Welcome to the May 10, 2023, edition of ACM TechNews, providing timely information for IT professionals three times a week.
ACM TechNews mobile apps are available for Android phones and tablets (click here) and for iPhones (click here) and iPads (click here).
To view "Headlines At A Glance," hit the link labeled "Click here to view this online" found at the top of the page in the html version.
The online version now has a button at the top labeled "Show Headlines."
|
|
Mass Event Will Let Hackers Test Limits of AI Technology
Associated Press Matt O'Brien May 10, 2023
Major artificial intelligence (AI) providers are working with the White House to offer thousands of hackers the opportunity to "jailbreak" their AI language models and uncover vulnerabilities. Rumman Chowdhury, who is coordinating a mass hacking event for this summer's DEF CON hacker convention, explained, "We need a lot of people with a wide range of lived experiences, subject matter expertise, and backgrounds hacking at these models and trying to find problems that can then go be fixed." Chowdhury described hackathons like the White House-associated exercise as "a direct pipeline to give feedback to companies," with participants compiling reports and detailing common flaws and patterns.
|
Snake-Like EELS Slithers into Robotics Terrain
NASA Jet Propulsion Laboratory May 8, 2023
A self-propelled snake-like robot developed by researchers at NASA's Jet Propulsion Laboratory (JPL) can map and navigate previously inaccessible terrain autonomously. The EELS (Exobiology Extant Life Surveyor) robot can choose a safe path through terrains ranging from steep craters to underground lava tubes, and can adapt to uncertainty. The robot weighs 220 pounds, is 13 feet long, and is comprised of 10 rotating segments featuring screw threads for propulsion, traction, and grip. EELS employs four pairs of stereo cameras and LiDAR to produce three-dimensional maps of its surroundings, which navigation algorithms analyze to identify the robot’s safest route. JPL's Matthew Robinson said, "It has the capability to go to locations where other robots can't go. Though some robots are better at one particular type of terrain or other, the idea for EELS is the ability to do it all."
|
Drone Can Hitch a Ride on Moving Car
IEEE Spectrum Michelle Hampson May 2, 2023
The Hitchhiker quadcopter drone developed by Sensen Liu and colleagues at China's Shanghai Jiao Tong University can land on moving inclined surfaces, like the side of a moving car. The drone uses a trajectory planning algorithm that factors in each of its four rotors' individual thrust, applying a two-stage tracking approach that analyzes position and attitude. A wheel-like array of ventral suction cups helps Hitchhiker latch onto surfaces, boosting the odds they will contact the desired landing surface and offset any trajectory planning errors. The researchers added an adjustable surface to a car that could be positioned at different inclines, and found Hitchhiker can reliably perch-land on the surface at least 70% of the time at speeds up to 1.07 meters per second (about 2.4 miles per hour) and inclinations up to 90 degrees.
|
Training Machines to Learn More Like Humans Do
MIT News Adam Zewe May 9, 2023
Scientists at the Massachusetts Institute of Technology (MIT) and Toyota subsidiary Woven Planet found computer vision models can be trained to produce more stable, predictable visual representations, similar to those humans learn through perceptual straightening. The researchers taught the models millions of examples via adversarial training, which enhanced their perceptual straightness while reducing their reactivity to slight errors within images. They discovered the models trained on more perceptually straight representations could correctly classify objects in videos with greater consistency. MIT's Vasha DuTell said, "One of the take-home messages here is that taking inspiration from biological systems, such as human vision, can both give you insight about why certain things work the way that they do and also inspire ideas to improve neural networks."
|
Mechanical Backpack Boosts Sensation of Jumping in VR
New Scientist Alex Wilkins May 5, 2023
JumpMod, a mechanical backpack developed by researchers at the University of Chicago, slides a weight up or down to generate the sensation of jumping or falling when worn during virtual reality (VR). JumpMod can sense within milliseconds the need to create the sensation and moves a 2-kg. weight to trick the user's brain into thinking they are jumping or falling.. University of Chicago's Pedro Lopes said, "You can play lots of tricks just by playing with perception, rather than physically having these massive infrastructures," referring to the bulky hardware traditionally used to simulate forces on the body in VR, such as with a roller coaster or race car.
*May Require Paid Registration
|
Quantum Computer in Reverse Gear
Universität Innsbruck (Austria) May 4, 2023
Martin Lanthaler, Ben Niehoff, and Wolfgang Lechner at Austria's University of Innsbruck and quantum spin-off ParityQC have created a template for a quantum computer that can solve the factorization problem using an inverted version of algorithms. A classical logic circuit, which multiplies two numbers, is the starting point; entering two integers as the input values causes the circuit to return their product. Lanthaler explained that although the circuit is constructed for irreversible operations, "The logic of the circuit can be encoded within ground states of a quantum system. Thus, both multiplication and factorization can be understood as ground-state problems and solved using quantum optimization methods."
|
Data Compression Scheme Facilitates Measurement of Blood Flow to the Brain
SPIE May 8, 2023
A data compression scheme developed by researchers at the U.K.'s University of Edinburgh allows most calculations involving single-photon avalanche diode (SPAD) data to be performed directly on a field-programmable gate array (FPGA). This paves the way for improvements in multispeckle diffuse correlation spectroscopy (DCS) techniques to better measure blood flow to the brain. The researchers connected a 192-by-128-pixel SPAD sensor array, packaged into the Quanticam camera module, to an FPGA and applied an autocorrelation algorithm. This allowed the computational burden of calculating autocorrelations to be shifted from the host computing system to the hardware linked directly to the SPAD sensors. University of Edinburgh's Robert K. Henderson said, "Our proposed system achieved a significant gain in the signal-to-noise ratio, which is 110 times higher than that possible on a single-speckle DSC implementation and three times higher than other state-of-the-art multispeckle DSC systems."
|
U.S. Says It Dismantled Russia's 'Most Sophisticated' Malware Network
The New York Times Charlie Savage May 9, 2023
The U.S. Department of Justice said the U.S. and its allies have dismantled a major cyberespionage operation that Russian intelligence had long used to surveil computers worldwide. The Cybersecurity and Infrastructure Security Agency described the "Snake" malware network as "the most sophisticated cyberespionage tool" used by Russia's Federal Security Service. Its purported activities included stealing international relations documents and other diplomatic communications from a NATO country, and infiltrating computers across more than 50 nations and within various American institutions. Cybersecurity agent Taylor Forry explained in a newly unsealed court filing how the Federal Bureau of Investigation used a U.S.-based malware-infected computer to penetrate and "permanently disable" the Snake network by overriding the code on all of its compromised computers.
*May Require Paid Registration
|
Lithography-Free Photonic Chip Offers Speed, Accuracy for AI
Penn Engineering Today Devorah Fischler May 1, 2023
Researchers at the University of Pennsylvania School of Engineering and Applied Science (SEAS) have constructed a lithography-free photonic chip that offers programmable on-chip information processing, yielding photonics-level speed enhanced with superior accuracy and flexibility for artificial intelligence (AI) applications. The chip uses lasers to beam light onto a semiconductor wafer without defined lithographic pathways. Explained SEAS' Liang Feng, "Our chip overcomes [reprogrammability, damage, and cost] obstacles and offers improved accuracy and ultimate reconfigurability given the elimination of all kinds of constraints from predefined features." SEAS' Zihe Gao said the device's active light control can be used "to reroute optical signals and program optical information processing on-chip."
|
Algorithm Improves Understanding of Particle Beams in Accelerators
SLAC National Accelerator Laboratory David Krause May 1, 2023
Scientists at the U.S. Department of Energy's Stanford Linear Accelerator Center (SLAC) National Accelerator Laboratory, Argonne National Laboratory, and the University of Chicago have created an algorithm to more precisely forecast particle beams' positions and velocities in an accelerator. The researchers formulated a machine learning model that taps beam-dynamics knowledge to predict the beam's phase space distribution. They applied the algorithm to experimental data from the Argonne Wakefield Accelerator, and reconstructed fine beam details from only 10 data points by incorporating the physics of beam dynamics. SLAC's Auralee Edelen said, "We've shown that we can infer very complicated high-dimensional beam shapes from astonishingly small amounts of data."
|
Data Class-Specific Image Encryption Using Optical Diffraction
UCLA Samueli Newsroom May 3, 2023
University of California, Los Angeles researchers have developed diffractive deep neural networks that can perform class-specific all-optical image encryption at both near-infrared and terahertz wavelengths using no external computing power aside from the illumination light. After training the networks using deep learning, the researchers used three-dimensional printing to physically fabricate the networks, transform the input images, and produce encrypted, uninterpretable output patterns. The encrypted images can be restored only by applying the correct decryption keys. The transformations performed by the diffractive encryption network are pre-determined and specifically and exclusively assigned to a single data class, which makes it difficult to use reverse-engineering to decipher the original images belonging to the target data classes. Additionally, different decryption keys can be distributed to multiple end-users based on their data access permission, allowing only the appropriate portion of the input data to be shared.
|
Quantum LiDAR Prototype Acquires Real-Time 3D Images While Fully Submerged
Optica May 4, 2023
Researchers from the U.K., France, and Spain have engineered a prototype quantum LiDAR (light detection and ranging) system that can obtain real-time three-dimensional (3D) images underwater. The system illuminates scenes of interest with a green pulsed laser, while an array of single-photon detectors catches the reflected illumination to enable ultrafast low-light detection and to shorten measurement time in photon-sparse settings. Said Aurora Maccarone at the U.K.'s Heriot-Watt University, "Our approach also allows us to distinguish the photons reflected by the target from those reflected by particles in the water, making it particularly suitable to performing 3D imaging in highly turbid waters where optical scattering can ruin image contrast and resolution." The system successfully captured images under controlled highly scattering conditions at distances of 3 meters (9.8 feet) while fully submerged.
|
Model Aims to Help First Responders Reach Accident Sites Faster
NC State University News Matt Shipman May 8, 2023
A model developed by North Carolina State University (NC State) researchers outperformed existing methods of quickly deploying first responders to vehicle accident scenes. The model maximizes the coverage area while minimizing the amount of time it takes responders to reach accident sites. It considers where first responders should be based to respond to the most likely accident sites based on historical data, whether traffic conditions make it more efficient for those based farther away to respond to a particular scene, and how accident severity impacts response times. Said NC State's Leila Hajibabai, "The model can be used for both long-term planning and for allocating incident response resources on a day-to-day basis."
|
|