Jess Weatherbed reports via The Verge: It's alive! A day after announcing it was retiring Atlas, its hydraulic robot, Boston Dynamics has introduced a new, all-electric version of its humanoid machine. The next-generation Atlas robot is designed to offer a far greater range of movement than its predecessor. Boston Dynamics wanted the new version to show that Atlas can keep a humanoid form without limiting "how a bipedal robot can move." The new version has been redesigned with swiveling joints that the company claims make it "uniquely capable of tackling dull, dirty, and dangerous tasks."

The teaser showcasing the new robot's capabilities is as unnerving as it is theatrical. The video starts with Atlas lying in a cadaver-like fashion on the floor before it swiftly folds its legs backward over its body and rises to a standing position in a manner befitting some kind of Cronenberg body-horror flick. Its curved, illuminated head does add some Pixar lamp-like charm, but the way Atlas then spins at the waist and marches toward the camera really feels rather jarring. The design itself is also a little more humanoid. Similar to bipedal robots like Tesla's Optimus, the new Atlas now has longer limbs, a straighter back, and a distinct "head" that can swivel around as needed. There are no cables in sight, and its "face" includes a built-in ring light. It is a marked improvement on its predecessor and now features a bunch of Boston Dynamics' new AI and machine learning tools. [...] Boston Dynamics said the new Atlas will be tested with a small group of customers "over the next few years," starting with Hyundai.

Robotics firm Boston Dynamics, owned by Hyundai, has retired its humanoid robot Atlas after a decade, despite significant funding pouring into the category. TechCrunch adds: Boston Dynamics has been focused on commercializing technologies for a number of years now. Hyundai's 2021 acquisition of the firm, coupled with the appointment of Rob Playter as its second-ever CEO, has further accelerated that path. Given the tremendous interest around companies like Agility, Figure, 1X and Apptronik, it stands to reason that -- at the very least -- the Waltham, Massachusetts-based company has -- at the very least -- seriously explored the commercial humanoid category.

Boston Dynamics was, of course, well ahead of the current humanoid robotics curve. Last July marked the 10th anniversary of the bipedal robot's debut. The company teamed with DARPA for Atlas' early development, leading the robot to be heavily incorporated into challenges of the era.

Foreign Policy magazine visits a U.S. military training exercise that pitted Lt. Isaac McCurdy and his platoon of infantry troops against machines with camera lenses for eyes and sheet metal for skin: Driving on eight screeching wheels and carrying enough firepower on their truck beds to fill a small arms depot, a handful of U.S. Army robots stormed through the battlefield of the fictional city of Ujen. The robots shot up houses where the opposition force hid. Drones that had been loitering over the battlefield for hours hovered above McCurdy and his team and dropped "bombs" — foam footballs, in this case — right on top of them, a perfectly placed artillery shot. Robot dogs, with sensors for heads, searched houses to make sure they were clear.

"If you see the whites of someone's eyes or their sunglasses, [and] you shoot back at that, they're going to have a human response," McCurdy said. "If it's a robot pulling up, shooting something that's bigger than you can carry yourself, and it's not going to just die when you shoot a center mass, it's a very different feeling."

In the United States' next major war, the Army's brass is hoping that robots will be the ones taking the first punch, doing the dirty, dull, and dangerous jobs that killed hundreds — likely thousands — of the more than 7,000 U.S. service members who died during two decades of wars in the Middle East. The goal is to put a robot in the most dangerous spot on the battlefield instead of a 19-year-old private fresh out of basic training... [Several] Army leaders believe that almost every U.S. Army unit, down to the smallest foot patrols, will soon have drones in the sky to sense, protect, and attack. And it won't be long before the United States is deploying ground robots into battle in human-machine teams.
The robots haven't been tested with live ammunition yet — or in colder temperatures, the magazine notes. (And at one point in the exercise, "Army officials jammed themselves, and a swarm of drones dropped out of the sky.) But the U.S. Army is "considering a proposal to add a platoon of robots, the equivalent of 20 to 50 human soldiers, to its armored brigade combat team."

Six generals and several colonels watched the exercise, according to the article, which notes that the ultimate goal isn't to replace all human soldiers. "The point is to get the advantage before China or Russia do."

An anonymous reader shares a report: The story of warehouse robotics is a story of attempting to keep up with Amazon. It's been more than a decade since the online giant revolutionized its delivery services through its Kiva Systems acquisition. As Walmart works to remain competitive, it's taking a more piecemeal approach to automation, through partnerships with a range of different robotics firms.

On Thursday, the mega-retailer announced a partnership with Fox Robotics, which brings 19 of the Austin-based startup's robotic forklifts to its distribution centers. Today's news follows a 16-month pilot, which found Walmart trialing the technology in Distribution Center 6020. That Florida distribution center is the first of what the company calls its "high-tech DC." These are warehouses where it trials automation and various other technologies, before rolling them out to its wider channel of distribution and fulfillment centers. DC 6020 is the place where Walmart began trials with Symbotic's package sortation and retrieval technologies.

As reported by Bloomberg (paywalled), Apple is exploring the development of personal home robots following the shut down of its electric vehicle project. CNBC reports: Engineers at Apple have been looking into a robot that can follow users around their homes and a tabletop device that uses robotics to adjust a display screen, Bloomberg reported, citing people familiar with the research team. [...] Apple's hardware engineering division and its artificial intelligence and machine learning group are overseeing the work on personal robotics, Bloomberg reported. The home robot project is still in the early research and development phase, according to the report.

In a state where housing is expensive to build, to rent, or to buy — and not especially energy efficient — can a big blue robot make a difference?

The Boston Globe reports on Reframe Systems, one of the companies "trying robots to make construction more efficient" — in this case, "working alongside humans in an assembly line to build small houses in a factory." [Its cofounders] learned to get robots and humans to work together while at Amazon, which has built more than 750,000 bots in Massachusetts and deployed them to distribution centers around the world. Advising the company are Amy Villeneuve, former chief operating officer of that Amazon division, and Charly Mwangi, a veteran of the carmakers Nissan, Tesla, and Rivian...

Standing at one end of Reframe's factory, [cofounder Aaron] Small explained that the company's ambition is to build net-zero houses — houses that produce as much energy as they use — "twice as fast as traditional methods, twice as cheap, and with 10 times lower carbon" emissions. That means using large screws called helical piles to fix the house to the site, instead of a concrete foundation. (Concrete production generates large amounts of carbon dioxide.) The company buys recycled cellulose insulation to fill the walls. Solar panels go on the roof and triple-paned windows in the walls...

Reframe's "microfactory" can produce between 30 and 50 homes a year, [cofunder Vikas] Enti said. Eventually, the company aims to set up larger factories around the country, all within an hour's drive of big cities.
After a home is trucked to its final destination, "Electrical wires and plumbing are installed in both floors and walls as they're built," according to the article.

"Employees toting iPads can refer to digital construction drawings and get step-by-step instructions about tasks from cutting lumber to connecting pipes." One of the co-founders says, "We like to compare it to Lego instructions."

Mercedes-Benz is the latest automotive company to trial how humanoid robots could be used to automate "low skill, physically challenging, manual labor." From a report: On Friday, robotics company Apptronik announced it had entered into a commercial agreement with Mercedes to pilot how "highly advanced robotics" like Apollo -- Apptronik's 160-pound bipedal robot -- can be used in manufacturing. The news follows a similar pilot announced by BMW in January.

Apptronik says that Mercedes is exploring use cases like having Apollo inspect and deliver components to human production line workers. Neither company has disclosed any figures for the agreement or how many Apollo robots are being trialed. According to Apptronik, humanoid robots would allow vehicle manufacturers to start automating manufacturing tasks without having to redesign their existing facilities. The company says its approach instead "centers on automating some physically demanding, repetitive and dull tasks for which it is increasingly hard to find reliable workers."

"NASA said Friday it is shutting down a $2 billion satellite refueling project," reports UPI, "after criticizing the project's contractor for poor performance." The agency in a statement said it will discontinue the On-orbit Servicing, Assembly and Manufacturing 1 project after nearly a decade of work due to "continued technical, cost, and schedule challenges, and a broader community evolution away from refueling unprepared spacecraft, which has led to a lack of a committed partner." [...] The spacecraft would have utilized an attached Space Infrastructure Dexterous Robot (SPIDER) to refuel the Landsat, assemble a communications antenna and demonstrate in-space manufacture of a 32-foot carbon fiber composite beam to verify the capability of constructing large spacecraft structures in orbit... An audit from NASA's Inspector General, however, found OSAM-1 was on track to exceed the projected $2.05 billion budget and would not make its December 2026 launch date, laying the blame on the "poor performance of Maxar."

"NASA and Maxar officials acknowledged that Maxar underestimated the scope and complexity of the work, lacked full understanding of NASA technical requirements, and were deficient in necessary expertise," the report read.

The report also noted Maxar was "no longer profiting from their work on OSAM-1," after which the xproject appeared not "to be a high priority for Maxar in terms of the quality of its staffing."
Thanks to long-time Slashdot reader schwit1 for sharing the news.

OpenAI, Microsoft, Nvidia, and Jeff Bezos are all part of a pack of investors in a business "developing human-like robots," reports Bloomberg, "according to people with knowledge of the situation..."

At the startup — which is named "Figure" — engineers "are working on a robot that looks and moves like a human. The company has said it hopes its machine, called Figure 01, will be able to perform dangerous jobs that are unsuitable for people and that its technology will help alleviate labor shortages." Figure is raising about $675 million in a funding round that carries a pre-money valuation of roughly $2 billion, said the people, who asked not to be identified because the matter is private. Through his firm Explore Investments LLC, Bezos has committed $100 million. Microsoft is investing $95 million, while Nvidia and an Amazon.com Inc.-affiliated fund are each providing $50 million... Other technology companies are involved as well. Intel Corp.'s venture capital arm is pouring in $25 million, and LG Innotek is providing $8.5 million. Samsung's investment group, meanwhile, committed $5 million. Backers also include venture firms Parkway Venture Capital, which is investing $100 million, and Align Ventures, which is providing $90 million...

The AI robotics industry has been busy lately. Earlier this year, OpenAI-backed Norwegian robotics startup 1X Technologies AS raised $100 million. Vancouver-based Sanctuary AI is developing a humanoid robot called Phoenix. And Tesla Inc. is working on a robot called Optimus, with Elon Musk calling it one of his most important projects. Agility Robotics, which Amazon backed in 2022, has bots in testing at one of the retailer's warehouses. Bloomberg calls the investments in Figure "part of a scramble to find new applications for artificial intelligence."

The Washington Post reports: Google's self-driving car company, Waymo, is hitting resistance in its quest to expand 24/7 robotaxi service to other parts of California, including a series of incidents that have fed public officials' safety concerns about the vehicles coming to their cities. Over eight days in February, for example, a Waymo vehicle smashed into a closing gate while exiting the University of Southern California's campus; the next day, another collided with a cyclist in San Francisco. Later that week, a mob of people vandalized and lit one of its cars on fire. Days later, the company announced a voluntary recall of its software for an incident involving a pickup truck in Phoenix. [Though it occurred three months ago, the Post reports that after the initial contact between the vehicles, "A second Waymo vehicle made contact with the pickup truck a few minutes later."]

This string of events — none of which resulted in serious injuries — comes after Waymo's main competitor, General Motors-owned Cruise, recalled its fleet of driverless cars last year... [Waymo] is now the lone company trying to expand 24/7 robotaxi service around California, despite sharp resistance from local officials. "Waymo has become the standard-bearer for the entire robotaxi industry for better or for worse," said David Zipper, a senior fellow at the MIT Mobility Initiative. While Waymo's incidents are "nowhere near what Cruise is accused of doing, there is a crisis of confidence in autonomous vehicle companies related to safety right now."

The California Public Utilities Commission (CPUC) delayed deciding whether Waymo could expand its service to include a portion of a major California highway and also Los Angeles and San Mateo counties, pending "further staff review," according to the regulator's website. While Waymo said the delay is a part of the commission's "standard and robust review process," the postponement comes as officials from other localities fear becoming like San Francisco — where self-driving cars have disrupted emergency scenes, held up traffic and frustrated residents who are learning to share public roads with robot cars... Zipper said it is a notable disparity that "the companies are saying the technology is supposed to be a godsend for urban life, and it's pretty striking that the leaders of these urban areas really don't want them," he said.
Waymo offers ride-hailing services in San Francisco and Phoenix — as well as some free rides in Los Angeles, according to the article. It also cites a December report from Waymo estimated that overich 7.1 million miles of testing, there were 17 fewer injuries and 20 fewer police-reported crashes "compared to if human drivers with the benchmark crash rate would have driven the same distance in the areas we operate."

On Thursday, the Odysseus Moon lander made history by becoming the first ever privately built and operated robot to complete a soft lunar touchdown. While the lander is "alive and well," the CEO of Houston-based Intuitive Machines, which built and flew the lander, said it tipped over during its final descent, coming up to rest propped up sideways on a rock. The BBC reports: Its owner, Texan firm Intuitive Machines, says Odysseus has plenty of power and is communicating with Earth. Controllers are trying to retrieve pictures from the robot. Steve Altemus, the CEO and co-founder of IM, said it wasn't totally clear what happened but the data suggested the robot caught a foot on the surface and then fell because it still had some lateral motion at the moment of landing. All the scientific instruments that planned to take observations on the Moon are on the side of Odysseus that should still allow them to do some work. The only payload likely on the "wrong side" of the lander, pointing down at the lunar surface, is an art project.

"We're hopeful to get pictures and really do an assessment of the structure and assessment of all the external equipment," Mr Altemus told reporters. "So far, we have quite a bit of operational capability even though we're tipped over. And so that's really exciting for us, and we are continuing the surface operations mission as a result of it." The robot had been directed to a cratered terrain near the Moon's south pole, and the IM team believes it got very close to the targeted site - perhaps within a couple of kilometers. A US space agency satellite called the Lunar Reconnaissance Orbiter will search for Odysseus in the coming days.

Tom McKay successfully masqueraded as a "Slackbot" on Slack after leaving Gizmodo in 2022, going unnoticed by the site's management for several months. The Verge reports: If you're not glued to Slack for most of the day like I am, then you might not know that Slackbot is the friendly robot that lives in the messaging service. It helps you do things like set reminders, find out your office's Wi-Fi password, or let you know when you've been mentioned in a channel that you're not a part of. When it was his time to leave, McKay swapped out his existing profile picture for one that resembled an angrier version of Slackbot's actual icon. He also changed his name to "Slackbot." You can't just change your name on Slack to "Slackbot," by the way, as the service will tell you that name's already been taken. It does work if you use a special character that resembles one of the letters inside Slackbot, though, such as replacing "o" with the Unicode character "o."

The move camouflaged McKay's active Slack account for months, letting his account evade deletion. It also allowed him to send bot-like messages to his colleagues such as, "Slackbot fact of the day: Hi, I'm Slackbot! That's a fact. Have a Slack-ly day!" My colleague Victoria Song, who previously worked at Gizmodo, isn't all that surprised that this situation unfolded, and says, "As Tom's former coworker and a G/O Media survivor, this tracks."

India's Space Research Organisation (ISRO) will send a humanoid robot astronaut into this space this year, then send it back alongside actual humans in 2025 on its long-delayed Gaganyaan orbital mission. From a report: According to the space agency, the robot-crewed Vyommitra Mission is scheduled for the third quarter of this year. The robot -- whose name translates to "Space Friend" in Sanskrit -- can monitor module parameters, issue alerts and execute life support operations. Vyommitra is also an excellent multitasker that can operate six panels while responding to queries and mimicking human functions. The humanoid speaks two languages: Hindi and English.

It's also been designated as female -- to the extent possible for a legless robot -- and sports coiffed hair, feminine facial features, and hands that look like they are wearing white gloves. It resembles a wax figurine or mannequin and The Register fancies it mostly manages to stay out of the Uncanny Valley -- the term applied to robots and digital depictions of humans that try to appear human but instead come off as creepy and/or unsettling.

In a new video released today, Boston Dynamics' Atlas robot is shown performing "kinetically challenging" work, like moving some medium-weight car parts and precisely picking stuff up. Ars Technica reports: In the latest video, we're on to what looks like "phase 2" of picking stuff up -- being more precise about it. The old clamp hands had a single pivot at the palm and seemed to just apply the maximum grip strength to anything the robot picked up. The most delicate thing Atlas picked up in the last video was a wooden plank, and it was absolutely destroying the wood. Atlas' new hands look a lot more gentle than The Clamps, with each sporting a set of three fingers with two joints. All the fingers share one big pivot point at the palm of the hand, and there's a knuckle joint halfway up the finger. The fingers are all very long and have 360 degrees of motion, so they can flex in both directions, which is probably effective but very creepy. Put two fingers on one side of an item and the "thumb" on the other, and Atlas can wrap its hands around objects instead of just crushing them.

Atlas is picking up a set of car struts -- an object with extremely complicated topography that weighs around 30 pounds -- so there's a lot to calculate. Atlas does a heavy two-handed lift of a strut from a vertical position on a pallet, walks the strut over to a shelf, and carefully slides it into place. This is all in Boston Dynamics' lab, but it's close to repetitive factory or shipping work. Everything here seems designed to give the robot a manipulation challenge. The complicated shape of the strut means there are a million ways you could grip it incorrectly. The strut box has tall metal poles around it, so the robot needs to not bang the strut into the obstacle. The shelf is a tight fit, so the strut has to be placed on the edge of the shelf and slid into place, all while making sure the strut's many protrusions won't crash into the shelf.

The "technology building blocks" for space solar are already available, reports Clean Technica. "It's just a matter of scaling, systems integration, and adjustments for space-hardiness."

And several groups are looking at it — including the U.S. Space Force To help push costs down, the California Institute of Technology has proposed a sandwich-type solar module that integrates solar harvesting along with conversion to a radio frequency into one compact package, accompanied by a built-in antenna. Last month researchers at the school wrapped up a months-long, in-space test of different types of solar cells. Another approach is illustrated by the Michigan startup Virtus Solis, an industry partner of the University of Bristol. Last June the company and the school received £3.3 million in funding from the UK Net Zero Innovation program, for developing an open-source model for testing the performance of large, centralized antennas in space. "The concept depends upon the use of gigascale antenna arrays capable of delivering over 2GW of power from space onto similar gigascale antenna arrays either at sea or on the ground," the school explained.

As for how such a thing would be launched into space, that's where the U.S. Space Force comes in. Last August, the Space Force awarded a small business contract to the U.S. startup Orbital Composites. The company is tasked with the mission of developing its patented "quantum antenna" and in-space fabrication tools for secure communications in space applications, including space-to-space as well as space-to-Earth and vice versa. The basic idea is to let 3D printing doing much of the work in space. According to Orbital, in-space fabrication would save more than 100 times the cost of applying conventional fabrication methods to large-scale orbiting antennas. "By harnessing the potential of In-Space Servicing, Assembly, and Manufacturing (ISAM), the company eyes the prospect of creating significantly larger space antennas," Orbital Composites explains. "By fabricating antennas in space, larger and more complex designs are possible that eliminate the constraints of launch and rocket fairings...."

If you're guessing that a hookup between Virtus and Orbital is in the works, that's a good guess. On February 1, at the SpaceCOM conference in Orlando, Florida, Virtus Solis let slip that it is working with Orbital Composites on a space solar pilot project. If all goes according to plan, the project will be up and running in 2027, deploying Virtus's robot-enabled fabrication system with Orbital's 3D printing. As of this writing the two companies have not posted details, but Space News picked up the thread. "The 2027 mission is designed to showcase critical power-generation technologies including in-space assembly of solar panels and transmission of more than one kilowatt to Earth," Space News explained. "The news release calls the 2027 mission "a precursor to large-scale commercial megawatt-class solar installations in space by 2030...."

To be clear, Orbital's press release about its new Space Force quantum antenna contract does not mention anything in particular about space solar. However, the pieces of the puzzle fit. Along with the Virtus and Grumman connections, in October of 2022 Orbital won a small business contract through SpaceWERX, the Space Force's innovative technologies funding arm, to explore the capabilities of ISAM systems.
"SpaceWERX comes under the umbrella of the U.S. Air Force's AFWERX innovation branch, which has developed a program called SSPIDR, short for Space Solar Power Incremental Demonstrations and Research Project," the article points out. (While Virtus believes most space-based solar power systems could deliver megawatt hours of electricity at prices comparable to today's market.)

Quora "used to be a thriving community that worked to answer our most specific questions," writes Slate. "But users are fleeing," while the site hosts "a never-ending avalanche of meaningless, repetitive sludge, filled with bizarre, nonsensical, straight-up hateful, and A.I.-generated entries..."

The site has faced moderation issues, spam, trolls, and bots re-posting questions from Reddit (plus competition for ad revenue from sites like Facebook and Google which forced cuts in Quora's support and moderation teams). But automating its moderation "did not improve the situation...

"Now Quora is even offering A.I.-generated images to accompany users' answers, even though the spawned illustrations make little sense." To top it all off, after Quora began using A.I. to "generate machine answers on a number of selected question pages," the site made clear the possibility that human-crafted answers could be used for training A.I. This meant that the detailed writing Quorans provided mostly for free would be ingested into a custom large language model. Updated terms of service and privacy policies went into effect at the site last summer. As angel investor and Quoran David S. Rose paraphrased them: "You grant all other Quora users the unlimited right to reuse and adapt your answers," "You grant Quora the right to use your answers to train an LLM unless you specifically opt out," and "You completely give up your right to be any part of any class action suit brought against Quora," among others. (Quora's Help Center claims that "as of now, we do not use answers, posts, or comments added to Quora to train LLMs used for generating content on Quora. However, this may change in the future." The site offers an opt-out setting, although it admits that "opting out does not cover everything.")

This raised the issue of consent and ownership, as Quorans had to decide whether to consent to the new terms or take their work and flee. High-profile users, like fantasy author Mercedes R. Lackey, are removing their work from their profiles and writing notes explaining why. "The A.I. thing, the terms of service issue, has been a massive drain of top talent on Quora, just based on how many people have said, Downloaded my stuff and I'm out of there," Lackey told me. It's not that all Quorans want to leave, but it's hard for them to choose to remain on a website where they now have to constantly fight off errors, spam, trolls, and even account impersonators....

The tragedy of Quora is not just that it crushed the flourishing communities it once built up. It's that it took all of that goodwill, community, expertise, and curiosity and assumed that it could automate a system that equated it, apparently without much thought to how pale the comparison is. [Nelson McKeeby, an author who joined Quora in 2013] has a grim prediction for the future: "Eventually Quora will be robot questions, robot answers, and nothing else." I wonder how the site will answer the question of why Quora died, if anyone even bothers to ask.
The article notes that Andreessen Horowitz gave Quora "a much-needed $75 million investment — but only for the sake of developing its on-site generative-text chatbot, Poe."

According to Neuralink founder Elon Musk, the first human received an implant from the brain-chip startup on Sunday and is recovering well. "Initial results show promising neuron spike detection," Musk added. Reuters reports: The U.S. Food and Drug Administration had given the company clearance last year to conduct its first trial to test its implant on humans. The startup's PRIME Study is a trial for its wireless brain-computer interface to evaluate the safety of the implant and surgical robot. The study will assess the functionality of the interface which enables people with quadriplegia, or paralysis of all four limbs, to control devices with their thoughts, according to the company's website.

Figure's first humanoid robot will be coming to a BMW manufacturing facility in South Carolina. TechCrunch reports: BMW has not disclosed how many Figure 01 models it will deploy initially. Nor do we know precisely what jobs the robot will be tasked with when it starts work. Figure did, however, confirm with TechCrunch that it is beginning with an initial five tasks, which will be rolled out one at a time. While folks in the space have been cavalierly tossing out the term "general purpose" to describe these sorts of systems, it's important to temper expectations and point out that they will all arrive as single- or multi-purpose systems, growing their skillset over time. Figure CEO Brett Adcock likens the approach to an app store -- something that Boston Dynamics currently offers with its Spot robot via SDK.

Likely initial applications include standard manufacturing tasks such as box moving, pick and place and pallet unloading and loading -- basically the sort of repetitive tasks for which factory owners claim to have difficulty retaining human workers. Adcock says that Figure expects to ship its first commercial robot within a year, an ambitious timeline even for a company that prides itself on quick turnaround times. The initial batch of applications will be largely determined by Figure's early partners like BMW. The system will, for instance, likely be working with sheet metal to start. Adcock adds that the company has signed up additional clients, but declined to disclose their names. It seems likely Figure will instead opt to announce each individually to keep the news cycle spinning in the intervening 12 months.

samleecole writes: A food delivery robot company instructed a public university to promote its service on campus with photographs and video featuring only students who "have a healthy-looking BMI," [body mass index] according to emails and documents I obtained via a public records request. The emails also discuss how ordering delivery via robot should become a "habit" for a "captured" customer base of students on campus.

These highly specific instructions show how universities around the country are going to extreme lengths to create a welcoming environment on campus for food delivery robots that sometimes have trouble crossing the street and need traffic infrastructure redesigned for them in order to navigate campus, a relatively absurd cache of public records obtained by 404 Media reveals.

Generative AI "doesn't easily carry over into robotics," write two researchers in IEEE Spectrum, "because the Internet is not full of robotic-interaction data in the same way that it's full of text and images."

That's why they're working on a single deep neural network capable of piloting many different types of robots... Robots need robot data to learn from, and this data is typically created slowly and tediously by researchers in laboratory environments for very specific tasks... The most impressive results typically only work in a single laboratory, on a single robot, and often involve only a handful of behaviors... [W]hat if we were to pool together the experiences of many robots, so a new robot could learn from all of them at once? We decided to give it a try. In 2023, our labs at Google and the University of California, Berkeley came together with 32 other robotics laboratories in North America, Europe, and Asia to undertake the RT-X project, with the goal of assembling data, resources, and code to make general-purpose robots a reality...

The question is whether a deep neural network trained on data from a sufficiently large number of different robots can learn to "drive" all of them — even robots with very different appearances, physical properties, and capabilities. If so, this approach could potentially unlock the power of large datasets for robotic learning. The scale of this project is very large because it has to be. The RT-X dataset currently contains nearly a million robotic trials for 22 types of robots, including many of the most commonly used robotic arms on the market...

Surprisingly, we found that our multirobot data could be used with relatively simple machine-learning methods, provided that we follow the recipe of using large neural-network models with large datasets. Leveraging the same kinds of models used in current LLMs like ChatGPT, we were able to train robot-control algorithms that do not require any special features for cross-embodiment. Much like a person can drive a car or ride a bicycle using the same brain, a model trained on the RT-X dataset can simply recognize what kind of robot it's controlling from what it sees in the robot's own camera observations. If the robot's camera sees a UR10 industrial arm, the model sends commands appropriate to a UR10. If the model instead sees a low-cost WidowX hobbyist arm, the model moves it accordingly.
"To test the capabilities of our model, five of the laboratories involved in the RT-X collaboration each tested it in a head-to-head comparison against the best control system they had developed independently for their own robot... Remarkably, the single unified model provided improved performance over each laboratory's own best method, succeeding at the tasks about 50 percent more often on average." And they then used a pre-existing vision-language model to successfully add the ability to output robot actions in response to image-based prompts.

"The RT-X project shows what is possible when the robot-learning community acts together... and we hope that RT-X will grow into a collaborative effort to develop data standards, reusable models, and new techniques and algorithms."

Thanks to long-time Slashdot reader Futurepower(R) for sharing the article.