ACM Technews, Wednesday, September 25, 2024

Welcome to the September 25, 2024 edition of ACM TechNews, providing timely information for computer professionals three times a week.

U.N. Experts Urge United Nations to Lay Foundations for Global Governance of AI

A United Nations advisory body comprised of 39 AI leaders from 33 countries is calling on the U.N. to lay the foundation for global regulation of AI and set forth principles, including both international and human rights law, to guide the establishment of new AI governance institutions. Among other things, the advisory group recommends the creation of an international scientific panel on AI to ensure global understanding of the technology's capabilities and risks.
[ » Read full article ]

Biden Proposes Banning Chinese Vehicles, 'Connected Car' Technology from U.S. Roads

A proposed regulation announced Monday by the U.S. Commerce Department would ban the use of Chinese software and hardware in connected vehicles on U.S. roads, in response to national security concerns. The proposed regulation would ban Chinese automakers from testing self-driving vehicles on U.S. roads. The prohibitions could be extended to vehicle software and hardware from Russia and other U.S. adversaries.
[ » Read full article ]

Neuromorphic Wires Amplify Their Own Signals

A research team led by Texas A&M University's R. Stanley Williams developed self-amplifying neuromorphic wires that mimic the axons that carry signals from human nerve cells. To amplify the signals traveling along the wire without the help of amplifiers or other devices, the researchers used a layer of lanthanum cobalt oxide with a 1 mm metal line on top; they biased the lanthanum cobalt oxide with a direct current and passed an alternating current signal through the metal. The researchers observed the signal traveling through the wires was amplified upwards of 70%.
[ » Read full article ]

McDonald's Touchscreen Kiosks Were Feared as Job Killers. Instead, Something Surprising Happened

Despite predictions that self-service kiosks at McDonald's and other fast-food chains would eliminate jobs, Placer.ai analyst RJ Hottovy argues that labor has been reallocated to handle issues with mobile ordering and delivery. Added Hottovy, "If kiosks really improved speed of service, order accuracy, and upsell, they'd be rolled out more extensively across the industry than they are today."
[ » Read full article ]

Microsoft AI Needs So Much Power It's Tapping Site of U.S. Nuclear Meltdown

Constellation Energy Corp. will spend $1.6 billion to revive the Three Mile Island nuclear plant in Pennsylvania, with Microsoft agreeing to purchase all the output energy for 20 years as it looks to access carbon-free electricity for its AI datacenters. Constellation said a reactor that was closed in 2019 will be placed back into service in 2028. The deal is part of a Microsoft initiative to run all of its datacenters on clean energy by 2025.
[ » Read full article *May Require Paid Registration ]

Study Finds Thousands of Browser Extensions Compromise User Data

Georgia Institute of Technology researchers developed the Arcanum web framework to study the more than 100,000 functional browser extensions available in the Google Chrome Web Store to determine whether they extract user data from websites containing sensitive data. The researchers found user-specific data is collected automatically by more than 3,000 browser extensions, more than 200 of which uploaded sensitive user data from webpages to servers.
[ » Read full article ]

AI 'Godfather' Says OpenAI's New Model May Be Able to Deceive, Needs 'Much Stronger Safety Tests'

ACM A.M. Turing Award recipient Yoshua Bengio is concerned about the ability of OpenAI's new o1 model to deceive, noting it has a "far superior ability to reason than its predecessors." Said Bengio, "In general, the ability to deceive is very dangerous, and we should have much stronger safety tests to evaluate that risk and its consequences in o1's case."
[ » Read full article *May Require Paid Registration ]

Telegram Will Now Provide Some User Data to Authorities

The Telegram messaging app has changed its terms of service and privacy policy to state that it will provide users' IP addresses and phone numbers in response to search warrants or other valid legal requests from authorities. Stanford University's Daphne Keller said that policy likely will not satisfy the requirements of French and European law, noting that some countries require authorities to be notified when child sexual abuse material and other illegal content is posted.
[ » Read full article ]

Is Math the Path to Chatbots That Don’t Make Stuff Up?

Silicon Valley startup Harmonic is focusing on mathematics as it works to develop an AI chatbot that never hallucinates. Harmonic's Aristotle not only produces correct answers, but also detailed computer programs proving its answers are right, which then can be used to improve its results. Some researchers believe the same techniques can be used to develop AI systems that can verify physical truths as well.
[ » Read full article *May Require Paid Registration ]

Some U.S. Kaspersky Customers Find Their Security Software Replaced

Following the U.S. government's ban on the sale of Kaspersky products over concerns that Russia could use them to spy on U.S. citizens, users of the firm's antivirus products on Windows systems have had their software replaced automatically by something labelled UltraAV. UltraAV software reportedly has not undergone testing by the Anti-Malware Testing Standards Organization, and most major antivirus testing laboratories have not seen UltraAV's code.
[ » Read full article ]

Meta to EU: Your Tech Rules Threaten to Squelch the AI Boom

In an open letter coordinated by Facebook parent firm Meta Platforms, executives warned the European Union risks missing out on the full benefits of artificial intelligence because of its tech regulations. More than two dozen companies signed the letter, which said AI can boost productivity and expand the economy. The letter called on the EU to harmonize its rules and provide what the signatories refer to as a modern interpretation of the bloc’s data-protection law.
[ » Read full article *May Require Paid Registration ]

A Bottle of Water Per Email: The Hidden Environmental Costs of Using AI Chatbots

The Washington Post worked with researchers at the University of California, Riverside to determine how much water and electricity are used to write the average 100-word email using ChatGPT. They determined such an email requires little more than a single bottle of water but sending it once weekly for a year would require an amount of water equivalent to the consumption of every household in Rhode Island for 1.5 days.
[ » Read full article *May Require Paid Registration ]

Google Chrome Switches to ML-KEM for Post-Quantum Cryptography Defense

Google said it will move from KYBER to ML-KEM (Module-Lattice-based Key-Encapsulation Mechanism) for post-quantum cryptography defense with version 131 of its Chrome browser, slated for release in early November. The company attributed the switch to changes in the final version of ML-KEM, which make it incompatible with KYBER.
[ » Read full article ]