Welcome to the March 17, 2021 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."
|
|
ML Models for Diagnosing Covid-19 Not Yet Suitable for Clinical Use
University of Cambridge (U.K.) March 15, 2021
A review of studies containing descriptions of machine learning (ML) models for diagnosing Covid-19 by researchers at the U.K.'s University of Cambridge concluded that none are yet suitable for detecting or diagnosing the virus from standard medical imaging. The team ultimately reviewed 62 studies, and invalidated each model's suitability due to biases in study design, methodological flaws, lack of reproducibility, and publicly available "Frankenstein datasets." Many ML models were trained on sample datasets that were too small to be effective, failed to specify their data's origins, were trained and tested on the same data, or lacked involvement from radiologists and clinicians. Cambridge's James Rudd said, "These early studies show promise, but they suffer from a high prevalence of deficiencies in methodology and reporting, with none of the literature we reviewed reaching the threshold of robustness and reproducibility essential to support use in clinical practice."
|
Ford Partners with U-M on Robotics Research, Building
AP News Corey Williams March 16, 2017
The University of Michigan (U-M) and car manufacturer Ford Motor have partnered to develop robots and roboticists to improve lives, human safety, and society. Tuesday's grand opening of U-M's Ford Robotics Building highlights this collaboration, as robots that fly, walk, roll, and enhance the human body are being designed at the facility. U-M's Alec Gallimore said researchers are collaborating on robots for people, to create human-robotic synergy. Said Gallimore, “Robots aren’t people and people aren’t robots, but we think — together — there can be synergy. So, we’re designing robots that are going to help you. First responders for example. Can we put robots in harm’s way so we don’t have to have people there?”
|
California Passes Regulation Banning 'Dark Patterns' Under Landmark Privacy Law
Gizmodo Brianna Provenzano March 15, 2021
New rules enacted under California's Consumer Privacy Act (CCPA) will bar so-called dark patterns, or underhanded practices used by websites or applications to get users to behave atypically. Examples include website visitors suddenly being redirected to a subscription page, even when they have no interest in the product being marketed. According to an infographic from the California Attorney General's office, dark-pattern strategies rely on "confusing language or unnecessary steps such as forced clicking or scrolling through multiple screens or listening to why you shouldn't opt out of their data sale." The new CCPA regulations will further add a Privacy Options icon, which Internet users can use as a visual cue to opt out of the sale of their personal data.
|
FSU Researchers Enhance Quantum ML Algorithms
Florida State University News Bill Wellock March 16, 2021
Florida State University (FSU) researchers have developed a method to automatically infer parameters utilized in a critical quantum Boltzmann machine algorithm for machine learning applications. A restricted Boltzmann machine algorithm employs probability to learn based on inputs fed to the network, and FSU's William Oates and Guanglei Xu invented a technique to automatically calculate a parameter associated with effective temperature used in that algorithm. Oates said, "That parameter in the model replicates what the quantum annealer is doing. If you can accurately estimate it, you can train your neural network more effectively and use it for predicting things."
|
Virtual Reality at Your Fingertips
ETH Zurich (Switzerland) Leo Herrmann March 16, 2021
Researchers at ETH Zurich in Switzerland have developed a dual-sensor wristband that facilitates intuitive free-hand interaction within virtual productivity spaces. The prototype TapID technology incorporates two acceleration sensors in a rubber wristband, which detect when the hand touches a surface and which finger the user has employed. This design senses tiny differences in the vibration profile on the wrist and differentiates between each unique finger movement, while a custom machine learning pipeline processes the data in real time. TapID generates extremely precise input when used with cameras embedded within virtual reality (VR) glasses, which capture hand positions. The researchers designed a virtual keyboard and piano to demonstrate TapID's capabilities, and ETH Zurich's Christian Holz said the portable technology "has the potential to make VR systems suitable for productivity work on the go."
|
Researchers Find Better Way to Measure Consciousness
University of Wisconsin-Madison News Chris Barncard March 16, 2021
Analysis of neural signals in monkeys by University of Wisconsin–Madison (UWM) researchers combined traditional telltales of consciousness with computational metrics describing the signals' complexities and interaction in different brain regions. The authors used machine learning to determine whether the monkeys were conscious or not and the activity levels of their brain areas by processing those signals through a computer. UWM's Mohsen Afrasiabi said the results indicated the back of the brain and the deep brain areas are more predictive of states of consciousness than the front. UWM's Yuri Saalmann said, "We could use what we've learned to optimize electrical patterns through precise brain stimulation and help people who are, say, in a coma maintain a continuous level of consciousness."
|
Simulation of Self-Driving Fleets Brings Their Deployment in Cities Closer
Imperial College London (U.K.) Hayley Dunning March 15, 2021
Researchers at the U.K.'s Imperial College London (ICL) have modeled the impact of self-driving fleets based on real-world data, to suggest optimal approaches for urban deployment. The team analyzed tens of thousands of potential deployment scenarios, and a range of service parameters and fleet management algorithms, as part of a project called SHIFT. The algorithms can help optimize such a fleet and ensure that only enough autonomous vehicles (AVs) are running in the right areas to meet demand, while minimizing energy consumption. The report provides driver safety guidelines, an AV build manual, and a data infrastructure framework to help operators scale up AV demonstrations to service deployments.
|
Drive-Throughs That Predict Your Order? Restaurants Are Thinking Fast
The New York Times Julie Creswell March 13, 2021
Many restaurants expect digital ordering and drive-throughs to remain key business channels, and some are testing artificial intelligence (AI) to predict and suggest personalized orders. McDonald's acquired Israeli AI company Dynamic Yield to boost sales via personalized digital promotions. Burger King is modernizing its drive-through with its Deep Flame AI system to suggest foods based on daily popularity, and testing Bluetooth technology to identify loyal customers and show their previous orders to calculate their probability of ordering the same items. Restaurant Brands International (RBI) hopes to deploy predictive personalized systems at more than 10,000 of its North American restaurants by mid-2022. RBI's Duncan Fulton envisions customers having "the ability to automatically reorder things and pay for the items at the board, which, ultimately, speeds up the window time, allowing you to collect your food and go on your way."
*May Require Paid Registration
|
Fire Safety App Simulates Wildfires, Shows Route to Avoid Them
New Scientist Edd Gent March 11, 2021
Researchers at the CYENS Center of Excellence in Cyprus have built a mobile wildfire simulation application that provides personalized evacuation routes to anyone in the path of a fire. The app connects to a Web server running the simulation program, which uses publicly available data to update predictions of the spread of fires every 15 minutes. A fire management tool allows local fire departments to quickly tag when and where a fire starts, which is applied to real-time simulations. The app then employs the global positioning system (GPS) location of each user to map out potential escape routes, selecting optimal routes by comparing how fast each route gets them to safety against how close it brings them to the fire's path. The algorithm displays the best option either as turn-by-turn directions or as a route overlaid on a regional map.
|
Researchers Blur Faces That Launched a Thousand Algorithms
Wired Will Knight March 15, 2021
Privacy concerns prompted the researchers who manage ImageNet to blur every human face within the dataset to determine whether doing so would affect the performance of object-recognition algorithms trained on the dataset. ImageNet features 1.5 million images with about 1,000 labels, but only 243,198 images were blurred. The researchers blurred the faces using Amazon's AI service Rekognition and found it did not impact the performance of several object-recognition algorithms trained on ImageNet. Princeton University's Olga Russakovsky said, "We hope this proof-of-concept paves the way for more privacy-aware visual data collection practices in the field." However, Massachusetts Institute of Technology's Aleksander Madry said training an AI model on a dataset with blurred faces could have unintended consequences; said Madry, "Biases in data can be very subtle, while having significant consequences."
|
IBM Develops AI to Invent Antibiotics—and It's Made Two Already
New Atlas Michael Irving March 11, 2021
IBM Research is using artificial intelligence (AI) to develop new antibiotics more quickly, and has already produced two promising drug candidates. Potential molecules are comprised of countless possible chemical combinations, which is why drug development generally takes years. To speed up the process, the researchers used a deep generative autoencoder model to examine a range of peptide sequences, collecting data about their function and the molecules within them and searching for similarities to other peptides. The researchers then used a Controlled Latent attribute Space Sampling (CLaSS) system to generate new peptide molecules with specific properties based on the data gathered by the model. The AI system identified, synthesized, and experimented with 20 new antibiotic peptide candidates over 48 days, producing two that were effective against a range of Gram-positive and Gram-negative bacteria.
|
Coming 'Vaccine Passports' Aim for Simplicity
The Wall Street Journal Katie Deighton March 16, 2021
Developers of the first digital "vaccine passports" for post-pandemic travel said the applications are designed for ease of use, and engineered to eventually mate with other travel platforms. The Clear trusted-traveler program is evaluating a Covid-19 test or vaccination-verification app on certain flights into Hawaii. Meanwhile, organizations like the nonprofit Commons Project Foundation and the International Air Transport Association (IATA) are unveiling apps that allow cross-border travelers to demonstrate they have had a COVID-19 test or a vaccination. Health pass apps aim to navigate users through data entry, while minimizing the information they must input, partly through functions like passport-chip scanning. IATA's Alan Murray Hayden thinks gradual adoption of such apps will provide a kind of herd immunity for lines at the airport.
*May Require Paid Registration
|
Cybersecurity Report: 'Smart Farms' Are Hackable Farms
IEEE Spectrum Payal Dhar March 15, 2021
Researchers at China's Nanjing Agricultural University (NAU) surveyed smart farming and its underlying technologies and utilities, and discovered unique cybersecurity issues stemming from agricultural Internet of Things (IoT) applications. Possible threats to IoT integrity include facility damage, sensor failures in poultry and livestock breeding, and control system intrusions in greenhouses. NAU's Xing Yang said the most pressing vulnerability in smart agriculture concerns the physical environment, like plant factory control system intrusion and unmanned aerial vehicle false positioning; for example, rural areas are prone to poor network signals, which Yang said leads to false base station signals. Yang and his colleagues suggested the use of countermeasures in response, including artificial intelligence to detect malicious users, and the application of existing industrial security standards to design a targeted security framework for agricultural IoT.
|
|