Welcome to the August 4, 2025 edition of ACM TechNews, providing timely information for computer professionals three times a week.

HEADLINES AT A GLANCE

• Chinese Universities Want Students to Use More AI, Not Less
• A Quantum Computer Goes to Space
• CISA Open-Sources Thorium Platform for Malware, Forensic Analysis
• AI Is Fast-Tracking Climate Research, from Weather Forecasts to Sardines
• Scientists Create Light-Powered Microphone
• Google AI Model Maps World in 10-Meter Squares for Machines to Read
• TikTok Asks Users to Help Police Misinformation
• YouTube Channels Are Being Sold, Repurposed to Spread Scams, Disinformation, Says New Research
• Tech Giants Revise AI Product Claims That Faced Scrutiny
• Teaching Robots with Generated Video
• U.S. Airlines Oppose New Limits on Facial Recognition at Airport Checkpoints
• OpenAI Launches Stargate in Europe with Norwegian Deal

Chinese Universities Want Students to Use More AI, Not Less

Almost all faculty and students at Chinese universities use generative AI, according to a survey by Chinese higher-education research group the Mycos Institute. A study of the 46 top Chinese universities' AI strategies by MIT Technology Review found nearly all have added interdisciplinary AI general-education classes, AI-related degree programs, and AI literacy modules. All students at China’s Remin, Nanjing, and Fudan universities can enroll in general-access AI courses and degree programs.
[ » Read full article ]

A Quantum Computer Goes to Space

A quantum computer has successfully been deployed into Earth’s orbit. The compact photonic device, designed by University of Vienna researchers, launched June 23 and is now operational aboard a satellite. The new computer runs on just 10 watts and fits within 3 liters. Researchers say it will soon begin demonstrating its capabilities, which may include efficient onboard data processing and future quantum communication.
[ » Read full article ]

CISA Open-Sources Thorium Platform for Malware, Forensic Analysis

The open source Thorium platform developed by researchers at the U.S. Cybersecurity and Infrastructure Security Agency (CISA) and Sandia National Laboratories is intended for use by government-, public-, and private-sector malware and forensic analysts. Available through CISA's official GitHub repository, Thorium automates numerous cyberattack investigatory tasks. Integrating commercial, open source, and custom tools, Thorium can schedule more than 1,700 jobs per second and handle more than 10 million files per hour per permission group.
[ » Read full article ]

AI Is Fast-Tracking Climate Research, from Weather Forecasts to Sardines

Climate researchers increasingly are turning to AI to automate routine tasks amid funding cuts and other challenges. Researchers at Spain's AZTI marine research center are using AI models to monitor water quality, the presence of different types of marine life, and more to inform decision-making. AI also is being used to produce more accurate weather forecasts and to facilitate citizen science projects.
[ » Read full article *May Require Paid Registration ]

Scientists Create Light-Powered Microphone

A visual microphone developed by researchers at China's Beijing Institute of Technology shines light on an object and measures changes in reflected intensity resulting from sound-induced vibrations. The microphone is equipped with smart algorithms that decode the fluctuations into audible sound. The use of single-pixel imaging combined with Fourier-based localization methods means it is simpler and less expensive than optical microphones that rely on high-speed cameras or lasers.
[ » Read full article ]

Google AI Model Maps World in 10-Meter Squares for Machines to Read

Google's new AlphaEarth Foundations AI model provides a comprehensive view of Earth over time by mapping it in 10-meter squares that can be read by deep learning applications. Trained on Earth observation data from satellites and other sources, AlphaEarth integrates the data into "embeddings" that are easily processed by computer systems. The embeddings have 64 dimensions, each representing a 10-meter pixel that encodes data about territorial conditions for that plot over a year.
[ » Read full article ]

TikTok Asks Users to Help Police Misinformation

TikTok is launching "Footnotes," a crowdsourced fact-checking feature that allows approximately 80,000 qualified U.S. users to add contextual and background information and source links to videos. Contributors can rate these notes for helpfulness, with top-rated footnotes displayed to all users. The program, available to users aged 18+ with six months on the platform and clean community guidelines records, supplements existing content moderation.
[ » Read full article *May Require Paid Registration ]

YouTube Channels Are Being Sold, Repurposed to Spread Scams, Disinformation, Says New Research

Carnegie Mellon University researchers found that 1 of every 400 YouTube channels were sold on third-party platforms and completely repurposed within 30 days. The overhauled channels saw subscriber counts increase during the 12 weeks following each revamp, indicating viewers may not have been aware of the change in handles, titles, and descriptions. The researchers found 37% of these channels later promoted political disinformation, crypto schemes, gambling ads, and other material flagged by YouTube as potentially harmful.
[ » Read full article ]

Tech Giants Revise AI Product Claims That Faced Scrutiny

Apple, Google, Microsoft, and Samsung have revised or retracted AI marketing claims following investigations by BBB National Programs' National Advertising Division (NAD). NAD found several misleading advertisements, including Apple's promotion of unreleased iPhone AI features as "available now," a YouTube video from Google showing sped-up Gemini assistant capabilities, Microsoft's claim that Copilot's Business Chat function works "seamlessly across all your data," and Samsung's claim that its AI-powered refrigerator "automatically recognizes what's in your fridge" when it only identifies 33 specific items if they are clearly visible.
[ » Read full article *May Require Paid Registration ]

Teaching Robots with Generated Video

University of Illinois Urbana-Champaign researchers developed a video generation model to rapidly produce demonstration videos to teach robots to perform specific tasks. With RIGVid (Robots Imitating Generated Videos), users need only provide a text command and an image of the current scene to produce a demonstration video tailored to the robot's specific task and environment, with no additional training required. RIGVid's success rate across tasks was 85%, outperforming other robot imitation methods.
[ » Read full article ]

U.S. Airlines Oppose New Limits on Facial Recognition at Airport Checkpoints

In a letter to the U.S. Senate, trade association Airlines for America called on lawmakers to reject legislation that would restrict the U.S. Transportation Security Administration's ability to use biometrics and facial recognition technology at security checkpoints. The organization expressed concerns that the bill could "increase wait times considerably by slowing down identity verification at every airport security checkpoint," as well as slowing aviation innovation.
[ » Read full article ]

OpenAI Launches Stargate in Europe with Norwegian Deal

A datacenter being built by Nscale Global Holdings Ltd. in Kvandal, Norway, with funding from Norwegian investor Aker ASA, will be the first European site for OpenAI's Stargate datacenter infrastructure project. The site will offer 230 megawatts of capacity initially, with an additional 290 megawatts to be added in the future. By the end of 2026, OpenAI will deliver 100,000 Nvidia GPUs to the datacenter, with more chips to be added afterward.
[ » Read full article *May Require Paid Registration ]