L’intelligence artificielle (IA) est souvent saluée comme la solution aux défis les plus complexes de la médecine moderne. Qu’il s’agisse de prédire des infections chez les prématurés ou de recommander des traitements personnalisés, les outils d’IA promettent plus de précision et d’efficacité. En tant que chercheurs et médecins, nous voyons l’immense potentiel de l’IA. Mais nous constatons aussi un problème croissant : bien que l’IA soit conçue pour alléger la charge des cliniciens, elle pourrait en réalité l’aggraver. Comme nous l’avons souligné en mars dans un article publié dans le JAMA Health Forum, on demande aux médecins d’accomplir l’impossible. L’IA s’introduit dans les hôpitaux à un rythme que les réglementations ne peuvent pas suivre. Les médecins doivent décider – souvent seuls – quand se fier aux recommandations d’un algorithme et quand les rejeter. Et si le résultat est mauvais, ce sont eux qui en portent la responsabilité. Des recherches montrent que le public tend à blâmer davantage les médecins qui suivent une recommandation erronée d’IA que ceux qui suivent un avis humain fautif. Les médecins restent perçus comme les décideurs finaux, même lorsque la technologie échoue. Résultat : on demande aux médecins d’évaluer parfaitement la fiabilité d’outils qu’ils n’ont pas conçus et ne peuvent entièrement comprendre, tout en les tenant seuls responsables du résultat. Lire aussi | Intelligence artificielle et médecine : des promesses et beaucoup de questions Nous appelons cela le « dilemme surhumain du médecin ». La société attend depuis longtemps des médecins qu’ils soient infaillibles et l’IA ne fait qu’amplifier ces attentes. On demande désormais aux médecins d’interpréter instantanément des résultats algorithmiques, en naviguant entre deux risques constants : les faux positifs (faire confiance à une IA défectueuse) et les faux négatifs (rejeter une IA fiable). « Boîtes noires » Les voilà ainsi pris dans un double piège. S’ils suivent l’IA et que celle-ci se trompe, ils peuvent être blâmés pour avoir fait confiance à une machine. S’ils la rejettent et qu’un préjudice survient, ils peuvent être blâmés pour l’avoir ignorée. Cela ajoute de la pression, brouille leur jugement et érode la confiance, tant envers la technologie qu’envers eux-mêmes. Il vous reste 60.99% de cet article à lire. La suite est réservée aux abonnés.

Why it matters: A technological leap in fiber optics has shattered previous limitations, achieving what experts once considered impossible: transmitting data at 1.02 petabits per second – enough to download every movie on Netflix 30 times over – across 1,808 kilometers using a single fiber no thicker than a human hair. At the heart of this breakthrough – driven by Japan's National Institute of Information and Communications Technology (NICT) and Sumitomo Electric Industries – is a 19-core optical fiber with a standard 0.125 mm cladding diameter, designed to fit seamlessly into existing infrastructure and eliminate the need for costly upgrades. Each core acts as an independent data channel, collectively forming a "19-lane highway" within the same space as traditional single-core fibers. Unlike earlier multi-core designs limited to short distances or specialized wavelength bands, this fiber operates efficiently across the C and L bands (commercial standards used globally) thanks to a refined core arrangement that slashes signal loss by 40% compared to prior models. The experiment's success relied on a complex recirculating loop system. Signals traveled through an 86.1-kilometer fiber segment 21 times, simulating a cross-continental journey equivalent to linking Berlin to Naples or Sapporo to Fukuoka. To maintain integrity over this distance, researchers deployed a dual-band optical amplification system, comprising separate devices that boosted signals in the C and L bands. This enabled 180 distinct wavelengths to carry data simultaneously using 16QAM modulation, a method that packs more information into each pulse. // Related Stories At the receiving end, a 19-channel detector, paired with advanced MIMO (multiple-input multiple-output) processing, dissected interference between cores, much like untangling 19 overlapping conversations in a crowded room. Schematic diagram of the transmission system This digital signal processor, leveraging algorithms developed over a decade of multi-core research, extracted usable data at unprecedented rates while correcting for distortions accumulated over 1,808 km. The achievement caps years of incremental progress. In 2023, the same team achieved 1.7 petabits per second, but only across 63.5 km. Earlier efforts using 4-core fibers reached 0.138 petabits over 12,345 km by tapping the less practical S-band, while 15-mode fibers struggled with signal distortion beyond 1,001 km due to mismatched propagation characteristics. The new 19-core fiber's uniform core design sidesteps these issues, achieving a capacity-distance product of 1.86 exabits per second per kilometer – 14 times higher than previous records for standard fibers. Image diagram of 19-core optical fiber. Presented as the top-rated post-deadline paper at OFC 2025 in San Francisco, this work arrives as global data traffic is projected to triple by 2030. While challenges remain, such as optimizing amplifier efficiency and scaling MIMO processing for real-world use, the technology offers a viable path to petabit-scale networks. Researchers aim to refine production techniques for mass deployment, potentially enabling transoceanic cables that move entire data centers' worth of information hourly. Researchers aim to refine production techniques for mass deployment, potentially enabling transoceanic cables that move entire data centers' worth of information hourly. Sumitomo Electric's engineers, who designed the fiber's coupled-core architecture, note that existing manufacturing lines can adapt to produce the 19-core design with minimal retooling. Meanwhile, NICT's team is exploring AI-driven signal processing to further boost speeds. As 6G and quantum computing loom, this breakthrough positions fiber optics not just as a backbone for tomorrow's internet, but as the central nervous system of a hyperconnected planetary infrastructure.

Imagine: China deploys hundreds of thousands of autonomous drones in the air, on the sea, and under the water—all armed with explosive warheads or small missiles. These machines descend in a swarm toward military installations on Taiwan and nearby US bases, and over the course of a few hours, a single robotic blitzkrieg overwhelms the US Pacific force before it can even begin to fight back.  Maybe it sounds like a new Michael Bay movie, but it’s the scenario that keeps the chief technology officer of the US Army up at night. “I’m hesitant to say it out loud so I don’t manifest it,” says Alex Miller, a longtime Army intelligence official who became the CTO to the Army’s chief of staff in 2023. Even if World War III doesn’t break out in the South China Sea, every US military installation around the world is vulnerable to the same tactics—as are the militaries of every other country around the world. The proliferation of cheap drones means just about any group with the wherewithal to assemble and launch a swarm could wreak havoc, no expensive jets or massive missile installations required.  While the US has precision missiles that can shoot these drones down, they don’t always succeed: A drone attack killed three US soldiers and injured dozens more at a base in the Jordanian desert last year. And each American missile costs orders of magnitude more than its targets, which limits their supply; countering thousand-dollar drones with missiles that cost hundreds of thousands, or even millions, of dollars per shot can only work for so long, even with a defense budget that could reach a trillion dollars next year. The US armed forces are now hunting for a solution—and they want it fast. Every branch of the service and a host of defense tech startups are testing out new weapons that promise to disable drones en masse. There are drones that slam into other drones like battering rams; drones that shoot out nets to ensnare quadcopter propellers; precision-guided Gatling guns that simply shoot drones out of the sky; electronic approaches, like GPS jammers and direct hacking tools; and lasers that melt holes clear through a target’s side. Then there are the microwaves: high-powered electronic devices that push out kilowatts of power to zap the circuits of a drone as if it were the tinfoil you forgot to take off your leftovers when you heated them up.  That’s where Epirus comes in.  When I went to visit the HQ of this 185-person startup in Torrance, California, earlier this year, I got a behind-the-scenes look at its massive microwave, called Leonidas, which the US Army is already betting on as a cutting-edge anti-drone weapon. The Army awarded Epirus a $66 million contract in early 2023, topped that up with another $17 million last fall, and is currently deploying a handful of the systems for testing with US troops in the Middle East and the Pacific. (The Army won’t get into specifics on the location of the weapons in the Middle East but published a report of a live-fire test in the Philippines in early May.)  Up close, the Leonidas that Epirus built for the Army looks like a two-foot-thick slab of metal the size of a garage door stuck on a swivel mount. Pop the back cover, and you can see that the slab is filled with dozens of individual microwave amplifier units in a grid. Each is about the size of a safe-deposit box and built around a chip made of gallium nitride, a semiconductor that can survive much higher voltages and temperatures than the typical silicon.  Leonidas sits on top of a trailer that a standard-issue Army truck can tow, and when it is powered on, the company’s software tells the grid of amps and antennas to shape the electromagnetic waves they’re blasting out with a phased array, precisely overlapping the microwave signals to mold the energy into a focused beam. Instead of needing to physically point a gun or parabolic dish at each of a thousand incoming drones, the Leonidas can flick between them at the speed of software. The Leonidas contains dozens of microwave amplifier units and can pivot to direct waves at incoming swarms of drones.EPIRUS Of course, this isn’t magic—there are practical limits on how much damage one array can do, and at what range—but the total effect could be described as an electromagnetic pulse emitter, a death ray for electronics, or a force field that could set up a protective barrier around military installations and drop drones the way a bug zapper fizzles a mob of mosquitoes. I walked through the nonclassified sections of the Leonidas factory floor, where a cluster of engineers working on weaponeering—the military term for figuring out exactly how much of a weapon, be it high explosive or microwave beam, is necessary to achieve a desired effect—ran tests in a warren of smaller anechoic rooms. Inside, they shot individual microwave units at a broad range of commercial and military drones, cycling through waveforms and power levels to try to find the signal that could fry each one with maximum efficiency.  On a live video feed from inside one of these foam-padded rooms, I watched a quadcopter drone spin its propellers and then, once the microwave emitter turned on, instantly stop short—first the propeller on the front left and then the rest. A drone hit with a Leonidas beam doesn’t explode—it just falls. Compared with the blast of a missile or the sizzle of a laser, it doesn’t look like much. But it could force enemies to come up with costlier ways of attacking that reduce the advantage of the drone swarm, and it could get around the inherent limitations of purely electronic or strictly physical defense systems. It could save lives. Epirus CEO Andy Lowery, a tall guy with sparkplug energy and a rapid-fire southern Illinois twang, doesn’t shy away from talking big about his product. As he told me during my visit, Leonidas is intended to lead a last stand, like the Spartan from whom the microwave takes its name—in this case, against hordes of unmanned aerial vehicles, or UAVs. While the actual range of the Leonidas system is kept secret, Lowery says the Army is looking for a solution that can reliably stop drones within a few kilometers. He told me, “They would like our system to be the owner of that final layer—to get any squeakers, any leakers, anything like that.” Now that they’ve told the world they “invented a force field,” Lowery added, the focus is on manufacturing at scale—before the drone swarms really start to descend or a nation with a major military decides to launch a new war. Before, in other words, Miller’s nightmare scenario becomes reality.  Why zap? Miller remembers well when the danger of small weaponized drones first appeared on his radar. Reports of Islamic State fighters strapping grenades to the bottom of commercial DJI Phantom quadcopters first emerged in late 2016 during the Battle of Mosul. “I went, ‘Oh, this is going to be bad,’ because basically it’s an airborne IED at that point,” he says. He’s tracked the danger as it’s built steadily since then, with advances in machine vision, AI coordination software, and suicide drone tactics only accelerating.  Then the war in Ukraine showed the world that cheap technology has fundamentally changed how warfare happens. We have watched in high-definition video how a cheap, off-the-shelf drone modified to carry a small bomb can be piloted directly into a faraway truck, tank, or group of troops to devastating effect. And larger suicide drones, also known as “loitering munitions,” can be produced for just tens of thousands of dollars and launched in massive salvos to hit soft targets or overwhelm more advanced military defenses through sheer numbers.  As a result, Miller, along with large swaths of the Pentagon and DC policy circles, believes that the current US arsenal for defending against these weapons is just too expensive and the tools in too short supply to truly match the threat. Just look at Yemen, a poor country where the Houthi military group has been under constant attack for the past decade. Armed with this new low-tech arsenal, in the past 18 months the rebel group has been able to bomb cargo ships and effectively disrupt global shipping in the Red Sea—part of an effort to apply pressure on Israel to stop its war in Gaza. The Houthis have also used missiles, suicide drones, and even drone boats to launch powerful attacks on US Navy ships sent to stop them. The most successful defense tech firm selling anti-drone weapons to the US military right now is Anduril, the company started by Palmer Luckey, the inventor of the Oculus VR headset, and a crew of cofounders from Oculus and defense data giant Palantir. In just the past few months, the Marines have chosen Anduril for counter-drone contracts that could be worth nearly $850 million over the next decade, and the company has been working with Special Operations Command since 2022 on a counter-drone contract that could be worth nearly a billion dollars over a similar time frame. It’s unclear from the contracts what, exactly, Anduril is selling to each organization, but its weapons include electronic warfare jammers, jet-powered drone bombs, and propeller-driven Anvil drones designed to simply smash into enemy drones. In this arsenal, the cheapest way to stop a swarm of drones is electronic warfare: jamming the GPS or radio signals used to pilot the machines. But the intense drone battles in Ukraine have advanced the art of jamming and counter-jamming close to the point of stalemate. As a result, a new state of the art is emerging: unjammable drones that operate autonomously by using onboard processors to navigate via internal maps and computer vision, or even drones connected with 20-kilometer-long filaments of fiber-optic cable for tethered control. But unjammable doesn’t mean unzappable. Instead of using the scrambling method of a jammer, which employs an antenna to block the drone’s connection to a pilot or remote guidance system, the Leonidas microwave beam hits a drone body broadside. The energy finds its way into something electrical, whether the central flight controller or a tiny wire controlling a flap on a wing, to short-circuit whatever’s available. (The company also says that this targeted hit of energy allows birds and other wildlife to continue to move safely.) Tyler Miller, a senior systems engineer on Epirus’s weaponeering team, told me that they never know exactly which part of the target drone is going to go down first, but they’ve reliably seen the microwave signal get in somewhere to overload a circuit. “Based on the geometry and the way the wires are laid out,” he said, one of those wires is going to be the best path in. “Sometimes if we rotate the drone 90 degrees, you have a different motor go down first,” he added. The team has even tried wrapping target drones in copper tape, which would theoretically provide shielding, only to find that the microwave still finds a way in through moving propeller shafts or antennas that need to remain exposed for the drone to fly.  EPIRUS Leonidas also has an edge when it comes to downing a mass of drones at once. Physically hitting a drone out of the sky or lighting it up with a laser can be effective in situations where electronic warfare fails, but anti-drone drones can only take out one at a time, and lasers need to precisely aim and shoot. Epirus’s microwaves can damage everything in a roughly 60-degree arc from the Leonidas emitter simultaneously and keep on zapping and zapping; directed energy systems like this one never run out of ammo. As for cost, each Army Leonidas unit currently runs in the “low eight figures,” Lowery told me. Defense contract pricing can be opaque, but Epirus delivered four units for its $66 million initial contract, giving a back-of-napkin price around $16.5 million each. For comparison, Stinger missiles from Raytheon, which soldiers shoot at enemy aircraft or drones from a shoulder-mounted launcher, cost hundreds of thousands of dollars a pop, meaning the Leonidas could start costing less (and keep shooting) after it downs the first wave of a swarm. Raytheon’s radar, reversed Epirus is part of a new wave of venture-capital-backed defense companies trying to change the way weapons are created—and the way the Pentagon buys them. The largest defense companies, firms like Raytheon, Boeing, Northrop Grumman, and Lockheed Martin, typically develop new weapons in response to research grants and cost-plus contracts, in which the US Department of Defense guarantees a certain profit margin to firms building products that match their laundry list of technical specifications. These programs have kept the military supplied with cutting-edge weapons for decades, but the results may be exquisite pieces of military machinery delivered years late and billions of dollars over budget. Rather than building to minutely detailed specs, the new crop of military contractors aim to produce products on a quick time frame to solve a problem and then fine-tune them as they pitch to the military. The model, pioneered by Palantir and SpaceX, has since propelled companies like Anduril, Shield AI, and dozens of other smaller startups into the business of war as venture capital piles tens of billions of dollars into defense. Like Anduril, Epirus has direct Palantir roots; it was cofounded by Joe Lonsdale, who also cofounded Palantir, and John Tenet, Lonsdale’s colleague at the time at his venture fund, 8VC. (Tenet, the son of former CIA director George Tenet, may have inspired the company’s name—the elder Tenet’s parents were born in the Epirus region in the northwest of Greece. But the company more often says it’s a reference to the pseudo-mythological Epirus Bow from the 2011 fantasy action movie Immortals, which never runs out of arrows.)  While Epirus is doing business in the new mode, its roots are in the old—specifically in Raytheon, a pioneer in the field of microwave technology. Cofounded by MIT professor Vannevar Bush in 1922, it manufactured vacuum tubes, like those found in old radios. But the company became synonymous with electronic defense during World War II, when Bush spun up a lab to develop early microwave radar technology invented by the British into a workable product, and Raytheon then began mass-producing microwave tubes—known as magnetrons—for the US war effort. By the end of the war in 1945, Raytheon was making 80% of the magnetrons powering Allied radar across the world. From padded foam chambers at the Epirus HQ, Leonidas devices can be safely tested on drones.EPIRUS Large tubes remained the best way to emit high-power microwaves for more than half a century, handily outperforming silicon-based solid-state amplifiers. They’re still around—the microwave on your kitchen counter runs on a vacuum tube magnetron. But tubes have downsides: They’re hot, they’re big, and they require upkeep. (In fact, the other microwave drone zapper currently in the Pentagon pipeline, the Tactical High-power Operational Responder, or THOR, still relies on a physical vacuum tube. It’s reported to be effective at downing drones in tests but takes up a whole shipping container and needs a dish antenna to zap its targets.) By the 2000s, new methods of building solid-state amplifiers out of materials like gallium nitride started to mature and were able to handle more power than silicon without melting or shorting out. The US Navy spent hundreds of millions of dollars on cutting-edge microwave contracts, one for a project at Raytheon called Next Generation Jammer—geared specifically toward designing a new way to make high-powered microwaves that work at extremely long distances. Lowery, the Epirus CEO, began his career working on nuclear reactors on Navy aircraft carriers before he became the chief engineer for Next Generation Jammer at Raytheon in 2010. There, he and his team worked on a system that relied on many of the same fundamentals that now power the Leonidas—using the same type of amplifier material and antenna setup to fry the electronics of a small target at much closer range rather than disrupting the radar of a target hundreds of miles away.  The similarity is not a coincidence: Two engineers from Next Generation Jammer helped launch Epirus in 2018. Lowery—who by then was working at the augmented-reality startup RealWear, which makes industrial smart glasses—joined Epirus in 2021 to run product development and was asked to take the top spot as CEO in 2023, as Leonidas became a fully formed machine. Much of the founding team has since departed for other projects, but Raytheon still runs through the company’s collective CV: ex-Raytheon radar engineer Matt Markel started in January as the new CTO, and Epirus’s chief engineer for defense, its VP of engineering, its VP of operations, and a number of employees all have Raytheon roots as well. Markel tells me that the Epirus way of working wouldn’t have flown at one of the big defense contractors: “They never would have tried spinning off the technology into a new application without a contract lined up.” The Epirus engineers saw the use case, raised money to start building Leonidas, and already had prototypes in the works before any military branch started awarding money to work on the project. Waiting for the starting gun On the wall of Lowery’s office are two mementos from testing days at an Army proving ground: a trophy wing from a larger drone, signed by the whole testing team, and a framed photo documenting the Leonidas’s carnage—a stack of dozens of inoperative drones piled up in a heap.  Despite what seems to have been an impressive test show, it’s still impossible from the outside to determine whether Epirus’s tech is ready to fully deliver if the swarms descend.  The Army would not comment specifically on the efficacy of any new weapons in testing or early deployment, including the Leonidas system. A spokesperson for the Army’s Rapid Capabilities and Critical Technologies Office, or RCCTO, which is the subsection responsible for contracting with Epirus to date, would only say in a statement that it is “committed to developing and fielding innovative Directed Energy solutions to address evolving threats.”  But various high-ranking officers appear to be giving Epirus a public vote of confidence. The three-star general who runs RCCTO and oversaw the Leonidas testing last summer told Breaking Defense that “the system actually worked very well,” even if there was work to be done on “how the weapon system fits into the larger kill chain.” And when former secretary of the Army Christine Wormuth, then the service’s highest-ranking civilian, gave a parting interview this past January, she mentioned Epirus in all but name, citing “one company” that is “using high-powered microwaves to basically be able to kill swarms of drones.” She called that kind of capability “critical for the Army.”  The Army isn’t the only branch interested in the microwave weapon. On Epirus’s factory floor when I visited, alongside the big beige Leonidases commissioned by the Army, engineers were building a smaller expeditionary version for the Marines, painted green, which it delivered in late April. Videos show that when it put some of its microwave emitters on a dock and tested them out for the Navy last summer, the microwaves left their targets dead in the water—successfully frying the circuits of outboard motors like the ones propelling Houthi drone boats.  Epirus is also currently working on an even smaller version of the Leonidas that can mount on top of the Army’s Stryker combat vehicles, and it’s testing out attaching a single microwave unit to a small airborne drone, which could work as a highly focused zapper to disable cars, data centers, or single enemy drones.  Epirus’s microwave technology is also being tested in devices smaller than the traditional Leonidas. EPIRUS While neither the Army nor the Navy has yet to announce a contract to start buying Epirus’s systems at scale, the company and its investors are actively preparing for the big orders to start rolling in. It raised $250 million in a funding round in early March to get ready to make as many Leonidases as possible in the coming years, adding to the more than $300 million it’s raised since opening its doors in 2018. “If you invent a force field that works,” Lowery boasts, “you really get a lot of attention.” The task for Epirus now, assuming that its main customers pull the trigger and start buying more Leonidases, is ramping up production while advancing the tech in its systems. Then there are the more prosaic problems of staffing, assembly, and testing at scale. For future generations, Lowery told me, the goal is refining the antenna design and integrating higher-powered microwave amplifiers to push the output into the tens of kilowatts, allowing for increased range and efficacy.  While this could be made harder by Trump’s global trade war, Lowery says he’s not worried about their supply chain; while China produces 98% of the world’s gallium, according to the US Geological Survey, and has choked off exports to the US, Epirus’s chip supplier uses recycled gallium from Japan.  The other outside challenge may be that Epirus isn’t the only company building a drone zapper. One of China’s state-owned defense companies has been working on its own anti-drone high-powered microwave weapon called the Hurricane, which it displayed at a major military show in late 2024.  It may be a sign that anti-electronics force fields will become common among the world’s militaries—and if so, the future of war is unlikely to go back to the status quo ante, and it might zag in a different direction yet again. But military planners believe it’s crucial for the US not to be left behind. So if it works as promised, Epirus could very well change the way that war will play out in the coming decade.  While Miller, the Army CTO, can’t speak directly to Epirus or any specific system, he will say that he believes anti-drone measures are going to have to become ubiquitous for US soldiers. “Counter-UAS [Unmanned Aircraft System] unfortunately is going to be like counter-IED,” he says. “It’s going to be every soldier’s job to think about UAS threats the same way it was to think about IEDs.”  And, he adds, it’s his job and his colleagues’ to make sure that tech so effective it works like “almost magic” is in the hands of the average rifleman. To that end, Lowery told me, Epirus is designing the Leonidas control system to work simply for troops, allowing them to identify a cluster of targets and start zapping with just a click of a button—but only extensive use in the field can prove that out. Epirus CEO Andy Lowery sees the Leonidas as providing a last line of defense against UAVs.EPIRUS In the not-too-distant future, Lowery says, this could mean setting up along the US-Mexico border. But the grandest vision for Epirus’s tech that he says he’s heard is for a city-scale Leonidas along the lines of a ballistic missile defense radar system called PAVE PAWS, which takes up an entire 105-foot-tall building and can detect distant nuclear missile launches. The US set up four in the 1980s, and Taiwan currently has one up on a mountain south of Taipei. Fill a similar-size building full of microwave emitters, and the beam could reach out “10 or 15 miles,” Lowery told me, with one sitting sentinel over Taipei in the north and another over Kaohsiung in the south of Taiwan. Riffing in Greek mythological mode, Lowery said of drones, “I call all these mischief makers. Whether they’re doing drugs or guns across the border or they’re flying over Langley [or] they’re spying on F-35s, they’re all like Icarus. You remember Icarus, with his wax wings? Flying all around—‘Nobody’s going to touch me, nobody’s going to ever hurt me.’” “We built one hell of a wax-wing melter.”  Sam Dean is a reporter focusing on business, tech, and defense. He is writing a book about the recent history of Silicon Valley returning to work with the Pentagon for Viking Press and covering the defense tech industry for a number of publications. Previously, he was a business reporter at the Los Angeles Times. This piece has been updated to clarify that Alex Miller is a civilian intelligence official. 

Imagine downloading 10,000 4K movies in just a second. A team of Japanese researchers has achieved such a mind-blowing internet speed using a specially designed optical fiber that’s no thicker than what we use today. The researchers set a new world record, transmitting 1.02 petabits (1.02 x 106 GB) of data per second over a distance of 1,808 kilometers (~1,118 miles) using their special coupled 19-core optical fiber. However, this achievement isn’t just about faster internet.  In their new study, the researchers claim that their newly developed optical-fiber technology can help us prepare our networks for a future where data traffic will skyrocket, thanks to AI, 6G, the Internet of Things, and beyond. The science of insane internet speed For years, scientists have tried to increase the amount of data that can travel through optical fibers. While they’ve managed to send petabits per second before, they could only do it over short distances (less than 1,000 km or 621 miles).  Long-distance transmission has always been challenging. That’s because the signal weakens as it travels, and amplifying it across many fiber cores without creating interference is a major technical challenge. The study authors tackled the problem by designing a special type of optical fiber—a 19-core fiber.  Think of it like replacing a single-lane road with a 19-lane superhighway, all bundled into a fiber just 0.125 mm thick, the same size as those used in existing infrastructure. Each core carried data independently, and together they allowed a huge amount of information to move simultaneously. The researchers also developed a smart amplification system. Optical signals lose strength as they move along the fiber, so amplifiers are used to boost them. However, there’s one catch:  each core had to be amplified at the same time, and across two different bands of light (C-band and L-band).  The team built a system that used a combination of special amplifiers to do this in all 19 cores without mixing up the signals. They set up 19 recirculating loops, each using one core of the fiber, and passed the signals through them 21 times to simulate a total distance of 1,808 kilometers.  At the end of the journey, a 19-channel receiver caught the signals, and a multi-input multi-output (MIMO)-based digital processor cleaned them up, removing interference and calculating the data rate.  The result was astonishing. A total capacity of 1.02 petabits per second over 1,808 km was achieved, setting a new world record for optical fiber communication using standard-sized fibers. Even more impressive, the capacity-distance product, a key measure of fiber performance, reached 1.86 exabits per second-km, the highest ever recorded. A powerful and practical fiber technology A table comparing the performance of different fiber-optic cables. Source: NICT This isn’t the first time a 19-core optical fiber has been put to the test. “The transmission over an earlier generation of 19-core coupled-core fiber was limited to 1.7 petabits per second over a relatively short distance of 63.5 km,” the study authors added. However, this is indeed the first time that this revolutionary technology has broken the distance limits by carrying data over 1,800 km. This success could completely reshape how we build the internet of tomorrow.  As the world moves into the post-5G era, with self-driving cars, AI assistants, real-time VR, and billions of connected devices, we’ll need massive data highways to keep everything running.  “In the post-5G society, the volume of data traffic is expected to increase explosively due to new communication services, and the realization of advanced information and communication infrastructure is required,” the study authors added.RECOMMENDED ARTICLES This research shows that it’s possible to build ultra-high-speed, long-distance fiber networks without changing the size of existing infrastructure, which makes real-world deployment much easier. However, when this new optical fiber technology will actually roll out remains to be seen. The study was presented at the 48th Optical Fiber Communication Conference (OFC 2025).

With its gold accents, prominent control knobs, and guitar amp styling, Marshall’s hefty Heston 120 looks like no other soundbar on the planet. But what fascinates me about the company’s first TV speaker isn’t the styling (it looks exactly like I’d expect from a Marshall product), it’s how it’s been engineered to work with the company’s equally iconic portable Bluetooth speakers: It uses Bluetooth. Wait, I know that sounds obvious, but bear with me because this is actually a new and intriguing change to the way soundbars work. Recommended Videos Marshall Heston 120 Marshall First, a quick 101 on the Heston 120. It’s priced at $1,000, which should tell you right away that Marshall isn’t messing around. That’s the same price as the Sonos Arc Ultra and Bowers & Wilkins Panorama 3, and only $100 more than the Bose Smart Ultra Soundbar. It packs 11 drivers, including two dedicated subwoofers, and can process both Dolby Atmos and DTS:X in a 5.1.2-channel configuration. It has onboard mics that are used for room calibration, and it supports a wide array of protocols, including Apple AirPlay, Google Cast, Spotify Connect, and Tidal Connect. On the back panel, you get an Ethernet jack, an HDMI passthrough input with 4K/120Hz/Dolby Vision support, stereo RCA analog jacks (for a turntable or other gear), and a dedicated subwoofer output — something you rarely find on soundbars.  Marshall has redesigned its mobile app to give people deep controls over the Heston as well as the company’s full range of existing headphones, earbuds, and speakers. Expansion via Bluetooth Marshall Where things get interesting is on the wireless side of the equation. The Heston 120 supports Wi-Fi 6 and Bluetooth 5.3. That’s not unusual — all three of its competitors I mentioned above have the same or similar specs. What *is* unusual is how it uses these connections, specifically Bluetooth. Marshall considers the Heston 120 an all-in-one speaker that’s designed to work equally well for movies and music. However, the company also recognizes that some people want even more immersion from their TV sound systems, so it offers expansion via wireless speakers. Normally, when a soundbar is expandable with additional speakers, those connections are made via Wi-Fi (Sonos, Bluesound, Denon) or dedicated onboard transmitter/receivers (Bose, Sony, Klipsch). Bluetooth has never been considered a viable option because of issues around latency and limitations on transmitting multiple audio channels (e.g. low frequency, surround left, surround right) simultaneously. However, the Heston 120 is Bluetooth Auracast compatible — as far as I know, that’s a first for a soundbar — a technology that overcomes traditional Bluetooth limitations. Unlike earlier Bluetooth standards, which could create audio lag of 100-300 milliseconds, Auracast can achieve a latency of as little as 30 milliseconds. That should be almost imperceptible for dialogue synchronization, and even less noticeable for low-frequency bass or surround sound effects. Moreover, an Auracast device, like a TV or soundbar, can transmit multiple discrete broadcasts. In theory, it could handle multiple wireless subwoofers, two or four surround speakers, plus one or more wireless headphones or hearing aids — each with a dedicated sound stream. More choice, more flexibility Marshall Emberton III Marshall So what does this mean? Marshall’s ultimate goal is to let you use any pair of Auracast-capable Bluetooth speakers as your Heston 120 left/right surrounds, and an additional Auracast subwoofer for low-frequency effects. Initially, however, the plan is more conservative. At launch, the Heston 120 will support a single Marshall-built wireless subwoofer and later in the year you’ll be able to add two Marshall Bluetooth speakers as left/right surrounds. You’ll have a lot of choice — all of Marshall’s third-gen Homeline Bluetooth speakers are Auracast-ready — from the small but mighty Emberton III to the 120-watt Woburn III. Once they receive a planned firmware update, you can expect them all to work with the Heston as satellite speakers via Bluetooth. Typically, wireless surround speakers and subwoofers need to be plugged into a wall at all times. That provides power to the built-in amplifiers and their Wi-Fi network connections. Bluetooth, as a wireless technology, requires way less power than Wi-Fi, so if your Marshall portable Bluetooth speaker has a 20-hour battery, that’s 20 hours of completely wire-free home theater listening. And if, for some reason, you don’t have a Wi-Fi network, you can still assemble a multi-speaker system. Marshall points out that while Auracast is an open standard, each company can implement it as it sees fit, and that could mean that some Auracast speakers won’t work with the Heston 120. JBL Auracast speakers like the Charge 6 — for example — can only share and access audio from other JBL Auracast speakers. Still, Auracast-enabled soundbars like the Heston are opening up a new era in home theater technology; one where we’ll have a lot more freedom to choose the kind, number, and placement of speakers. It will also reduce the number of gadgets we buy. When your portable Bluetooth speaker can double as a surround speaker, that’s one less device in our ever-expanding world of tech. More options coming soon Auracast-enabled soundbars are the first step toward greater flexibility and choice in home theater. Soon, there will be more alternatives. Dolby has promised it will launch a soundbar alternative technology called Dolby Atmos FlexConnect, which will let a compatible TV send multichannel audio to a variety of wireless speakers that you’ll be able to place almost anywhere in your room. Fraunhofer IIS, the entity that gave us the MP3 file format, has its own version of FlexConnect — the somewhat awkwardly named UpHear Flexible Rendering. We haven’t seen any commercially available systems based on either Dolby’s or Fraunhofer’s tech so far, but I expect that to change in 2025.

The Unity for Humanity Summit is back for its third year on Wednesday, November 2, 2022. Join us for a virtual celebration of real-time 3D creators using Unity to change the world – registration is free and open to all.This year’s Summit includes an inspiring keynote conversation with actor and activist Rosario Dawson, who will discuss her sustainable fashion and nonprofit work, why positivity breeds innovation, and how technology can bring us all together around the virtual fire.In addition, you can attend a wealth of sessions led by artists, game developers, educators, nonprofit leaders, activists, and other changemakers across the core themes of sustainability, education, health and wellbeing, and inclusivity. Here are just a few of the exciting sessions on the agenda:Players first: Accessibility insights for creators features Dave Evans, Jazmin Cano, and Lukáš Hosnedl in a conversation about prioritizing accessibility in game development and the future of accessible gaming.From imagining to creating better worlds in RT3D brings together visionary creators Binh Minh Herbst, Katerina Cizek, Kathryn Evans, and Tony Patrick to discuss the process of translating ambitious visions into real-world action and impact.Create with AR Live and Unity Essentials Live showcase hands-on tools and resources for getting started in Unity and developing mobile AR experiences, led by Unity Learn team members Joy Horvath and Thomas Winkley.Activate your players: Nudging sustainable behavior and driving engagement provides insight into how game developers Hunter Bulkeley, Sheila Ndungu, Jude Ower, Galina Fedulova, and Jens Isensee are driving sustainability and empowering players to protect the planet through their games.Leveraging digital twins for sustainability in the built environment spotlights how organizations across the nonprofit, public, and private sectors are using real-time 3D for positive impact, hosted by Dr. Max Mallia-Parfitt, Dr. Sarah Whateley, and Ursula Smolka.Crafting Heroes: Students building the future using HoloLens 2 with Deidre LaCour and Sean Wybrant takes you straight into the inspiring XR work students are creating.Check out the full schedule and register now.Conversations with Creators are unique opportunities to meet fellow creators, ask them your burning questions, and immerse yourself in their stories, projects, and practical advice. Topics for these live Q&A sessions include:Addressing the youth mental health crisis with Cleo Barnett, co-creative director of AmplifierImmersive storytelling and impact with Gone to Water team Cat Ross and Marin VeselyThe world of immersive theater with Ferryman Collective cofounders Deirdre V. Lyons and Stephen ButchkoUnity for Humanity technical support Q&A with Andy Ellis, software development consulting manager at UnityDigital health and wellbeing with Gabriel G. Torres, creator of Haus of DustAR-tivism with Damien McDuffie, creator of Black TerminusPersonal storytelling for impact with Ondřej Moravec, Hana Blaha Šilarová, and Robin Pultera of DarkeningApplications for the Unity for Humanity 2023 Grant will open during the Summit, and we’ll be sharing tips for applying for the $500,000 available to bring your real-time 3D projects to life. Be sure to tune into the Unity for Humanity Grant Q&A session during the Summit to ask our program leads all of your questions, and sign up for the Social Impact mailing list to be notified of future grant announcements.Register now and join us at the Unity for Humanity Summit on November 2 to connect with social impact creators from across the globe and be part of the movement working towards a more equitable, inclusive, and sustainable world.

Scientists have designed an amplifier that can transmit 10 times more information per second than current fiber-optic systems can, which could be helpful for medical treatment and diagnosis.

Macworld Apple’s AirPods 2025 plans: In summary New AirPods Pro could launch featuring heart‑rate and temperature tracking and better ANC sound Some AirPods to get live translation feature via software update  AirPods Max could gain H2 chip, Adaptive Audio, Conversation awareness, Voice Isolation, Bluetooth 5.3 and more In September 2024, Apple introduced the AirPods 4. At the same time the company announced that hearing aid features would be coming to the 2nd generation AirPods Pro and the AirPods Max got a minor update (Lightning switched for USB-C and a color refresh). Those hoping that the AirPods Pro 3 and the AirPods Max 2 would launch were disappointed. It is now almost three years since the AirPods Pro 2 were introduced in September 2022, and more than four years since the AirPods Max introduction in December 2020. Updates are long overdue. In this article we are tracking all the rumors about the launch date, new features and specs and any other changes coming to the AirPods Pro and AirPods Max. Read on for the latest news about the 2025 AirPods. We’ll look first at the rumours and leaks relating to the AirPods Pro 3 before moving on to the AirPods Max 2 below (click here to jump to the AirPods Max section).  AirPods Pro 3 release date: When will the AirPods Pro 3 be released? In summary September 2025 launch of AirPods Pro 3 likely… …But launch could be delayed until 2026 Production delays indicated It would be typical to expect a September launch for AirPods Pro, alongside iPhone 17. The original AirPods Pro were introduced in October 2019, the AirPods 2nd-gen launched in September 2022, and those were updated with USB-C in September 2023. In a February 2025, Bloomberg report Mark Gurman suggested that the AirPods Pro 3 were “Still many months away”. That could still line up with a September (or October) launch though. Also in February an unverified leaker claimed in a tweet that there would be an AirPods Pro launch in May or June. Twitter However, in May 2025 analyst Ming-Chi Kuo suggested that we might not see an update to the AirPods until 2026. This could relate to production delays due to a new infra-red camera Apple is said to be equipping the AirPods with (more on that below). Twitter AirPods Pro 3 price: How much will the AirPods Pro 3 cost? In summary Price hike not expected… …But prices may rise due to economic factors We don’t expect the price to change, although it may increase in certain parts of the world and remain the same in others.  For example, when Apple launched the AirPods Pro in the U.K. the price was £249, that price was later reduced to £229, so it might go back up to £249. The price in the U.S. hasn’t changed since launch. Other factors may affect pricing. Pricing for the AirPods range starts at $129/£129 for the AirPods 4, and rises to $499/£549 for the AirPods Max:  Apple Earbuds: $19/£19 AirPods 4: $129/£129 AirPoids 4 with ANC: $179/£179 AirPods Pro 2nd-gen $249/£229 AirPods Max $549/£499 Ignoring the $300 premium for the AirPods Max, that’s an extra $70 between the AirPods 4 with ANC and the AirPods Pro. It would be feasible for Apple to increase the price of the AirPods Pro if the new features made the upgrade worthwhile.  We have a round-up of the Best wireless earbuds for iPhones and the best wired headphones if you’d like to consider alternatives to AirPods. Also read AirPods Pro vs AirPods to find out how the models compare. AirPods Pro 3 features: What specs and features will the AirPods Pro 3 have? In summary Heart rate monitoring Temperature monitoring Translation Better active noise cancellation It’s fairly easy to make predictions about the new features coming to AirPods Pro 3 just by looking at the AirPods 4 and the Beats Powerbeats Pro 2 (Beats is an Apple-owned subsidiary). For example, Powerbeats Pro 2, which launched in February 2025, includes heart rate monitoring. Read our review of the Powerbeats Pro 2 and our AirPods 4 review. As for the features rumored to be coming to the AirPods Pro 3, here’s what we’ve heard: Heart Rate monitoring: In December 2024, Gurman said Apple is working on in-ear heart rate monitoring. This will enable users to track heart rate while exercising without having to wear an Apple Watch.  The inclusion of heart rate monitoring in the Powerbeats Pro 2 certainly indicates that the same feature will be available on the AirPods Pro 3, but you can expect it to be superior. One issue with the Powerbeats Pro 2 is that they can’t steam music and capture heart data simultaneously. Another issue is with the interplay between the heart rate monitoring on the Apple Watch and the Powerbeats.  Temperature monitoring: Mark Gurman has also said Apple was researching in-ear temperature sensors, https://www.macworld.com/article/1978329/new-airpods-details-major-leak.html and Apple has a patent on such technology. In-ear temperature sensing could also be faster and more reliable than those via the Apple Watch. https://www.patentlyapple.com/2025/01/apple-reveals-an-all-new-temperature-sensor-that-could-be-used-with-apple-watch-airpods-and-more.html#:~:text=Apple Reveals an all-new,Electronic Devices With Temperature Sensors Translation: As part of the iOS 19 update Apple is said to be adding a live translation feature to AirPods. It’s unknown which AirPods will be compatible, but the AirPods Pro 3 are sure to be. https://www.macworld.com/article/2636989/airpods-to-get-live-translation-feature-as-part-of-ios-19-update.html We expect to hear more at WWDC. https://www.macworld.com/article/678333/wwdc-ios-macos-watchos-hardware-keynote.html Better active noise cancellation: According to Mark Gurman (in October 2023) the AirPods Pro 3 will feature a new H3 chip, which could lead to better active noise cancellation. Crash detection: An Apple patent describes a way that motion sensors inside AirPods could aid Crash Detection. AirPods Pro 3 design: Will Apple update the AirPods Pro design? In summary Changes to the charging case possible Slimmer design suggested Touch controls may work with gloves Mahmoud Itani / Foundry As mentioned above, we can look to the AirPods 4 for some clues about changes that may come to the AirPods Pro. One such change related to the charging case. The AirPod 4 charging case lacks a physical setup button, instead featuring a sensor under the status light on the front that lights up when active. There may be more changes coming. According to Mark Gurman (back in October 2023) the AirPods Pro will get a new design. We don’t expect new colors, but the buds may slim down further. The changes may be less visible, more tactile: A number of Apple patents indicate that Apple is looking to improve the way that the touch controls work on the earbuds work when the user is wearing gloves. For example in February 2025 a patent was granted to Apple that could enable this. AirPods Max 2 release date: When will the AirPods Max be updated? In summary Update long overdue Could be waiting until 2027 Bloomberg’s Mark Gurman wrote of the AirPods Max in a December 2023 newsletter, saying that “The cans don’t sell well enough for the company to invest in entirely new hardware or software features”. However, it’s now more than four years since the AirPods Max were introduced and with such a premium price Apple owes it to customers to bring the AirPods Max into line with their competition and other AirPods Apple sells.  In 2024 Mark Gurman predicted that the AirPods Max would get an updated at the iPhone event. Sadly, the update included only a color change and a switch from Lightning for USB-C connectivity.  I expect Apple to launch new AirPods Max headphones tomorrow with better noise cancellation, adaptive audio & USB-C. They’ll launch with the low-end AirPods 4. This is line with my reporting for months but runs counter to lack of low inventory in stores. https://t.co/KIJQEwywcp— Mark Gurman (@markgurman) September 9, 2024 Even worse, the change meant that the AirPods Max could no longer be used as wired headphones (a feature of the Lightning-based model). This was the case until a software update in March 2025 added this ability along with lossless and ultra low latency audio. As a result of that software update, users are now able to play lossless audio and ultra-low-latency audio through a wired connection. Lossless audio is a key feature for audio producers, while ultra‑low latency audio is key for gaming and live streams. Both very important features for the AirPods Mac. Despite these ‘updates’ the over four year old headphones are lagging behind the competition and an update is long over due. The bad news is that, if analyst Ming-Chi Kuo is correct, we could be waiting until 2027 for the second generation of AirPods Max. Twitter AirPods Max 2 price: How much will the AirPods Max 2 cost? In summary Price unlikely to change… …Although Apple could unify the U.S. and U.K. prices At $549/£499 the AirPods Max are an expensive purchase. The fact that they are now more than four years old makes that price look incredibly inflated. It’s no wonder they are frequently discounted at Amazon and elsewhere. When the new models launch we don’t expect a price change though. These are designed to fit at the premium headphones end of the market, and their competition has solar prices. That said, the price in the U.S. and U.K. may come in to line, with Apple choosing to ‘match’ the price as it usually does (many Apple products share an, e.g. $499/£499 price point) which Apple calculated based on exchange rates and local taxes. AirPods Max 2 specs: What specs and features will the AirPods Max 2 have In summary Necessary updates to bring into line with rest of the AirPods lineup New chip  Removal of ‘audio howl’ There is so much lacking from he AirPods Max compared to the AirPods Pro that it is easy to assume that the following will be included:  H1 chip upgraded to H2 (or H3)  Adaptive Audio (which combines transparency mode and ANC) Conversation Awareness  Conversation Boost Voice Isolation  Personalized Volume Option to activate with ‘Siri’ voice command (not just ‘Hey Siri’) Custom high dynamic range amplifier Bluetooth 5.3 (up from Bluetooth 5.0)  Find My  Also expect better battery life.  One patent identified in 2022 indicates Apple is looking at detecting “audio howl”, which is described as distracting audio feedback. AirPods Max 2 design: Will Apple update the AirPods Max design? In summary More protective carry case Removable ear cushions  Waterproofing  Jason Cross/IDG Improvements to gesture controls which are currently via the Digital Crown. Could Apple forgo the Digital Crown in favor of touch controls on the ear cups?  We could see changes to the carry case. Ideally it should cover the whole product and be more protective. A patent from 2022 suggested that Apple was developing a better case with magnetic clasps. The Audio Howl patent mentioned above also refers to removable ear cushions which could be reattached using magnets.  We’d also like to see a foldable design and better waterproofing. If you’re keen to hear the latest rumors about Apple’s plans, the best place to start is our roundup of New Apple products. Those interested in the current range should take a look at our roundup of the best AirPods deals for the latest bargains.