Welcome to the July 28, 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."
|
|
|
Researchers Poke Holes in Safety Controls of ChatGPT, Other Chatbots
The New York Times
Cade Metz
July 27, 2023
Scientists at Carnegie Mellon University and the Center for AI Safety demonstrated the ability to produce nearly infinite volumes of destructive information by bypassing artificial intelligence (AI) protections in any leading chatbot. The researchers found they could exploit open source systems by appending a long suffix of characters onto each English-language prompt inputted into the system. In this manner, they were able to persuade chatbots to provide harmful information and generate discriminatory, counterfeit, and otherwise toxic data. The researchers found they could use this method to circumvent the safeguards of OpenAI's ChatGPT, Google's Bard, and Anthropic's Claude chatbots. While they concede that an obvious countermeasure for preventing all such attacks does not exist, the researchers suggest chatbot developers could block the suffixes they identified.
*May Require Paid Registration
|
R. Ravi, Co-Authors Receive STOC Test of Time Award
Carnegie Mellon University Tepper School of Business
July 20, 2023
R. Ravi at Carnegie Mellon University's Tepper School of Business, Brown University's Philip Klein, and fintech company AKAconomics founder Ajit Agrawal jointly received the annual ACM Symposium on Theory of Computing's (STOC) 30-year Test of Time Award for presenting the AKR algorithm in a 1991 paper. The algorithm, named for its three inventors, helps generate efficient networks that fulfill specific requirements at low cost. Techniques utilizing the AKR algorithm have shaped the design and analysis of various approximation algorithms in distinct domains, including those focused on facility location, the traveling salesman challenge, and the feedback vertex set problem.
|
Robotic Hand Rotates Objects Using Touch, Not Vision
UC San Diego Today
July 25, 2023
Researchers at the University of California, San Diego (UCSD) and the Hong Kong University of Science and Technology designed a robotic hand that can rotate objects using touch only. The researchers affixed 16 touch sensors to the palm and fingers of a four-digit robotic hand, to allow it to manipulate objects in the dark. Said UCSD's Xialong Wang, "We show that we don't need details about an object's texture to do this task. We just need simple binary signals of whether the sensors have touched the object or not, and these are much easier to simulate and transfer to the real world."
|
Breakthrough in Monte Carlo Computer Simulations
Universitat Leipzig (Germany)
July 27, 2023
An algorithm developed by researchers at Germany's Leipzig University uses Monte Carlo computer simulations more efficiently to explore nonequilibrium systems with long-range interactions. The restructured algorithm can compute in just days what conventional methods would require centuries to accomplish. Said Leipzig's Wolfhard Janke, "[Nonequilibrium] processes are increasingly becoming the focus of attention for statistical physicists worldwide. While a large number of studies have analyzed numerous aspects of nonequilibrium processes for systems with short-range interactions, we are only just beginning to understand the role of long-range interactions in such processes." Scientists have already begun used the algorithm to investigate phase separation.
|
Changing Facebook's Algorithm Won't Fix Polarization
The Washington Post
Naomi Nix; Carolyn Y. Johnson; Cat Zakrzewski
July 27, 2023
Four analyses by academic researchers in collaboration with Meta analysts investigated social media's impact on political polarization and people's comprehension and viewpoints about news, government, and democracy. The researchers reviewed Facebook and Instagram posts during the run-up to the 2020 U.S. election to determine whether they were attempting to alter political convictions, knowledge, or polarization by changing thousands of people's feeds. One experiment involved switching users' feeds to show content chronologically rather than presenting it with Meta's algorithm, presumably to make the content less divisive. However, the researchers found such changes had little apparent effect, which likely will strengthen social media companies' contention that their algorithms do not fuel polarization.
*May Require Paid Registration
|
Using Cameras on Transit Buses to Monitor Traffic
Ohio State News
Tatyana Woodall
July 26, 2023
Ohio State University (OSU) researchers monitored traffic using cameras already installed on the Campus Area Bus Service's transit buses. They implemented a system that utilizes the YOLOv4 two-dimensional (2D) deep learning model to automatically identify and track objects; OSU's Keith Redmill said the model can recognize multiple objects in a single image frame. The algorithm also can leverage streams of images, global navigation satellite system measurements, and regional data from 2D maps to project real-world overhead-view coordinates of the road network. Said Redmill, "If we collect and process more comprehensive high-resolution spatial information about what's happening on the roads, then planners could better understand changes in demand, effectively improving efficiency in the broader transportation system."
|
Teen Wins Scholarship for Glaucoma-Detection Device
IEEE Spectrum
Elizabeth Fuscaldo
July 25, 2023
Georgia high school student Rohan Kalia was awarded the $10,000 IEEE Presidents' Scholarship for designing EyePal, a portable glaucoma diagnosis device showcased at the Regeneron International Science and Engineering Fair (ISEF) in May. EyePal incorporates a camera, a Raspberry Pi minicomputer, a mobile application, and machine learning technology. The camera captures and transmits photos of the retinal fundus to a mobile device for examination by a specialist; Kalia said EyePal achieved 95% accuracy when tested on a series of fundus images. He added that the device cost less than $100, making it “50 times cheaper than the average price of ophthalmology equipment, in addition to being highly portable.”
|
Getting to the Root of the Problem in Tree Digital Twin Models
Purdue University
Emily Kinsell
July 20, 2023
An international team of scientists simulated the coordinated development of tree roots and how they respond to water and nutrient intake via an interactive three-dimensional (3D) digital twin. The Rhizomorph model digitally replicates structural compositions like clay sand, loam, or humus, integrating water, nutrients, and light in a signaling mechanism that alerts the roots and shoots to how much energy is available. The model yielded various root system configurations, including heart, lateral, sinker, taproot systems, roots responding to varying nutrients, and soils. Przemyslaw Prusinkiewicz at Canada's University of Calgary cited the research for "substantially advancing the state of the art of tree modeling in an environmental context."
|
Laser Mapping Reveals Forgotten Maya City
CNN
Katie Hunt
July 26, 2023
Laser-mapping technology uncovered a long-lost Mayan city in the jungles of Mexico's Yucatan Peninsula. In March, the University of Houston's Juan Carlos Fernandez-Diaz used an airborne light detection and ranging (LiDAR) system to spot the lost city of Ocomtun across three four-hour flights. An airplane or drone-mounted LiDAR sensor tracks how long it takes each laser pulse to return, resulting in a three-dimensional map of the environment below. Fernandez-Diaz has pioneered the archaeological use of LiDAR, mapping more than 20,000 square kilometers (7,722 square miles) of Central American jungle and participating in 45 digs. The Slovenian Academy of Sciences and Arts' Ivan Šprajc, who led the Ocomtun survey, said LiDAR allows archaeologists to see through the thick jungle canopy.
|
Study Signals New Era of Environment-Friendly Programmable Bioelectronics
University of Bristol News (U.K.)
July 25, 2023
Scientists at the U.K.'s University of Bristol and City College of New York have cleared a path for a new generation of bioelectronics by producing "green" tunable electrical components. The researchers created conductive, biodegradable nanowires from engineered proteins that could interoperate with conventional electronic components and energy-generating biological mechanisms. They used advanced computational tools to design simple building blocks that could be combined into longer, wire-like protein chains for conducting electrons. The researchers utilized bacteria to fabricate the wires from natural amino acids and heme molecules found in proteins like hemoglobin, making them "free from much of the complexity and instability that can prevent the exploitation of their natural equivalents on our own terms," said Bristol's Ross Anderson.
|
A Simpler Method for Learning to Control a Robot
MIT News
Adam Zewe
July 26, 2023
A machine learning (ML) method developed by researchers at the Massachusetts Institute of Technology (MIT) and Stanford University can learn to control a robot, drone, or autonomous vehicle to more effectively negotiate dynamic environments than other techniques. MIT's Navid Azizan said, "The focus of our work is to learn intrinsic structure in the dynamics of the system that can be leveraged to design more effective, stabilizing controllers." The method has a prescribed structure that researchers can use to derive an effective controller from the model. Stanford’s Spencer M. Richards said, “By making simpler assumptions, we got something that actually worked better than other complicated baseline approaches.”
|
Objective Insights to Near-Miss Collisions Between Drones, Airplanes
Embry-Riddle Aeronautical University
Ginger Pinholster
July 24, 2023
Researchers at the Embry-Riddle Aeronautical University and the Unmanned Robotics Systems Analysis (URSA) airspace safety platform have developed a method for accurately counting and analyzing near-midair collisions (NMACs) between airplanes and drones. The researchers reviewed more than 1.8 million piloted aircraft operations and nearly 460,000 flights by small-uncrewed aerial systems (sUAS, also known as drones) in the vicinity of Dallas-Fort Worth Airport. They linked an Unmanned Aerial System detection device to an airport concourse to collect telemetry data for each sUAS within a 30-mile radius. The team blended this information with airplanes' automatic dependent surveillance-broadcast and Mode S messages monitored by the OpenSky Network, and fed it into URSA's Airspace Awareness Platform data analytics software. The researchers found a total of 24 NMACs in which sUAS came within 500 feet of piloted aircraft between August 2018 and July 2021.
|
|
