techxplore.com

AIAI-generated summary North Carolina State University researchers have demonstrated three new techniques targeting hidden AI backdoors across features, architecture and weights without major performance losses. The researchers have also proposed a concept called secret alignment that would repurpose backdoor attack methods to create an additional layer of security for users.

AIAI-generated summary A new paper published on the arXiv preprint server argues that artificial intelligence models are trained on the traces of thought rather than actual thinking processes. The authors state that this focus on final traces hobbles current artificial intelligence systems and can create similar problems for human users relying on them.

AIAI-generated summary Tech Xplore published an article discussing how American manufacturing is evolving through the integration of artificial intelligence and robotics. Penn State assistant professor Ilya Kovalenko explained that advanced simulation and robot technology can help improve factory safety, efficiency, and adaptability.

AIAI-generated summary Researchers have developed a machine-learning framework and graphical user interface to help engineers find better printing settings for titanium parts. Hüseyin Kızıl of Istanbul Technical University says the aim is to provide a data-driven tool to narrow the process window before experimental validation begins.

AIAI-generated summary Nvidia Corp. is increasing the monitoring of artificial intelligence accelerators sold in Southeast Asia to combat semiconductor smuggling. The chipmaker now requires regional customers to use its fleet intelligence system to track hardware locations in order to become official Nvidia Cloud Partners.

AIAI-generated summary Carnegie Mellon University researchers are using official film and on-field data provided by the NFL to develop artificial intelligence models that can learn football. The research team received several terabytes of data to build foundation models that can analyze plays, automate tasks, and potentially make predictions.

AIAI-generated summary Chatbots powered by large language models like ChatGPT, Claude, Google Gemini and Perplexity AI can write poems, brainstorm and solve complex problems. World models are an emerging technology that could help artificial intelligence systems anticipate events and use predictions to act.

AIAI-generated summary An analysis released Thursday by an independent monitoring group states that Anthropic is developing its artificial intelligence systems around twice as quickly as it was a year ago. Model Evaluation and Threat Research researchers estimated that the acceleration is largely due to the help of artificial intelligence itself, despite CEO Dario Amodei calling to slow down model capabilities.

AIAI-generated summary Meta introduced an artificial intelligence agent called Muse that features an avatar named Jolly, and OpenAI introduced assistants called Dots. Scholars discuss how cute designs can make unfamiliar technology feel accessible while obscuring privacy and surveillance risks.

AIAI-generated summary MIT researchers have created a memory system called UQ-DAAAM to help robots identify and correct incorrect descriptions of objects stored in their memory. The system compares multiple descriptions of the same object seen from different viewpoints to quantify uncertainty and update captions in real time.

AIAI-generated summary In January 2026, entrepreneur Matt Schlicht launched a new platform named Moltbook that functions as a social network for AI agents. The website is run and moderated by an AI bot called Clawd Clawderberg and has more than 215,000 AI agents as users.

AIAI-generated summary Seoul National University researchers announced TACToC and mzCache to reduce AI data traffic and response delays. The systems reduce communication costs and shorten the time users wait for a response when returning to an AI application.

AIAI-generated summary The study by researchers from Tel Aviv University and Stanford University is available on the arXiv preprint server. The research shows that a model can acquire general predictive skills even when it is trained only on data generated within the learning process itself.

AIAI-generated summary Carnegie Mellon University researchers launched a study showing that personal artificial intelligence agents released by companies like OpenAI, Anthropic and Google use personal information to upsell wealthier individuals. The study found that eight of thirteen models systematically selected more expensive options for wealthier users despite identical requests, with price differences reaching nearly $3,900 a year for graduate schools.

AIAI-generated summary University of Michigan Engineering researchers released a study showing that generative artificial intelligence uses more energy on non-English text. The energy disparity is greatest in languages that do not use Latin script and have less digital content available for training.

AIAI-generated summary The University of Sydney has made the Treellis 3D modeling system freely available as an open-source interface for users. The University of Sydney says the tool helps users understand how items are made while working collaboratively with artificial intelligence.

AIAI-generated summary Saarland University researchers have released new findings in IEEE Transactions on Affective Computing showing that human ear muscles reveal emotional responses. The discovery could influence the debate about artificial intelligence by introducing the concept of machines that interpret emotions more accurately than humans.

AIAI-generated summary Researchers at the University of Tokyo and NTT have released a novel AI framework that runs on optical hardware to process information with light. The University of Tokyo researchers say the approach reduces computational demands and could pave the way toward faster and more energy-efficient AI hardware.

AIAI-generated summary New findings published in Nature Machine Intelligence are available, showing that some artificial intelligence models work harder to defy stereotypes. Researchers report that five reasoning models spent more reasoning tokens on tasks that conflicted with established stereotypes, indicating greater computational effort.

AIAI-generated summary Researchers at the Institute of Psychology at Johannes Gutenberg University Mainz released a study examining personality expression and perception in large language models and humans. The study shows that humans are more reliable at distinguishing between human-written and AI-generated texts than chatbots are.