Most contemporary artificial intelligence (AI) systems learn to complete tasks via machine learning and deep learning. Machine learning is a computational approach that allows models to uncover patterns in data that are useful for making predictions. Deep learning, on the other hand, is a subset of machine learning that entails the use of multi-layered neural networks, which can autonomously extract features and learn complex patterns from unstructured data, sometimes with little or no human supervision.
Many AI systems trained with these approaches also produce confidence scores for their predictions. These scores are essentially estimates of how probable it is for a specific prediction to be accurate. Past studies suggest that in many cases, AI systems are overconfident and assign high confidence scores to wrong answers, or even present inaccurate information as a fact. This limits their reliability, particularly in high-stakes applications where wrong predictions can have serious consequences.
Researchers at the Korea Advanced Institute of Science and Technology recently introduced a new brain-inspired training approach that could yield more realistic AI confidence estimates. Their proposed strategy, introduced in a paper published in Nature Machine Intelligence, entails briefly training artificial neural networks on random noise (i.e., data with no meaningful patterns) and arbitrary outputs, so that they can learn to produce more realistic confidence estimates before learning specific tasks.
Guide dogs are powerful allies, leading the visually impaired safely to their destinations, but they can’t talk with their owners—until now. Using large language models, a team of researchers at Binghamton University, State University of New York has created a talking robot guide dog system that determines an ideal route and safely guides users to their destination, offering real-time feedback along the way.
The paper, “From Woofs to Words: Towards Intelligent Robotic Guide Dogs with Verbal Communication,” was presented at the 40th Annual AAAI Conference on Artificial Intelligence (AAAI 2026), held January 20–27 in Singapore. It is also available on the arXiv preprint server.
“For this work, we’re demonstrating an aspect of the robotic guide dog that is more advanced than biological guide dogs,” said Shiqi Zhang, an associate professor at the Thomas J. Watson College of Engineering and Applied Science’s School of Computing. “Real dogs can understand around 20 commands at best. But for robotic guide dogs, you can just put GPT-4 with voice commands. Then it has very strong language capabilities.”
GET MY FREE GUIDE: 📘 The Content Creator’s AI Blueprint: From 25 Hours to 5 Minutes https://FirstMovers.ai/blueprint/
This quantum duality discovery shows a material acting as both conductor and insulator… confirmed in a real lab.
A 35 Tesla experiment revealed quantum oscillations inside an insulator’s core. This “conductor-insulator duality” is being compared to wave-particle duality… raising deeper questions about how reality behaves.
Inside this breakdown:
• University of Michigan quantum physics finding
• Conductor-insulator duality explained
• Wave-particle and observer effect links
• Faith and science parallels from Scripture.
If quantum duality keeps expanding… what does it suggest about how reality actually works?
A thermoelectric generator with a shape that no human designer would likely have imagined has now been created by a computer—and it performs more than eight times better than conventional designs. Rather than relying on intuition or repeated trial and error, the breakthrough was achieved through advanced computational optimization.
A joint research team led by Professor Jae Sung Son of the Department of Chemical Engineering at POSTECH (Pohang University of Science and Technology), in collaboration with Professor Hayoung Chung of the Department of Mechanical Engineering at UNIST (Ulsan National Institute of Science and Technology), has developed a general design framework that enables computers to autonomously identify the optimal structure of thermoelectric generators, which convert waste heat into electricity.
Their work is published online in Nature Communications.
Despite affecting half of the world’s population, menopause has historically been understudied and misunderstood, both in biomedical research and clinical practice. However, with the increase in life expectancy, the number of women in the postmenopausal stage continues to grow and, in 2021, those over 50 already represented 26% of the world’s population, according to the WHO.
Its effects go far beyond the reproductive system and are associated with an increased risk of cardiovascular, metabolic, neurodegenerative, and bone diseases. Nevertheless, few studies analyzed in depth how this process affects the female reproductive system as a whole, beyond the ovaries.
In this context, a new study by the Barcelona Supercomputing Center—Centro Nacional de Supercomputación (BSC-CNS), published in Nature Aging, presents the first large-scale atlas of female reproductive system aging, providing a new vision of how this process impacts health.
To reliably complete various manual tasks, robots should be able to handle a variety of objects, ranging from items found in households to tools used in specific professional settings. While many existing robotic systems can now complete basic manual tasks, such as picking up objects and carrying them to a set location, most systems still struggle with tasks that entail the dexterous manipulation of objects.
The term dexterous manipulation describes the ability to skillfully and precisely move objects in nuanced ways, which is central to the completion of many of the tasks that humans tackle daily. Replicating this ability in robots can be very difficult, as it typically requires gathering and interpreting different types of sensory information.
Conventional approaches for robot manipulation rely on visual sensors, such as cameras, and tactile sensors, devices that pick up tactile information. Yet most existing tactile sensors only provide feedback after a robot touches an object, which makes it difficult to plan manipulation strategies in advance.
Anyone who has ever used a microscope knows that it takes time to bring a sample into sharp focus. Each time you move the slide, the image blurs, and you have to stop and carefully turn a knob to bring everything back into clear view. For scientists and clinicians, even if the motion is semi-automated, that time quickly adds up as they work with dozens or hundreds of samples.
Now a team of scientists at Caltech has developed an inexpensive, robust fix for this problem that involves little more than a couple of LED lights and some physics-based processing. They describe the new autofocus technique, which they call Digital Defocus Aberration Interference (DAbI), in a paper published in Nature Communications.
The lead authors of the paper are graduate students Haowen Zhou, Ph.D., and Shi “Josh” Zhao, who completed the work in the lab of Changhuei Yang, the Thomas G. Myers Professor of Electrical Engineering, Bioengineering, and Medical Engineering at Caltech and a Heritage Medical Research Institute Investigator.
From wearable technology to industrial heat recovery, thermoelectric generators which convert waste heat into electricity have an enormous range of potential applications. So far, however, designing high-performing versions of these devices has remained a painstaking task.
Now, through new research published in Nature, Airan Li and colleagues at the National Institute for Materials Science in Japan have developed an AI-based tool that predicts device performance with greater than 99% accuracy, all while cutting computational time by around 10,000-fold.