Tech & AI

Researchers found that redesigning how AI systems show their reasoning work significantly improves users' ability to spot mistakes and verify correctness. A graph-based interface boosted error detection to 85.6%, suggesting that better presentation of AI logic—not just better AI—could unlock safer deployment in education and other high-stakes fields.EN

Researchers found that when AI systems use persuasive language to justify decisions, people are more likely to accept them—but only in healthcare contexts. In finance, the opposite happens: users trust straightforward explanations more. The finding matters because high-stakes industries rely on user acceptance of AI recommendations, yet few have tested whether rhetorical tactics undermine or strengthen genuine trust.EN

A new review shows that optimizing nozzle design and printing parameters in plastic 3D printing can simultaneously improve product durability and reduce material waste. For manufacturers scaling additive manufacturing, these findings offer a roadmap to lower costs while meeting sustainability targets without sacrificing performance.EN

Researchers have developed a faster optimization method that improves how engineers design high-performance antennas used in 5G, radar, and wireless communications. The advance cuts design time and production costs by automating complex geometry calculations that previously required manual trial-and-error.EN

Researchers have created a computer model that predicts how perfluorooctanoic acid (PFOA)—a persistent chemical found in drinking water and food—moves through the human body. The breakthrough could help regulators assess thousands of untested PFAS chemicals faster and cheaper than traditional animal testing, potentially accelerating safety decisions for manufacturers and public health agencies.EN

Scientists have engineered a two-layer magnetic material that produces complex, linked magnetic patterns with potential applications in next-generation data storage and processing. The breakthrough demonstrates how tuning the layers' interaction and external fields can create previously unreachable magnetic configurations, opening pathways for more efficient information devices.EN

Computer scientists have dramatically improved the mathematical bounds linking two key measures of graph complexity, shrinking exponential gaps to linear relationships. The breakthrough enables faster algorithms for counting graph colorings—a problem with applications in network optimization, scheduling, and constraint satisfaction that affects industries from logistics to chip design.EN

A new review reveals that wave and tidal energy systems fail not from design flaws but from interconnected corrosion, wear, and biological fouling that current testing cannot predict. Understanding these coupled failure mechanisms is critical: ocean energy could supply massive renewable capacity, but commercial viability depends on solving these tribological challenges that engineers have largely overlooked.EN

A groundbreaking 2005 thesis on optimizing how mobile phones manage memory—developed by Lund University and Sony Ericsson—has been publicly released for the first time. The work, which inspired a European patent, could help engineers design faster, more efficient devices as computing demands continue to grow.EN

Researchers have developed a system that treats physical exhaustion as a measurable variable that can be tracked and managed moment-by-moment in human-robot teams. The innovation could reduce workplace injuries and boost productivity in manufacturing and logistics by redistributing demanding tasks before workers hit dangerous fatigue thresholds.EN

Researchers analyzing electric bus charging in Sweden found that modest shifts in when buses charge can cut electricity costs significantly—but only if pricing structures reward such flexibility. The finding matters because most transit agencies pursuing electrification haven't optimized for this lever, and tariff rules vary wildly between neighboring grid operators, creating hidden inefficiencies.EN

A new framework maps 30 practical steps to let researchers and companies share medical data across the Atlantic despite clashing privacy laws. The BRIDGE pilot study offers the first consensus blueprint for compliant data exchange, potentially unlocking billions in AI-driven drug discovery and clinical innovation currently blocked by regulatory gridlock.EN

Electrochemiska sensorer för marina miljöer förlorar snabbt sitt signal när marina mikroorganismer täcker upp dem — ett problem som begränsar långtidsövervakning av jonkoncentrationer. Forskare vid Qingdao University of Science and Technology har utvecklat en självreparerande sensor för kaliumjoner som motverkar denna biofouling genom ett peptidhydrogelskikt baserat på Fmoc-FF-dipeptider. Sensorn uppnår en signaldrift på endast 0,2 mV per timme under 24 timmar — betydligt bättre än konventionella elektroder. Arbetet genomfördes i samarbete med KTH och spanska universitet. Fyndet är relevant för alla som värderar marina sensor-IoT-system eller övervakar vattenkvalitet i kustnära områden. Självreparerande material minskar behovet av ofta utbytade sensorelement, vilket sänker driftskostnader och möjliggör äkta långtidsövervakning. För leverantörer av miljösensorik öppnas nya marknader inom akvakultur och kustövervakning.

**Ingen sammanfattning möjlig — data otillräcklig** Källan saknar abstract och innehållsbeskrivning. Utan åtkomst till studiens faktiska resultat, metoder eller slutsatser kan jag inte producera en faktakorrekt sammanfattning enligt dina krav. **Vad behövs för att fortsätta:** - Tillgång till full text eller abstract - Eller en konkret beskrivning av studiens huvudfynd, metod och implikationer Om du kan dela studiens abstract eller nyckelresultat, skriver jag omedelbar en sammanfattning utformad för Teknik & AI-chefer och investerare — med fokus på affärsrelevans och implikationer för systembeslut.

Researchers have released POLEVAL, a Python package that automates the complex analysis of X-ray photoelectron spectroscopy experiments. The tool lets scientists process months of lab data in a single notebook, enabling faster decision-making during expensive beamtime sessions and reducing the technical barriers that have historically slowed materials research.EN

Researchers have developed Gelina, an AI system that generates speech and hand gestures together rather than separately, producing more natural and synchronized video avatars. The breakthrough matters for companies building customer service bots, virtual presenters, and video translation tools—applications that currently suffer from awkward, misaligned gestures that undermine user trust.EN

Researchers have developed a method that helps smaller language models produce better text summaries by simultaneously optimizing multiple quality measures—consistency, coherence, relevance, and fluency. A 7-billion-parameter model trained with this technique now performs on par with GPT-4, potentially offering companies a cheaper, faster alternative for summarization tasks.EN

Researchers have developed a faster, cheaper way to evaluate whether AI-generated speech and music sound good to human ears—using audio compression technology as a quality metric. The method works without expensive listening panels, potentially accelerating development of voice assistants, podcast generators, and other audio AI products.EN

Researchers have developed a faster way for autonomous vehicles to understand their surroundings by focusing computational power on safety-critical objects like pedestrians and other cars, rather than wasting resources on static backgrounds. The technique cuts training time significantly while improving accuracy where it matters most for collision avoidance.EN

Researchers have released Nord-Parl-TTS, a large open-source speech dataset for Finnish and Swedish drawn from parliamentary recordings. The 5,900+ hours of high-quality audio addresses a critical bottleneck: most AI voice technology is trained on English and Mandarin, leaving smaller language markets underserved and companies dependent on proprietary solutions.EN

Researchers have developed a framework that uses artificial intelligence to isolate and enhance individual speakers in real-time, even as they move around. The technology could transform consumer audio devices, hearing aids, and augmented reality platforms by letting users suppress background noise and lock onto chosen speakers without losing natural sound quality.EN

Researchers have developed a technique that assigns reliable probability estimates to AI predictions, addressing a critical gap in machine learning reliability. For companies deploying AI in high-stakes decisions—from autonomous vehicles to medical diagnostics—knowing when a model is actually confident versus guessing could significantly reduce costly errors and liability risks.EN

A new design concept describes massive rotating structures that generate their own energy, water, and air while hosting diverse ecosystems—positioning them as practical alternatives to interstellar exploration. The proposal challenges the space industry's focus on distant travel by arguing that building thousands of self-sustaining worlds within the solar system offers greater economic and scientific value.EN

Researchers have published a technical blueprint for integrating artificial intelligence and blockchain into a unified financial operating system designed for organizations. The framework, which emphasizes education as its foundation, aims to help businesses automate recurring financial processes at scale—a shift that could reshape how companies handle settlements, decision-making, and compliance across distributed networks.EN

Researchers have solved a decades-old puzzle about how random points naturally tessellate into rectangular grids, proving that segment lengths follow predictable statistical patterns. The findings could improve algorithms for network design, materials science, and spatial optimization problems where systems self-organize from random starting conditions.EN