news 2024

Biodiversity researchers tested vision systems on how well they could retrieve relevant nature images. More advanced models performed well on simple queries but struggled with more research-specific prompts. Try taking a picture of each of North America's roughly 11,000 tree species, and you'll have a mere fraction of the millions of photos within nature image datasets.

Artificial intelligence (AI) systems tend to take on human biases and amplify them, causing people who use that AI to become more biased themselves, finds a new study by UCL researchers. Human and AI biases can consequently create a feedback loop, with small initial biases increasing the risk of human error, according to the findings published in Nature Human Behaviour .

This year, Lawrence Berkeley National Laboratory (Berkeley Lab) led research that pushes the boundaries of science and technology. From advancing our understanding of the universe to innovations in microelectronics, artificial intelligence, quantum science, and biomanufacturing, the lab's teams have achieved remarkable milestones across diverse fields.

The energy network in the Netherlands is overcrowded and the demand for new connections cannot be kept up with. So if there is still space to be found, a network company like Alliander wants to know exactly where and when that space is available. Data scientists from Alliander and Radboud University worked together on a system to better map this out.

A new technique identifies and removes the training examples that contribute most to a machine-learning model's failures. Machine-learning models can fail when they try to make predictions for individuals who were underrepresented in the datasets they were trained on. For instance, a model that predicts the best treatment option for someone with a chronic disease may be trained using a dataset that contains mostly male patients.

Using LLMs to convert machine-learning explanations into readable narratives could help users make better decisions about when to trust a model. Machine-learning models can make mistakes and be difficult to use, so scientists have developed explanation methods to help users understand when and how they should trust a model's predictions.

Scientists from Universidad Carlos III de Madrid (UC3M), together with a research team from Ericsson and the KTH Royal Institute of Technology in Sweden, have developed an artificial intelligence-based software programme that can search for information and make recommendations for biomedical image analysis.

Researchers propose a simple fix to an existing technique that could help artists, designers, and engineers create better 3D models. Creating realistic 3D models for applications like virtual reality, filmmaking, and engineering design can be a cumbersome process requiring lots of manual trial and error.

The human mouth holds a wealth of information about overall health. From body temperature to head and jaw movements during sleep, this data can be essential for understanding health conditions and dental issues. However, collecting such data is often cumbersome and uncomfortable. Researchers from TU Delft, in collaboration with Radboudumc, have developed a groundbreaking new technology: Densor, a battery-free sensor platform that can be worn in the mouth via a standard dental brace or a -bite splint.

Scientists from NASA and our Department of Physics have used supercomputer simulations to reveal that Mars' moons may have been formed from destroyed asteroid fragments. The researchers found that an asteroid passing near Mars could have been pulled apart by the planet's gravity, with the resulting rock fragments strewn into a range of orbits.

A new advance in health monitoring which uses radar to 'listen' to patients' heart sounds with remarkable accuracy could lead to a new generation of contactless medical monitoring equipment. Researchers from the University of Glasgow led the development of the new system, which uses radar to track patients' heart sounds like a doctor uses a stethoscope.

By sidestepping the need for costly interventions, a new method could potentially reveal gene regulatory programs, paving the way for targeted treatments. By studying changes in gene expression, researchers learn how cells function at a molecular level, which could help them understand the development of certain diseases.

ISTA researchers investigate why graph isomorphism algorithms seem to be so effective Graphs are everywhere. In discrete mathematics, they are structures that show the connections between points, much like a public transportation network. Mathematicians have long sought to develop algorithms that can compare any two graphs.

An international research team led by QuTech has demonstrated a network connection between quantum processors over metropolitan distances. Their result marks a key advance from early research networks in the lab towards a future quantum internet. The team developed fully independently operating nodes and integrated these with deployed optical internet fibre, enabling a 25 km quantum link.

More accurate orbit predictions for satellites and space debris as well as a better understanding of the water masses present on Earth: Researchers at TU Graz have achieved both using satellite laser ranging. What do the Earth's gravitational field and the trajectories of satellites and space debris have in common? The Earth's gravitational field influences the orbits of our companions in space, while the changes in the orbits in turn allow conclusions to be drawn about changes in the gravitational field and thus existing water masses.

Some sick Black patients are likely labeled as "healthy- in AI datasets due to inequitable medical testing Black patients are less likely than white patients to receive medical tests that doctors use to diagnose severe disease, such as sepsis, researchers at the University of Michigan have shown. Because of the bias, some sick Black patients are assumed to be healthy in data used to train AI, and the resulting models likely underestimate illness in Black patients.

A research project led by the Desertification Research Centre (CIDE, UV-CSIC-GVA) has successfully predicted the ecological interactions that occur in little-analysed plant communities based on coexistence patterns from a well-studied ecosystem in Alicante. The study employed transfer learning, a machine learning technique that introduces a new approach to biodiversity research and conservation.

Scientists from the Autonomous University of Madrid (UAM) and Complutense University of Madrid (UCM) have created an early warning system based on artificial intelligence to predict massive outbreaks of cyanobacteria in fresh waters. Using machine learning and deep learning models, the system could help protect aquatic ecosystems and improve water management.

The EU AI Act is designed to ensure that AI is transparent and trustworthy. For the first time, ETH computer scientists have translated the Act into measurable technical requirements for AI. In doing so, they have shown how well today's AI models already comply with the legal requirements. Researchers from ETH Zurich, the Bulgarian AI research institute INSAIT - created in partnership with ETH and EPFL - and the ETH spin-off LatticeFlow AI have provided the first comprehensive technical interpretation of the EU AI Act for General Purpose AI (GPAI) models.

Predicting the behavior of many interacting quantum particles is a complicated process but is key to harness quantum computing for real-world applications. A collaboration of researchers led by EPFL has developed a method for comparing quantum algorithms and identifying which quantum problems are the hardest to solve.