ScienceDaily Robotics
Science news aggregator providing the latest research updates in robotics
AI Identifies Early Risk Patterns for Skin Cancer
Swedish researchers used nationwide registry data from over 6 million adults to train AI models that predict melanoma risk. The most advanced model reached 73% accuracy, far above the 64% baseline of age‑sex only methods, and identified sub‑populations with a 33% five‑year chance of developing melanoma. The findings suggest AI could enable precision‑targeted screening, improving early detection while conserving healthcare resources. Further validation and policy work are needed before routine clinical adoption.
This Simple Change Stops Robot Swarms From Getting Stuck
Researchers at Harvard SEAS discovered that injecting a modest amount of randomness into robot swarm movement dramatically reduces congestion and boosts task completion rates. By combining mathematical models, computer simulations, and real‑world robot experiments, they identified a “Goldilocks zone” of...
AI Breakthrough Cuts Energy Use by 100x While Boosting Accuracy
Artificial intelligence consumed roughly 415 terawatt‑hours of electricity in 2024—over 10% of U.S. power generation—and demand is set to double by 2030. Researchers at the University of Wisconsin‑Madison unveiled a neuro‑symbolic visual‑language‑action (VLA) system that slashes energy use by up to...
Truckloads of Food Are Being Wasted because Computers Won’t Approve Them
Digital approval systems now govern every food shipment, and when they malfunction, trucks loaded with produce sit idle, leading to waste. Recent cyber‑attacks on U.S. grocery networks and the 2021 JBS ransomware incident illustrate how a broken digital manifest can...
Deepfake X-Rays Are so Real Even Doctors Can’t Tell the Difference
A study published in Radiology reveals that both radiologists and leading multimodal large language models struggle to differentiate AI‑generated deepfake X‑rays from authentic scans. When unaware of synthetic images, radiologists detected only 41% of fakes; awareness boosted accuracy to 75%....
AI-Powered Robot Learns How to Harvest Tomatoes More Efficiently
Osaka Metropolitan University researchers unveiled an AI‑driven robot that evaluates the “harvest‑ease” of each tomato before attempting to pick it. The system blends image recognition with statistical analysis to select optimal picking angles, achieving an 81% success rate in field...
Study Finds ChatGPT Gets Science Wrong More Often than You Think
Washington State University researchers tested ChatGPT on 719 business‑journal hypotheses, finding it answered correctly 76.5% in 2024 and 80% in 2025. After adjusting for chance, the model was only about 60% better than random guessing and struggled especially with false...
Scientists Discover AI Can Make Humans More Creative
Swansea University researchers found that AI can act as a creative collaborator, not just an efficiency tool. In a study of over 800 participants designing virtual cars, an AI system using MAP‑Elites generated diverse galleries of designs, including intentionally flawed...
Scientists Built the Hardest AI Test Ever and the Results Are Surprising
Researchers from Texas A&M and nearly 1,000 global experts created Humanity’s Last Exam, a 2,500‑question benchmark designed to be unsolvable by current AI models. The test spans mathematics, humanities, natural sciences, ancient languages and other specialist domains, with any question...
AI Breakthrough Could Replace Rare Earth Magnets in Electric Vehicles
Scientists at the University of New Hampshire used artificial intelligence to compile a searchable database of 67,573 magnetic compounds, uncovering 25 previously unknown high‑temperature magnets. The AI system extracts experimental data from literature, predicts magnetic behavior, and records temperature thresholds....
How Everyday Foam Reveals the Secret Logic of Artificial Intelligence
Engineers at the University of Pennsylvania discovered that foam interiors remain in perpetual motion, contradicting the long‑held view that foams behave like frozen glass. Using simulations, they showed bubbles wander through countless configurations, a behavior mathematically identical to the parameter...
