accuse Tongyang of not putting his money where his mouth is. In a video that went viral last December, he donned a suit of armor and let himself take blows from one of his T800 robots. One kick was so powerful that it slammed him to the ground.

It’s hard to tell how much of a splash the event made in the country, but it must’ve been a wild spectacle to witness in person. Legendary martial arts movie star and filmmaker Donnie Yen was in attendance and sounded blown away by what he saw.

“Honestly, I used to see robot fighting only in science fiction movies,” Yen said at the event’s opening, per Global Times. “But today, for the first time, I got to watch real robots fighting up close.”

“The weight of the machines, the precision of their movements — it was incredible,” he added.

Martial arts is one of Chinese robot companies’ favorite ways to show off their capabilities. China stunned the world earlier this year with a televised martial arts display performed by a group of Unitree G1 robots that shared the stage with human actors.

More on robots: Startup Upset After Journalist Points Out How Creepy Its Humanoid Robots’ Hands Are

The post Brutal Humanoid Robot Cage Match Stops Abruptly When One of the Robot’s Heads Pops Off appeared first on Futurism.

Hyundai workers in South Korea strike over pay and job protections as the automaker plans to deploy Atlas humanoid robots across Hyundai and Kia factories.

Nvidia is partnering with Japan’s robotics leaders to develop smarter factories using physical AI, new models, and domestic computing infrastructure.

Researchers at the University of Toronto Engineering have developed six new metal alloys using an artificial intelligence-driven discovery platform that could improve the durability of parts used in jet engines, nuclear power plants, and other extreme environments. The findings show that the AI-assisted system identified promising alloys within just a few weeks, significantly speeding up the search for high-performance materials.

These new alloys also work well with 3D metal printing, enabling the production be complex parts that are hard or impossible to produce with traditional methods.

AI speeds up alloy discovery

Led by Yu Zou, Canada Research Chair in Materials and Manufacturing for Extreme Environments, the team worked with Jason Hattrick-Simpers to build a system that uses computer modeling, machine learning, and robots to help with manufacturing.

Their method, called active learning, acts like a self-driving laboratory. Rather than testing thousands of metal combinations by hand, the system picks the best options, makes them, tests how they perform, and uses those results to guide the next experiments.

“There’s enormous demand for materials that can stand up to huge swings of temperature and pressure, such as what you would find inside a jet engine or in the steam generators inside nuclear power plants, anywhere conventional steel just can’t survive,” said Yu Zou, who led the project.

“We also need materials that can be printed layer by layer, enabling us to make components that can’t be created by traditional manufacturing processes. For example, to make a material that is both lightweight and strong, you can vary the composition: a hard, tough alloy on the outside to something softer and lighter on the inside.”

The project was partly supported by the University of Toronto’s Acceleration Consortium, a group that uses AI and automation to accelerate the discovery of new materials.

Active learning cuts the need for massive datasets

Most AI systems need a large amount of experimental data to make accurate predictions. This becomes a problem when researchers look at materials that have not been tested yet.

“One problem you often run into when trying to use AI to design materials is that most machine learning models require lots of data about material properties to learn from,” revealed Ajay Talbot, a Ph.D. student in Zou’s lab and lead author of the study.

“But if you’re working in part of the design space that hasn’t been explored yet, that data doesn’t exist, so you’re kind of flying blind.”

“The way we get around that challenge is to use data-lean models that essentially feel their own way along. Our active learning model strategically selects a few samples to manufacture and test, and the data from those experiments is fed back into the model to inform where we go next. It really speeds things up.”

<a href=”https://www.youtube.com/watch?v=RFYJV7WvtOM“>https://www.youtube.com/watch?v=RFYJV7WvtOM</a>

New alloys outperform industry benchmark

To demonstrate the system, the researchers focused on compositionally complex alloys made from nickel, cobalt, and chromium. Within weeks, the automated platform identified six new alloy formulations with strong performance.

“One of the properties we were targeting was puncture resistance at temperatures of up to 1,112°F (600°C), which is what you’d find in the front section of a jet engine,” Talbot said.

“The industry standard in this space is nickel-based alloys such as Inconel 625. But we found one made of 12 percent nickel, 62 percent cobalt, and 26 percent chrome that was great for retaining hardness at extremely high temperatures. Even with just three components, our alloy outperformed Inconel 625, an alloy of more than 10 different elements, by 4.5 percent in our lab tests.”

The team also developed another alloy designed for hotter sections of jet engines, where temperatures can reach 1,832°F (1,000°C).

“One of the things that happens in an environment like that is the formation of oxide scale, which essentially means that your material is just getting burned away,” Talbot explained.

“We found a material made of 36 percent nickel, 14 percent cobalt, and 50 percent chrome that was extremely good for oxidation resistance at these high temperatures: It even outperforms Inconel 625 by 85 percent. We’re eventually aiming to ramp up to even higher temperatures, up to 2,192°F (1,200°C).”

Researchers plan to expand to more complex materials

The researchers say the current alloys represent an early demonstration of what the AI-driven discovery platform can achieve.

“This nickel-cobalt-chrome system has just three elements in it. In the grand scheme of things, it’s a relatively simple system,” he added.

“But it’s great for showing that this whole closed-loop discovery platform really works. What we want to do next is ramp up the complexity a bit more to make even crazier stuff, with maybe up to 10 or 12 different elements.”

“As you add more components, you can get different strengthening mechanisms, different kinds of useful properties. There’s a lot more out there just waiting to be discovered.”

The study was published in the journal npj Advanced Manufacturing.

An era in which robots decide "how to walk" on their own has arrived. A four-legged robot has been developed that, much like a person or an animal, autonomously chooses the appropriate gait strategy for its surroundings—changing its gait on stairs, leaping over gaps and keeping its balance on forest trails.

Japanese communications company Fujitsu is leading a major push in artificial intelligence using Nvidia's technology, bringing together what it said was the best in Japan's manufacturing prowess in robotics with AI.

Xiaomi says its humanoid robot achieved a 98% success rate on a task at an EV factory, narrowing the gap with human operators.

Scientists hoped for years to make a machine that could emulate the movements of diving birds, such as puffins, and offer an affordable, unobtrusive way to monitor fragile marine ecosystems. A team of researchers at M.I.T. has been able to create such a waterproof winged robot, according to a study published July 9 in the journal Science.

A four-legged robot has learned to change the way it runs while navigating forests, staircases and obstacle courses. — seamlessly switching between a steady trot and a faster bounding gait without instructions from a human operator.

The 100-pound (45 kilograms) robot, called KAIST HOUND, uses cameras and lidar to scan the ground ahead, then selects an appropriate gait and adjusts its movements in real time. In outdoor tests, it crossed a 0.7-mile (1.1- kilometers) university campus route and a 0.2-mile (0.3 km) forest trail strewn with roots, logs and slippery leaves.

The researchers described the robotic framework on July 15 in the journal Science Robotics.

Changing gait

Animals naturally change their gait depending on their speed and surroundings. A dog might trot carefully across uneven ground, for example, before bounding over a fallen branch. Reproducing this adaptability in robots is tricky because different movements are often controlled by separate, highly specialized coding systems, and transitions between them can cause a lag that drives the robot to stumble.

To overcome this issue, researchers developed a special training framework called action pretrained transformer–based reinforcement learning (APT-RL). This is an artificial intelligence (AI) training system that first studies many examples of actions, uses a transformer to understand patterns across those actions, and then improves through rewards and penalties.

The training began with a simple, two-dimensional computer model of the robot. Using trajectory optimization — a technique that calculates physically workable movements for the robot — the team generated 180,000 short trotting and bounding sequences, including the joint forces the robot's legs need to perform. The dataset represented about 15.5 hours of movement but took only around eight minutes to produce.

During reinforcement learning — a machine learning technique where AI learns to make the best decisions by engaging with a particular environment through trial and error — an AI system then learned how to select and modify those skills while negotiating simulated stairs, stepping stones, hurdles, gaps and rough ground.

In digital simulations, the robot dog was not limited to copying its prerecorded movements. It could also make corrections for three-dimensional terrain and unexpected situations, such as jumping over a log — a behavior that wasn't included in the original, flat-ground training data.

Related stories

• Scientists found the optimal robot body, and it has 20 legs ‪—‬ watch it scale walls and move through trees
• This humanoid robot does all your housework for you ‪—‬ and its makers say it's ready for your home
• AI compressed billions of years of evolution into seconds to create 'Lego-like robots' that can recover even when they lose limbs

Finally, the researchers configured the system to include the robot's depth camera and lidar scanner in the simulation.

In one indoor test, HOUND bounded across an obstacle 2 feet (60 centimeters) high while briefly achieving 9.5 mph (15 km/h). It also jumped down a three-step staircase. The robot generally chose trotting at lower speeds on irregular ground, while bounding became more common at higher speeds or when it encountered larger steps, hurdles or gaps. The AI system that could select either gait performed more consistently across the different simulated environments than the version restricted to trotting or bounding alone.

The researchers suggest the technology could eventually help robots navigate disaster zones or other places inaccessible for wheeled machines. However, the current framework only allows two gait choices and mainly handles forward movement. Rapid turning, sideways motion and other behaviors like crawling remain future goals for the research team.

NAPA, Calif. – Icarus Robotics, the New York startup developing dexterous mobile robots for space missions, selected KULR Technology Group to provide batteries for Joy, a free-flying platform destined for […]

The post Icarus Robotics taps KULR to provide batteries for Space Station robots appeared first on SpaceNews.

Inspired by the physical feats of diving birds, researchers have created a robot that can plunge into the water and flap back into the air.

The VP of a humanoid robotics company was in for a surprise after finding that a Wired article pointed out something extremely unsettling about the company’s robot fingers.

Last week, Wired published a piece on Norwegian-American startup 1X and its Neo humanoid robot with an otherwise uncontroversial thesis: Though the marketing around Neo is decidedly carnal, its hands are impressive, matching and in some ways exceeding the dexterity of a human’s.

During the latter half of the article, Wired‘s Boone Ashworth noted a distinctly sensual pattern emerging in the company’s marketing materials: “smooth jazz plays in the background of a soft, warmly lit video. The robo digits curl around a wine glass, turn off a light, unzip a jacket, and gently fondle some grapes.”

“Not to kink shame, but this is a strange strategy for selling a robot that can also become a portal for human operators to peer into and interact with the things in your home,” Ashworth wrote, referring to the fact that the Neo will have remote human operators to start.

As reviews go, it’s far from the harshest critique of a robot ever written — and we should know — but 1X vice president of product and design Dar Sleeper still took Ashworth’s comments personally. Responding to the piece on X-formerly-Twitter, Sleeper lashed out at the author, using him as an example to lecture other tech journalists on what not to do.

“I gave Wired the exclusive on our hands launch, and they wrote a really weird article about how we are sexualizing robotics,” Sleeper wrote, displaying a remarkable amount of entitlement right out of the gate. “I felt pretty betrayed because that’s not what they told me they were writing about not is that what I’ve ever been about… but I’ve come to find a lot of dishonesty and malice in the journalism community so I wasn’t surprised.”

Continuing, Sleeper says he sent the following email to the Wired writer: “It was nice talking to you, but I wanted to let you know that I didn’t enjoy your article at all.”

I would like to note that I believe both Financial Times, Forbes & Bloomberg to be incredible institutions. You guys are killing it and I respect your ability to always innovate when things are changing.

— dar (@radbackwards) July 11, 2026

“I understand the need to be inflammatory because that seems to be the only thing that gets clicks these days but that doesn’t mean you shouldn’t recognize when something special is in front of you,” Sleeper carried on. “I trusted our PR team in saying we should offer you the exclusive on what is one of the most important technological developments in the history of Mankind and I deeply regret it.”

“Good luck with the rest of your writing career,” he added.

Clearly, 1X had something very different in mind with being featured in Wired, a highly reputable tech publication with decades of history. Sleeper’s comments highlight how the company was hoping to control the narrative.

If anything, the executive’s protest suggests that the “sexualization” of humanoid robots is an ongoing source of frustration for the company.

“This post makes it sound like the entire article is about the robot trying to make love to you,” tech journalist Joshua Topolsky tweeted, referring to Sleeper’s comments. “In fact, it’s roughly two paragraphs commenting on the marketing of the robot and the privacy issues of the Neo’s human control mode, where a person is remotely operating the bot, potentially inside your house and potentially in very private moments.”

“I appreciate that the founder embraces journalism and expected a certain framing, but this post kind of exaggerates a rather positive and mild take on his product into something it’s not,” Topolsky added. “If you want journalists to engage with your work, you must expect them to see things you might have missed. This seems to be one of those cases.”

More on robotics: Korean Workers Vote to Go On Strike, Fearing Robots Could Replace Them

The post Startup Upset After Journalist Points Out How Creepy Its Humanoid Robots’s Hands Are appeared first on Futurism.

Engineers have developed a new robot that can swim underwater, then rise to the surface and take flight, in an aerial-aquatic feat that mimics some of nature’s most formidable fliers and swimmers.

Developed by researchers at the Massachusetts Institute of Technology (MIT) and the Swiss Federal Institute of Technology Lausanne (EPFL), the new robot’s capabilities could help launch a new generation of aerial-aquatic exploration vehicles.

Dubbed the “Flapping-Wing Aerial-Aquatic Vehicle” (FAAV), this lightweight robotic platform weighs less than 300 grams and can transition seamlessly between water and air with its very flexible flapping wings. Now, the team’s findings, published in the journal Science, could help scientists better understand how diving birds, which include puffins, petrels, loons, and gulls, are so adept at navigating two vastly different environments.

Such capabilities, the team says, could also pave the way toward advanced drones in the coming years that could transform ocean research.

Inspired by Diving Birds

Departing from conventional modern designs employed for aircraft or underwater vehicles, the team based their design on diving birds, which evolved to fly over water, then plunge beneath the waves to hunt for food, before surfacing and resuming flight.

One problem when it comes to replicating these abilities in robots involves the density of water, which is around 1,000 times denser than air. Because of this, the engineering team had to carefully balance the flexibility of the robotic wings in their design, along with flapping frequency, tail angle, and other factors that allowed the robot to move efficiently through both environments without requiring propellers or separate propulsion systems.

Finding the Sweet Spot for Transmedium Travel

Experiments were undertaken by the team in the lab and in Switzerland’s Lake Geneva, and helped the team determine that the robot achieved best performance with medium-sized wings that flap at approximately five times per second. Under such conditions, the robot was able to swim one meter per second, then by pitching upward at a 70-degree angle, it could break through the surface, and continue this same movement to propel itself through the air.

Although the robot still has some limitations compared to the avian examples that inspired it, interestingly, it does appear to have one advantage: unlike diving birds, which paddle using their feet to achieve aquatic launches into the air, the new robot has no such requirement, and can transition from water to air relying only on its wings and adjustable tail.

Future Drones with “Aerial-Aquatic” Abilities

In the future, drones equipped with such capabilities could be leveraged to monitor marine ecosystems in a variety of ways. The team also says it could help to observe environmental concerns, such as water near glaciers or offshore infrastructure, or even collect data from hazardous locations that are either too difficult or simply too costly for ships to reach.

“Our dream vision is for oceanographers, marine biologists, and members of coastal communities to launch this robot from a boat, or from shore, and it would fly close to the area of interest, such as an iceberg or a port facility, or over a pod of whales,” according to Raphael Zufferey, assistant professor of mechanical engineering at MIT.

“It would dive into the water to take a measurement or collect a sample, and fly back to deliver the data at a fraction of the cost of traditional methods. Then it could go back out to dive for more,” Zufferey recently told MIT News.

Drawing from movements in nature influenced by evolution, the researchers hope their bird-inspired design may eventually help scientists by introducing cost-effective, but formidable new tools to help obtain information from some of the most challenging environments, all with help from robots that can do what only nature’s diving birds could once achieve.

The achievement is detailed in a new paper, “Leaping out of the water: aerial-aquatic locomotion with flapping wings,” which appeared in the journal Science.

Micah Hanks is the Editor-in-Chief and Co-Founder of The Debrief. A longtime reporter on science, defense, and technology with a focus on space and astronomy, he can be reached at micah@thedebrief.org. Follow him on X @MicahHanks, and at micahhanks.com.

Engineers at Queen Mary University of London have created a tactile sensor, described in a study published in Science Advances, that allows robots to detect touch using a simple camera.  

The sensor converts mechanical forces into visible color patterns, reducing the need for dense arrays of embedded sensors. Giacomo Sasso, a postdoctoral researcher at Queen Mary’s School of Engineering and Materials Science, led the team that developed this new sensor. Rather than relying on embedded electronic sensors, they created a soft material that changes color when pressed or bent.

This effect, known as structural color, is the same reason butterfly wings look iridescent without pigment. Pressure causes microscopic structures within the material to shift, altering how light reflects and creating a real-time visual indicator of force.

Touch Without the Computation

Most vision-based tactile sensors can map contact and force in detail, but only at the cost of heavy computation that slows them down. Simpler systems are faster, but sacrifice detail. The new material developed by Sasso’s team sidesteps this problem by building the sensing function into the material itself. With just a standard USB camera, researchers can track color changes on the surface and immediately infer pressure, strain, and contact shape without complex data reconstruction.

“You won’t guess how much information is generated when your finger presses a light switch. A human hand contains more than 10,000 mechanoreceptors to do the job, yet touch sensing remains one of the major challenges in robotics,” Sasso said. “We were happy to capture the finger ridges, as no existing technology can reproduce such sensor density at [a] comparable scale and simplicity. The key idea behind this project was to think outside the box: instead of embedding dense and overengineered sensor arrays, sensing is moved into the material itself, where mechanical cues are directly transformed into colour fields and captured using a simple low-cost USB camera.”

Professor James Busfield, one of the paper’s co-authors, described the advantage more simply: “What is particularly powerful is that the information is already in the light signal. You are no longer reconstructing touch – you are observing it directly.”

From Precision Manufacturing to the Operating Room

Reducing the computational demands of touch sensing opens up new possibilities. A robotic gripper made with this material could handle tiny parts more gently, since changes in force are immediately detectable. Prosthetic limbs equipped with similar sensors could also provide users with more detailed tactile feedback.

The researchers also see potential applications in surgery, where detecting subtle pressure differences could help clinicians interpret tissue properties.

A Decade of Groundwork Behind One Material

Professor Federico Carpi from the University of Florence and Professor James Busfield from Queen Mary worked together on this project, merging years of separate work on stretchable sensors and polymer behavior into one material system. This allowed the team to use a mechanochromic material as the sensor itself, not just as a base for electronics.

Many approaches to robotic touch have relied on increasingly complex sensor arrays, wiring, and computing systems. Rather than adding more electronics, Sasso’s team’s approach allows the material itself to encode touch into visible color. It remains unclear whether this color-based method can be scaled for use in commercial robotic hands, but this work suggests that the future of robotic touch might rely as much on smart materials as on advanced electronics.

Austin Burgess is a writer and researcher with a background in sales, marketing, and data analytics. He holds an MBA, a Bachelor of Science in Business Administration, and a data analytics certification. His work focuses on breaking scientific developments, with an emphasis on emerging biology, cognitive neuroscience, and archaeological discoveries.

Engineers have designed a waterproof "diving suit" for cyborg cockroaches that enables the hybrid insects to survive and roam underwater for up to three hours. This function expands the capabilities of cyborg insects and could one day be deployed in disaster zones, according to the team.

A built-in oxygen generator and silicone tubes deliver the gas directly to a cockroach's breathing holes, known as spiracles. The design is adapted for use in low-oxygen conditions as well as submerged environments, the researchers said in a new study published June 29 in the journal Nature Communications.

"Our approach combines a soft waterproof shell with a simple yet reliable chemical oxygen generator," study co-author Shinjiro Umezu, a professor in the School of Creative Science and Engineering at Waseda University in Japan, said in a statement. "This allows the insect to retain its natural mobility while being protected from an environment that it cannot normally survive in."

Cyborg insects are living insects that have been fitted with electronic controllers that guide their movements. Researchers have previously used them in search-and-rescue operations to access and investigate hard-to-reach areas; for example, they were used in rescue efforts after the devastating magnitude 7.7 earthquake in Myanmar in March 2025 that killed at least 3,700 people and injured 4,800 more. The advantage of cyborg insects over tiny robots is that the former employ insects' muscles to move, whereas the latter rely on high-power batteries that consume energy and can run out of steam.

The cyborg insects deployed in Myanmar were developed in the laboratory of Hirotaka Sato, senior author of the new study and a professor in the School of Mechanical and Aerospace Engineering at Singapore's Nanyang Technological University.

Sato has spent more than a decade pioneering cyborg insect technology. He and colleagues hope the new diving suit will extend cyborg insects' operational range to include flooded and partially submerged areas in disaster zones.

The suit consists of a flexible shell, four silicone tubes that attach to the spiracles and a transparent, 3D-printed oxygen tank. To make the tank produce oxygen, the researchers sprinkled manganese dioxide onto a highly absorbent sponge inside the tank. They then injected a small amount of diluted hydrogen peroxide, which breaks down slowly in the presence of manganese dioxide to produce oxygen. Finally, the team sealed the tank with ultraviolet adhesive to prevent leaks.

"The key engineering challenge was to build a system that was small, light and flexible enough for the insect to wear, while still producing enough oxygen for long-duration underwater movement," Umezu said.

The silicone tubes send oxygen straight into the thoracic spiracles, while the abdominal spiracles, which are lower down the insects' bodies, take in the oxygen contained in the suit.

"Our new insect diving suit works like the oxygen tank used by human divers," Sato said in the statement. The silicone tubes can be attached and removed without pain or harm to the insect, the researchers added.

RELATED STORIES

• Scientists found the optimal robot body, and it has 20 legs ‪—‬ watch it scale walls and move through trees
• MIT builds swarms of tiny robotic insect drones that can fly 100 times longer than previous designs
• AI compressed billions of years of evolution into seconds to create 'Lego-like robots' that can recover even when they lose limbs

The researchers tested the suit on a cyborg Madagascar hissing cockroach (Gromphadorhina portentosa), which they placed in a water tank and later sent into a plastic tube that simulated submerged and low-oxygen environments.

The suit enabled the cockroaches to roam underwater for up to three hours, raising the prospect that cyborg insects, including locusts and beetles, could one day be used to inspect flooded pipes, drains, tunnels and other hard-to-access places.

Next steps include improving the diving suit to potentially include sensors and a navigation system; and testing the design in simulated disaster environments, according to the statement.