Tech & AI

Researchers analyzing two decades of bridge management research identified artificial intelligence, digital twins, and machine learning as the field's fastest-growing priorities. The findings matter because aging infrastructure worldwide demands smarter maintenance strategies—and companies and agencies investing in these technologies now could gain competitive advantages in a sector facing mounting safety and cost pressures.EN

Researchers have developed a Bayesian approach that uses AI-generated priors to reconstruct scrambled signals from significantly fewer noisy observations than existing methods require. The breakthrough could accelerate medical imaging analysis, particularly in cryo-electron microscopy, by reducing data collection costs and processing time—a major efficiency gain for labs and biotech companies relying on these techniques.EN

Researchers have built a text-to-speech system that begins speaking from the first word with near-imperceptible delay—a breakthrough for real-time applications like live translation and voice assistants. The technology achieves this speed without requiring massive datasets, matching quality of larger systems trained on far more data.EN

A new study reveals that standard bias benchmarks for speech AI systems don't reliably forecast how those same models behave in practical applications. The finding threatens the validity of current fairness testing methods and suggests companies and regulators may be relying on flawed metrics to assess AI safety.EN

Researchers have developed mathematical tools to predict worst-case scenarios in AI decision-making systems—the rare failures that standard performance metrics miss. The work matters because autonomous systems in healthcare, finance, and autonomous vehicles need guarantees about tail risks, not just average performance.EN

Researchers have developed a technique that lets vision-language AI systems adapt to new environments without requiring explicit domain information — a major hurdle in real-world deployment. The approach automatically discovers hidden patterns in training data, enabling more robust and flexible models that could reduce costly retraining cycles across industries.EN

Researchers developed a new framework that lets drones automatically decide what to observe and communicate in real-time, reducing both control and sensing expenses. The approach could reshape how autonomous systems manage limited battery and bandwidth — critical economics for commercial drone operators and 6G infrastructure planners.EN

Wearable energy trackers significantly misalign with how people with disabilities actually experience their bodies, according to new research. The disconnect matters for tech companies designing wellness products—most devices ignore fluctuations from illness, travel, and menstrual cycles, limiting their utility and forcing users to abandon or distrust the tools.EN

Showing users how algorithms pick ads for news articles makes them less likely to click—even when they trust the match more. The finding, based on 498 users, suggests companies face a tradeoff: transparency can undermine engagement, forcing businesses to choose between trust-building and conversion.EN

Researchers have cracked a fundamental problem in how particles arrange themselves when trapped in bounded areas—work with implications for materials science, quantum computing, and optimization algorithms. The breakthrough connects abstract mathematical theory to practical physical systems, potentially improving how engineers design constraints for particle-based technologies.EN

Researchers have solved a long-standing technical problem: how to let autonomous systems use multiple communication styles simultaneously without creating conflicts or interference. The advancement could accelerate deployment of multi-agent AI in manufacturing, logistics, and autonomous vehicle fleets where different robots need to coordinate using different protocols.EN

Researchers have created BUD-MI, a data collection framework that makes it easier for companies and cities to track how much material goes into buildings—crucial information for circular economy planning. The tool combines standardized templates with automated analysis, letting construction firms and policymakers understand material stocks without expensive, time-consuming surveys.EN

Researchers have demonstrated a novel two-layer heat storage system that captures and releases energy at temperatures up to 600°C with minimal losses, potentially solving a critical bottleneck in renewable energy adoption. The breakthrough could enable power grids to store excess solar and wind output for later use, reducing the need for expensive backup fossil fuel plants.EN

Scientists have released ARpest, free software that automates complex analysis of electronic materials without requiring coding skills. The tool could accelerate discovery in semiconductors, quantum materials, and energy tech by letting researchers focus on experiments rather than data processing.EN

Researchers have developed a machine learning method that significantly reduces the amount of labeled data needed to group similar items—a fundamental task in recommendation systems, customer segmentation, and search engines. The breakthrough addresses a critical bottleneck: training AI models when no initial comparison data exists, potentially cutting implementation costs for companies deploying clustering systems.EN

Researchers have developed a method to make PageRank—the algorithm that powers how search engines and content platforms score importance—fairer to underrepresented sources. As AI transparency becomes a regulatory priority, the work offers a practical tool for companies managing large networks to ensure their ranking systems don't systematically disadvantage certain groups or domains.EN

Svenska hushåll kan sänka elnätsbelastningen betydligt genom flexibel elförbrukning — men framför allt om styrsystemen förstår både väder och personliga önskemål samtidigt. Forskargruppen vid Umeå universitet utvecklade en hemenergihanteringssystem (HEMS) som hanterar osäkerhet i solcellsproduktion medan den balanserar fyra konkurrerande mål: lägre kostnad, mindre CO₂, större energioberoende och mindre topplastvariationer. I ett pilotprojekt med ett svenskt enfamiljshus reducerade systemet elkostnaderna med 9,5 procent, koldioxidutsläppen med 9,3 procent och topplasten med 52 procent jämfört med regelbaserade alternativ — allt utan att offra boendekomfort. Den stokastiska planeringen minskade felmarginalen med 33,3 procent gentemot deterministiska metoder. För leverantörer av energistyrning och nätoperatörer är resultatet relevant: occupant-centrisk optimering under svenska tariffer möjliggör skalbar flexibilitet på stadsnivå utan att tvinga bort användare genom rigid automation.

A new analysis of over 400 documented conflicts shows North American and European financial institutions are bankrolling energy projects that displace Indigenous coastal communities worldwide. The finding exposes a supply-chain accountability gap for investors and regulators seeking to manage reputational and legal risks in emerging markets.EN

A new book argues that well-intentioned technology initiatives in developing countries often reinforce colonial power dynamics and economic dependence rather than solving local problems. The analysis challenges how corporations and nonprofits deploy digital solutions, with implications for corporate social responsibility strategies and international development policy.EN

Researchers discovered that water plays a critical role in making CO2 reduction more efficient at low temperatures using defect-rich titanium dioxide. The finding could accelerate commercial development of carbon capture technologies and chemical production processes that industries increasingly need to meet climate targets.EN

Researchers have mapped the complete mathematical landscape of simplified neural networks, uncovering that training dynamics follow predictable equations with no dead-end solutions. The finding could help engineers design more efficient AI systems and better understand how neural networks learn from data at a fundamental level.EN

Researchers discovered that speakers of Chinese and English show measurable cognitive differences in attention, mental rotation, and executive control—advantages tied directly to how their languages are structured. The findings suggest language choice may influence workforce capabilities in fields from aviation to software development, raising questions about hiring, training, and team composition in global enterprises.EN

Researchers have created automated software that predicts how well drug candidates bind to disease targets—a crucial step that typically takes months and costs millions. By testing 4 million+ molecular configurations at once, the new system could slash development timelines and reduce failures in clinical trials, potentially accelerating therapies to market.EN

Researchers developed a machine learning system that accurately forecasts individual travel behavior shifts in response to transit subsidies and policy changes—using far less data than previous methods. The breakthrough could help cities and transit agencies design more effective incentive programs and predict ridership patterns before implementing costly fare reforms.EN

Researchers in Scania, Sweden designed place-based connectivity infrastructure aimed at reducing energy consumption and environmental impact of wireless networks. The approach could reshape how cities and telecom operators balance coverage expansion with climate commitments, offering a template for decentralizing network architecture.EN