Welcome to the January 11, 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."
|
|
AI Turns Its Artistry to Creating Human Proteins
The New York Times Cade Metz January 9, 2023
Scientists are adapting techniques underlying art-producing artificial intelligence technology like OpenAI's DALL-E to model new proteins for fighting disease and other applications. The University of Washington (UW)'s Nate Bennett described a protein-structuring approach that, similar to DALL-E, "does what you tell it to do. From a single prompt, it can generate an endless number of designs." Such systems allow researchers to provide rough blueprints for desired proteins, whose three-dimensional shapes are produced by a diffusion model. These protein candidates then go to a wet lab to see if they function as expected. Said UW's Jue Wang, "What's exciting isn't just that they are creative and explore unexpected possibilities, but that they are creative while satisfying certain design objectives or constraints."
*May Require Paid Registration
|
Michael Fire Pioneers Method to Track Anomalous Users
Ben-Gurion University of the Negev (Israel) January 9, 2023
A team of researchers led by Michael Fire at Israel's Ben-Gurion University of the Negev has developed a technique to track groups of anomalous online users. The Co-Membership-based Generic Anomalous Communities Detection Algorithm (CMMAC) is not limited to a single type of network, which means "it can potentially work on different types of social media platforms," according to Fire. He said the researchers tested CMMAC on randomly generated and real-world networks, and found the technique outperformed many other methods in various settings. Fire said CMMAC often outclassed other algorithms on simulation and real-world data, and "successfully detected groups of anomalous users' communities who presented peculiar online activity."
|
Tech at CES Shows How Farmers Can Save Time, Money, Environment
France24 (France) January 8, 2023
Farmers can better protect the environment while reducing costs and saving time through data-collecting technologies unveiled at the 2023 Consumer Electronics Show (CES 2023). French startup Brad has created a sensor that supplies real-time temperature, moisture, and light data in fields to help growers make more accurate and timely decisions on irrigation, pesticide use, and soil treatment. Meanwhile, South Korean startup AimbeLab offers a system for monitoring the volume and condition of grain and feed stored in silos. U.S. company Simple Labs' sensor can read the temperature, humidity, pH value, and phenolic content of wine in a barrel or vat to facilitate more precise control over aging. French company Meropy's rolling wheel device is equipped with cameras to inspect crops for weeds, pests, or disease.
|
The Dawn of Solid-State Quantum Networks
SciTechDaily January 3, 2023
An international team of researchers has demonstrated high-visibility quantum interference between two independent quantum dots (QDs) connected by approximately 300-kilometer (km)/186.4-mile optical fibers. The arrangement generates single photons from resonantly driven single QDs paired with microcavities. Explained Chao-Yang Lu at the University of Science and Technology of China, “Our work jumped from the previous QD-based quantum experiments at a scale from [about] 1 km to 300 km, two orders of magnitude larger, and thus opens an exciting prospect of solid-state quantum networks." Lu added, "Feasible improvements can further extend the distance to about 600 km."
|
Google, DeepMind Launch MedPaLM Language Model
Interesting Engineering Loukia Papadopoulos January 4, 2023
Alphabet subsidiaries Google and DeepMind have launched the MedPaLM large language model (LLM), designed to yield safe, useful answers to questions in the medical field. The model merges HealthSearchQA, a free-response dataset of medical questions, with six open-question answering datasets encompassing professional medical exams, research, and consumer inquiries. MedPaLM can address multiple-choice questions and simple queries from medical professionals and non-professionals. The model was developed on the 540-billion-parameter PaLM LLM and its instruction-oriented Flan-PaLM variant to assess LLMs using MultiMedQA. A team of healthcare professionals found 92.6% of MedPaLM's responses were accurate, compared with 92.9% of clinician-generated responses.
|
Optical Fiber Keeps Data Safe Even When Twisted or Bent
University of Bath (U.K.) January 10, 2023
A new optical fiber developed by researchers at the U.K.'s University of Bath uses the mathematics of topology to protect light even as it is twisted or bent. By applying the principles of topology, the researchers were able to ensure a robust network that can overcome distortions in the fiber. Their fiber consists of several light-guiding cores connected in a spiral, which allows light to hop between the cores but stay within the edge due to the topological design. University of Bath's Anton Souslov said, "Using our fiber, light is less influenced by environmental disorder than it would be in an equivalent system lacking topological design. By adopting optical fibers with topological design, researchers will have the tools to preempt and forestall signal-degrading effects by building inherently robust photonic systems."
|
Security Researchers Say They Hacked California's Digital License Plates
Gizmodo Lucas Ropek January 9, 2023
Security researcher Sam Curry and colleagues identified a vulnerability in the app and website of Reviver, a company that sells digital license plates in California, Arizona, and Michigan. Taking advantage of the vulnerability, Curry gained "full super administrative access" to "all user accounts and for all Reviver connected vehicles." With such access, the researchers could track registered users' GPS locations, manipulate RPlate data, and report vehicles as stolen. Said Curry, "An actual attacker could remotely update, track, or delete anyone's Reviver plate. We could additionally access any dealer (e.g. Mercedes-Benz dealerships will often package Reviver plates) and update the default image used by the dealer when the newly purchased vehicle still had dealer tags."
|
China Turns Its Focus to Deepfakes
The Wall Street Journal Karen Hao January 8, 2023
The Cyberspace Administration of China Tuesday began enforcement of its "deep synthesis" technology regulations in an effort to prevent the production of "deepfakes." The regulations ban the use of artificial intelligence-generated content for disseminating "fake news" or information disruptive to the economy or national security. Additionally, providers of such technology must use prominent labels to indicate the images, video, and text it generates are synthetically generated or edited. Stanford University's Graham Webster said, "China is learning with the world as to the potential impacts of these things, but it's moving forward with mandatory rules and enforcement more quickly. People around the world should observe what happens."
*May Require Paid Registration
|
Project Aims to Expand Language Technologies
Carnegie Mellon University School of Computer Science Aaron Aupperle January 10, 2023
Researchers leading a project at the Carnegie Mellon University School of Computer Science's Language Technologies Institute (LTI) aim to vastly expand the volume of available languages for which automatic speech recognition is available. The researchers are developing a speech recognition model that relies on how languages share physical descriptions of words' sounds (phones), rather than on the sounds differentiating words from each other (phonemes). The method couples the model with a diagram charting relationships between languages to assist with pronunciation rules, enabling the team to approximate the speech model of thousands of languages without audio data. LTI's Xinjian Li said the elimination of audio data input "helps us move from 100 or 200 languages to 2,000."
|
Could Robots Do the Work of Master Marble Sculptors? This One is '99%' There, According to Its Creator
CBS News Chris Livesay January 3, 2023
The 1L robot developed by Robotor is being used to create sculptures using Italy's Carrara marble. The 13-foot zinc alloy robot can craft marble statues that would take humans months to finish in just days. Artists are collaborating with Robotor's Giacomo Massari to turn ideas into three-dimensional images that are then sculpted by 1L with precision. Massari said, "It saves a lot of waste. If something is wrong, or you don't like it, you can just go back...The cool thing about this technology is that we allow the artists to think without any limits." Humans are still required to put the finishing touches on 1L's sculptures. Massari said 1L has achieved 99% perfection, "but it's still the human touch [that] makes the difference. That 1% is so important."
|
Quantum Money Using Mathematics of Knots Could Be Unforgeable
New Scientist Karmela Padavic-Callaghan January 7, 2023
Scientists at computing and cryptography startup NTT Research, the University of Texas at Austin (UT Austin), and the Linux Foundation tapped knot theory to propose forging-immune quantum money. The researchers analyzed a quantum monetary system in which calculating invariants — two equivalent knots assigned identical value — for knots and similar classification problems forms the basis for authenticating money. Each unit of currency in the system features a collection of quantum bits (qubits), each with a corresponding knot and a list of invariants present. The check involves analyzing whether the qubits and their invariants match, but determining another list of matching knots is effectively impossible, which renders currency unforgeable. UT Austin's Scott Aaronson said a ledger of who owns what would be unnecessary with quantum money, as everyone could confirm currency themselves by running a calculation on a quantum computer.
*May Require Paid Registration
|
College Student Created App That Can Tell Whether AI Wrote Essay
NPR Emma Bowman January 9, 2023
Princeton University computer science major Edward Tian developed an app that can identify whether text has been written by a human or OpenAI's ChatGPT. The app comes amid concerns that the viral chatbot could be used by students to pass off assignments written by artificial intelligence (AI) as their own. The GPTZero app uses "perplexity," which measures the text's complexity, and "burstiness," which compares sentence variation, as indicators to determine whether a bot wrote the text in question. Text that is AI-generated is likely to have low complexity and more uniform sentences. Tian, who is working to improve the app's accuracy, said GPTZero is "not meant to be a tool to stop these technologies from being used. But with any new technologies, we need to be able to adopt it responsibly and we need to have safeguards."
|
|