TechXplore AI
AI news source. Articles are auto-selected and adapted by Hamidun News editors.
Latest publications
Wits Physicists Build a "Light Brain" That Reads Hidden Patterns in Light
Researchers at the University of the Witwatersrand built an optical system that recognizes hidden information in complex light patterns. The goal is faster communication links.
AI and Climate: Data Centers Use 415 TWh a Year, While the Benefits Remain Mostly Forecasts
Data centers use about 1.5% of the world's electricity, and the IEA expects nearly 945 TWh by 2030. A look at what is known about AI's harm and benefit for the climate.
Researchers Propose Teaching Engineers Ethics Through an AI Framework for Industry 4.0
A paper in the International Journal of Continuing Engineering Education and Life-Long Learning proposes using AI to teach engineers ethical decision-making, not just to automate tasks.
UK to keep early access to new US AI models for testing
The UK's AI minister confirmed London will keep testing new American models before release, after reports that the White House had restricted British access.
University of Michigan and NVIDIA to Build an Energy-Saving Method Into Megatron-LM, Cutting Up to 50% of Power Use
Software from the University of Michigan reportedly halves the energy used to train AI without slowing it down. It is being integrated into NVIDIA's Megatron-LM.
An Overlooked Sci-Fi Film Shows How AI Could Benefit Society
TechXplore argues that the AI-apocalypse news cycle is crowding out measured assessments from ethicists, while a film offers a positive scenario for change.
New Report: A Global Pause on Frontier AI Training Can Be Hardwired Into Infrastructure
A new report argues that an international moratorium on training frontier models is technically feasible if governments control chips, data centers, and energy.
Google Faces Protests in India Over $15 Billion AI Data Center
Protesters in India are taking to the streets against a Google data center with placards: the company logo ringed by handcuffs, and a slogan that says "you can't drink data." The project was valued at $15 billion.
Google Faces Local Protests Over 1 GW Data Center in Visakhapatnam
Residents of Visakhapatnam, India, are pushing back on Google's gigawatt campus, part of a $15 billion plan. Their main complaints: electricity, water and jobs.
Can AI Spot a Network "Black Hole" Before It Fails? A Look at Silent Packet Loss
A router looks healthy but silently drops some packets. Why these failures go unnoticed, and what AI could contribute to finding them.
"Human-Written" Book Labels Gain Traction After AI Scandals
Initiatives that label books as human-written are gaining momentum after scandals in which authors were suspected of secretly using AI. There is still almost no way to verify such claims.
Koç University Researcher Proposes Two Optical Systems for AI Computing
Assistant Professor Uğur Teğin has developed two systems that use light, not electronics, to run AI computations. The goal is faster models with a smaller energy appetite.
CISPA Researcher Tests Whether AI Image Watermarks Survive Training of New Models
Michel Meintz of CISPA's SprintML Lab studied whether watermarks on AI images persist after a new generative model is trained on them. The answer depends on the watermark scheme.
NTU Singapore Researchers Cut Decision-Software Development From Months to an Hour
NTU Singapore researchers built an AI framework that automates complex decision-making software, cutting months of engineering work to about one hour.
AI Candidate Scoring Looks Impartial, but That Doesn't Make It Fair
Companies are drowning in applications: popular openings draw up to 1000 a weekend. Algorithms sort the flood, but "no skew in the numbers" doesn't mean "fair selection."
LLMs Taught to See Every Letter Find Reversing Words Easier
Researchers modified language models so they see text letter by letter instead of as token chunks. Tasks like spelling words backwards became easier to solve.
Researchers Study How Chatbots Solve Problems Without Thinking in Symbols Like Humans Do
New neural network research explains why chatbots give coherent advice on math and writing, even though they don't manipulate symbols internally the way people do.
Indosat, Nvidia and Nokia Are Building the Zankore AI Platform in Central Java
Indosat Ooredoo Hutchison, Ooredoo, Nokia and Nvidia are preparing the Zankore platform. The key question for Central Java: how much water will an AI factory draw, and who will pay for it?
Japan raises cyber alert: attacks are more frequent and more sophisticated
Japan's government on Friday urged businesses and state bodies to step up vigilance: cyberattacks are growing in number and technical sophistication. The statement gives no figures.
FT: OpenAI's Revenue Falls About $20 Billion Short of What Investors Expected, and AI Stocks Wobble
The Financial Times reports that OpenAI's annualized revenue is roughly $20 billion below what investors were led to expect. The report hit AI stocks and raised questions about the industry's staying power.
A Simple Formula Predicts When Chatbots Will Go Off the Rails
Researchers propose a mathematical formula that forecasts when small on-device chatbots will start giving dangerous answers, including when offline.
AI Models Refuse to Judge Faces, But Will Draw a “Gay Man” or a “Criminal” on Request
Chatbots won't say whether a person looks gay, but they readily edit a photo to look “gay,” “straight,” or “criminal,” relying on appearance stereotypes.
Aalto University Researchers Build AI Reading Model for AR and Text Personalization
Aalto University researchers built the most accurate AI model of human reading to date, using reinforcement learning, aimed at content personalization and AR interfaces.
Zuckerberg's Manifesto: Meta Releases Open-Source AI Model Against AI Monopoly
On August 11, 2026, Meta released a new open AI model for developers, while Mark Zuckerberg published a manifesto on the threat of AI monopolization. *Meta has been designated an extremist organization and banned in Rus
New Method Scales AI Models Through Expert Blocks Without Retraining
A new method expands already-trained AI models by adding specialized expert modules, with no need to restart training from scratch.
New AI System Aims to Predict How a Fire Will Behave in a Modern Home
Researchers are combining AI, data science and applied math to predict fire dynamics in modern homes, with the goal of improving safety in emergencies.
KIST's Neuromorphic AI Trains Six Times Faster Than Google's Method, With No Extra Hardware
Korean researchers developed A²SG, a method for training spiking neural networks that needs six times less compute than Google's leading approach.
Researchers Build a Real-Time Deepfake Detection System for Video Conferencing
Deepfakes can now synthesize voice and face in real time so convincingly that, without a dedicated warning system, telling a real participant from a fake one is nearly impossible.
Northeastern University builds agentic AI that programs without a developer
Researchers in Boston have developed a framework in which an AI agent writes, tests and fixes code on its own, with no programmer involved at each step.
Researchers Build AI Framework for Long-Term Monitoring of Bridge Damage
Researchers have developed an AI framework for long-term bridge monitoring. It tracks cracks, concrete spalling and water leaks, recording not just that damage exists but how it changes over time.