Tech & AI

Scientists have discovered a new way to subtly corrupt AI systems that analyze aerial and satellite images, causing them to misclassify specific objects while leaving others untouched. The vulnerability threatens applications in urban planning, agriculture, and infrastructure monitoring—raising questions about the reliability of automated systems that governments and companies depend on for critical decisions.EN

A study of 24,000 health records in Ethiopia found that while basic demographic data are reliably captured, reports of live births, neonatal deaths, and child health events are inconsistent across tracking systems. For health systems and donors investing in data-driven interventions, the findings suggest that community-level health information may be less trustworthy than assumed.EN

Researchers developed an activity-tracking system that recognizes patient movements from accelerometer data no matter where or how a device is positioned on the body. The breakthrough addresses a real clinical problem: standard monitoring systems fail when devices shift during patient care, potentially missing health deterioration that hospitals need to catch.EN

Researchers identified over 69,000 colistin-resistance genes across bacterial species, revealing that dangerous antibiotic-resistance traits spread far more widely than previously known. The findings signal an urgent threat to treatment options for hard-to-kill infections and could reshape how health systems monitor and combat antimicrobial resistance.EN

Researchers have synthesized a new iron-calcium phosphate ceramic that matches the hardness of tooth enamel and resists acid better than existing materials. The breakthrough could enable stronger dental implants, protective coatings, and biomedical devices—opening a new product category for materials manufacturers and dental companies.EN

Studien undersöker hur generativa AI-modeller och LLM-arkitekturer kan lösa flaskhalsen mellan datahastighet och systemkapacitet i informationssystem. Forskare från Göteborgs universitet, ITT Technical Institute och Lviv Polytechnic analyserar hur språkmodeller kan optimera dataflödet när organisationer står inför ökande volym och växande krav på responsivitet. Resultaten visar att LLM-baserade system kan reducera latens och öka genomströmning genom intelligent datakomprimering och adaptiv routning. Implementeringen möjliggör snabbare beslutfattande utan att infrastrukturkostnaderna eskalerar proportionellt. För tekniska chefer och AI-produktledare är detta relevant vid val av arkitekturer för dataintensiva operationer. Många organisationer kämpar redan med denna flaskhals; LLM-lösningar erbjuder en tidseffektiv väg att adressera den utan omfattande omstrukturering. Slutsatserna tillämpas redan hos case-organisationen.

Researchers have developed a machine learning system that identifies pipeline failures by mapping cause-and-effect relationships in real time. The approach combines the transparency needed for safety-critical infrastructure with the accuracy advantages of AI, potentially reducing costly ruptures and environmental disasters across gas and oil networks.EN

Researchers have identified sustainable coatings made from agricultural waste—lignin, tannins, and plant proteins—that prevent dangerous biofilm growth on marine vessels and medical devices. The finding could disrupt a $2.5 billion antifouling market currently dominated by toxic copper-based paints, though manufacturers still need to solve durability and cost challenges before scaling production.EN

Scientists have developed interactive 3D virtual characters that can engage in natural conversations, combining advances in language models, voice synthesis, and digital avatars. The breakthrough could transform customer service, training, entertainment, and remote work by replacing text-based chatbots with lifelike conversational partners that feel more human and trustworthy.EN

Researchers released QDFlow, an open-source simulator that generates realistic synthetic datasets for quantum dot devices, eliminating the need for expensive and time-consuming experimental data collection. For companies developing quantum hardware, the tool could accelerate AI model development and reduce barriers to scaling quantum device research.EN

A new analysis reveals that language models require significantly larger volumes of aligned content to properly form and retain concepts compared to what search engines need to rank pages. The finding suggests that content strategies optimized for Google may leave AI systems unable to learn meaningful representations—a gap with implications for companies building AI-dependent products and platforms.EN

Researchers have developed an IoT system that automatically detects when machine coolant degrades, eliminating manual operator checks. The sensor network achieves 96% accuracy and enables predictive maintenance schedules, reducing downtime and environmental contamination from spent coolant.EN

Researchers have solved a longstanding puzzle about how mathematical ring systems preserve their fundamental properties when extended or transformed. The breakthrough could accelerate development of cryptography, coding theory, and quantum computing applications that rely on these abstract algebraic structures to function reliably at scale.EN

Researchers have developed VIVALDy, a machine-learning system that reconstructs complex fluid behavior from a handful of sensor readings, cutting computational costs for engineers. The breakthrough could accelerate design of energy-harvesting devices, offshore structures, and industrial systems where predicting fluid forces is critical but traditionally expensive.EN

Researchers have published a reference architecture for digital twins—virtual replicas of physical networks—designed specifically for next-generation 6G systems. The framework could help telecom operators and equipment makers reduce deployment costs, accelerate troubleshooting, and optimize network performance before problems reach customers.EN

A new analysis confirms that how leaders manage—not just what they decide—significantly impacts worker productivity and loyalty. The finding matters as companies wrestle with retention and performance: transformational leaders, who inspire and develop their teams, outperform those using command-and-control or transactional approaches in boosting employee commitment and initiative.EN

German and French research teams have successfully scaled complex engineering software to run on massive supercomputers, a milestone that could accelerate everything from aircraft design to drug discovery. The breakthrough combines new GPU optimization techniques with smarter data processing—clearing a path for companies to run prohibitively expensive simulations faster and cheaper.EN

A new ACM collection examines Alan Turing's lasting influence on artificial intelligence, cryptography, and computer science—fields now worth hundreds of billions annually. Understanding these intellectual roots matters for leaders navigating AI regulation, ethics, and the future of computational innovation.EN

A new study of an AI-powered voice service in Zambia reveals a paradox: the technology lets marginalized women communicate privately for the first time, yet still excludes them through language barriers and platform control. For companies and policymakers betting on AI to bridge digital divides, the finding warns that good intentions don't guarantee equitable outcomes.EN

Two major adaptations that help plants survive in hot, dry climates—annual life cycles and efficient photosynthesis—evolved independently rather than as linked strategies, a new study of 4,000 grass species reveals. The finding reshapes how scientists understand plant adaptation and could inform crop breeding and climate resilience programs.EN

Researchers have cracked a decades-old mathematical problem about solving complex boundary conditions—work that may have immediate applications in machine learning and artificial intelligence systems. The breakthrough enables computers to handle messy, incomplete data more reliably, potentially improving how AI models process real-world information.EN

Researchers have extended a foundational mathematical rule that helps organize and analyze complex datasets, making it faster to compute patterns in labeled hierarchies. The breakthrough simplifies calculations used in logistics, scheduling, and data science — areas where companies increasingly rely on algorithms to optimize operations and extract insights from intricate organizational structures.EN

Researchers have built a working system that uses traffic cameras and AI to track vehicle emissions across a city in near real-time, then forecast pollution hotspots hours ahead. The platform, tested in Stockholm, could help urban planners cut emissions and gives cities a cheaper alternative to expensive sensor networks for meeting climate targets.EN

Researchers have pinpointed why lithium-ion batteries lose capacity over time — and it's not what engineers thought. The weak link isn't the active particles themselves, but the conductive material binding them together. This finding could redirect billions in battery R&D spending toward materials that better withstand the stresses of repeated charging cycles.EN

Researchers have released QHDL, a specialized language that lets engineers write programs treating quantum and classical computers as a unified system rather than separate tools. The development could accelerate practical applications in industries relying on hybrid quantum computing, from finance to pharmaceuticals, by making these systems easier to design and debug.EN