Geometric alignment appears in a variety of applications, ranging from domain adaptation, optimal transport, and normalizing flows in machine learning; optical flow and learned augmentation in computer vision and deformable registration within biomedical imaging. A recurring challenge is the alignment of domains whose topology is not the same; a problem that is routinely ignored, potentially introducing bias in downstream analysis. As a first step towards solving such alignment problems, we propose an unsupervised algorithm for the detection of changes in image topology. The model is based on a conditional variational auto-encoder and detects topological changes between two images during the registration step. We account for both topological changes in the image under spatial variation and unexpected transformations. Our approach is validated on two tasks and datasets: detection of topological changes in microscopy images of cells, and unsupervised anomaly detection brain imaging.

If you thought robot dogs were the coolest animatronic animals out there, prepare to think again. Kawasaki Heavy Industries, a company better known for its high-end motorcycles, has unveiled a hydrogen-powered, ride-on robo-horse. The bizarre device was unveiled at the Osaka Kansai Expo on April 4 as part of Kawasaki's'Impulse to Move' project. Dubbed the CORLEO, this two-seater quadruped is capable of galloping over almost any terrain. The company calls it a'revolutionary off-road personal mobility vehicle' which swaps out the familiar wheels for four robotic legs. To steer, all you need to do is move your body and the machine's AI vision will pick out the best route to take.

Our old friend Spot the robot dog is joining the Big Apple's police force. New York City mayor Eric Adams announced that the New York Police Department will be acquiring some new semi-autonomous robotic canines in the coming weeks. The move comes almost exactly two years after the NYPD halted its first go at using a camera-carrying robot dog for surveillance, after a massive public outcry; citizens felt it was a dystopian overreach of police power. Now Adams, a former NYPD captain, is moving the program forward again. The NYPD says it will acquire two of Boston Dynamic's controversial Spot bots. While the robot dogs have autonomous capabilities, the NYPD says these units won't be patrolling the streets by themselves just yet.

U.S.-based Boston Dynamics' Spot is distinguishedly the market leader regarding these robots. Interesting Engineering has previously reported on many of Spot's antics and cuteness. Through videos such as the one shared above, Unitree is also looking to pique the cuteness quotient of its offerings. However, there are many other reasons why one could pick a Unitree robotic dog. Unitree's robotic dog, Go1, does not boast bright colors and only has a silvery metallic appearance.

For starters, it has no head. And instead of kibble and water, it runs on a lithium-ion battery. When the four-legged robot, which can climb stairs, open doors and transmit 360-degree video, was unveiled a few years ago, it was billed as a potent new tool for industries whose workers are often in dangerous conditions. It could, for example, detect radiation for an energy company or inspect the safety of a mining tunnel, its creator, Boston Dynamics, touted said in promotional material. And police officials around the U.S. realized Spot, which its inventors named, also offered an upgrade from the slower, less agile robots currently used in hostage situations, assessing suspicious packages and other high-risk situations.

Boston Dynamics has released an incredible video of a troupe its famous Spot robotic dogs pulling off some very impressive dance moves. Seven of the robots can be seen performing a choreographed routine to the hit song'Permission to Dance', by K-Pop band BTS. Initially, one dog appears to'sing' the solo parts of the songs by grasping its robotic arm in time to the words, while the others step in the background. When the chorus kicks in, they begin a series of synchronised moves in different formations, as if they were the boy band themselves. There is even a cameo from Atlas, a six-foot-tall bipedal humanoid robot also developed by the Boston-based firm, who jumps and claps to the beat.

Boston Dynamics, the DARPA-backed robotics company known for uncomfortable videos where nearly 200-pound humanoid robots perform backflips, uncomfortable dances, and various forms of horrifyingly aggressive parkour, says it isn't interested in weaponizing its robots. In an open letter this week, Boston Dynamics Dynamics joined five other robotics makers in a pledge not to weaponize their advanced-mobility, general-purpose robots, or the software that makes them tick. The companies said they would carefully review their customers' intended application for the bots "when possible" and pledged to explore features that could somehow mitigate risks. Stating the obvious, the companies wrote that weaponization of advanced robotics "raises new risks of harm and serious ethical issues," and could harm public trust in the technology. The robot makers went on to encourage policymakers to explore ways to promote the safe use of robots and encouraged other researchers and developers to join the pledge. "We are convinced that the benefits for humanity of these technologies strongly outweigh the risk of misuse, and we are excited about a bright future in which humans and robots work side by side to tackle some of the world's challenges," the companies wrote.

The days of Spot being leveraged as a weapons platform and training alongside special forces operators are already coming to an end; Atlas as a back-flipping soldier of fortune will never come to pass. Their maker, Boston Dynamics, along with five other industry leaders announced on Thursday that they will not pursue, or allow, the weaponization of their robots, according to a non-binding, open letter they all signed. Agility Robotics, ANYbotics, Clearpath Robotics, Open Robotics and Unitree Robotics all joined Boston Dynamics in the agreement. "We believe that adding weapons to robots that are remotely or autonomously operated, widely available to the public, and capable of navigating to previously inaccessible locations where people live and work, raises new risks of harm and serious ethical issues," the group wrote. "Weaponized applications of these newly-capable robots will also harm public trust in the technology in ways that damage the tremendous benefits they will bring to society."

Were you unable to attend Transform 2022? Check out all of the summit sessions in our on-demand library now! Since Boston Dynamics made its bright yellow robot, Spot, available to the general public to purchase in June 2020, the famously agile robo-dog has gone from cute-yet-creepy quadruped to must-have-helper on the factory floor and in industrial facilities. Now, recently-released updates to Spot make gaining data insights a top priority, while Spot's chief engineer, Zack Jackowski, says researchers are making strides in their long-term efforts to incorporate artificial intelligence (AI) and machine learning (ML) into Spot. The biggest overall difference in Spot's latest software update, he told VentureBeat, is that the Boston Dynamics team packaged the robot's applications into a more out-of-the-box experience.

Machine learning and artificial intelligence are among the most innovative forms of the tech that the industry is currently working on, but in spite of the fact that this is the case they also create some problems such as in the case of deepfakes. Deepfakes are AI generated videos that can initially seem indistinguishable from the real thing. Their existence is concerning because of the fact that this is the sort of thing that could potentially end up making it difficult to parse misinformation and fake news. With all of that having been said and now out of the way, it is important to note that AI company Metaphysic has revealed some cracks in the perfection of deepfakes. Some deepfake tech is capable of generating the face of a celebrity on top of your own during a video call, but something as simple as turning sideways might be enough to break the illusion.