Toggle light / dark theme

AI vision, reinvented: Vision-language models gain clearer sight through synthetic training data

In the race to develop AI that understands complex images like financial forecasts, medical diagrams and nutrition labels—essential for AI to operate independently in everyday settings—closed-source systems like ChatGPT and Claude currently set the pace. But no one outside their makers knows how those models were trained or what data they used, leaving open-source alternatives scrambling to catch up.

Now, researchers at Penn Engineering and the Allen Institute for AI (Ai2) have developed a new approach to train open-source models: using AI to create scientific figures, charts and tables that teach other AI systems how to interpret complex visual information.

Their tool, CoSyn (short for Code-Guided Synthesis), taps open-source AI models’ coding skills to render text-rich images and generate relevant questions and answers, giving other AI systems the data they need to learn how to “see” and understand scientific figures.

What The Last Century Of Cybersecurity Can Teach Us About What Comes Next In The Age Of AI

Now, with the introduction of AI systems trained on years of real-world data, many of those tasks can be automated at scale—in most cases, with greater speed and consistency than a human working alone. The business impact is immediate and measurable.

To use AI effectively in frontline defense, it must do more than process data. It has to understand how your organization assesses risk and learn to make decisions that protect both security and business continuity. We’re seeing that this is especially valuable for clients with high customer activity, where security teams are flooded with alerts that demand fast, accurate decisions to maintain service levels.

New scrubbing robot could contribute to automation of household chores

While the advent of robotic systems that can complete household chores has been widely anticipated, those commercially released so far are primarily robot vacuums that autonomously clean the floor. In contrast, robots that can reliably clean surfaces, tidy up, cook or perform other tasks in home environments are either too expensive or have not yet reached the market.

Researchers at Northeastern University recently developed SCCRUB, a soft that can complete a chore beyond hoovering and mopping, which many people find tedious, namely scrubbing surfaces clean. The new robotic arm, introduced in a paper on the arXiv preprint server, was found to successfully clean dirty, burnt and greasy surfaces, removing over 99.% of residue adhered to them.

“Our recent study builds on one of our earlier papers published in Science Robotics,” Jeffrey Lipton, senior author of the paper, told Tech Xplore. “We knew we had a new type of robot arm that could deliver the power of a drill through a soft robotic arm. We wanted to show what else we could do with this new platform.”

New approach allows drone swarms to autonomously navigate complex environments at high speed

Unmanned aerial vehicles (UAVs), commonly known as drones, are now widely used worldwide to tackle various real-world tasks, including filming videos for various purposes, monitoring crops or other environments from above, assessing disaster zones, and conducting military operations. Despite their widespread use, most existing drones either need to be fully or partly operated by human agents.

In addition, many drones are unable to navigate cluttered, crowded or unknown environments without colliding with nearby objects. Those that can navigate these environments typically rely on expensive or bulky components, such as advanced sensors, graphics processing units (GPUs) or .

Researchers at Shanghai Jiao Tong University have recently introduced a new insect-inspired approach that could enable teams of multiple drones to autonomously navigate complex environments while moving at high speed. Their proposed approach, introduced in a paper published in Nature Machine Intelligence, relies on both a deep learning algorithm and core physics principles.

Psychopathic traits linked to distinct brain networks in new neuroscience research

Psychopathy is often associated with impulsivity, aggression, and antisocial behavior. While past studies have focused heavily on how different brain regions function in people with psychopathic traits, less is known about how these regions are structurally connected. Structural connectivity refers to the physical links between brain areas—similar to the brain’s wiring system. The researchers aimed to go beyond earlier work that focused only on specific brain circuits and instead look across the entire brain to identify any structural patterns linked to psychopathy.

The researchers were especially interested in understanding whether structural differences in the brain might explain the relationship between psychopathic traits and externalizing behaviors. Previous models have suggested two possible brain-based explanations for these behaviors. One theory emphasizes problems in how people process emotional threats, while another highlights difficulties in attention control. Both theories have some support, but no study had comprehensively examined how structural brain networks might connect psychopathy with real-world behavioral problems.

The research team analyzed data from 82 young adults who participated in the Leipzig Mind-Brain-Body study. All participants were screened to rule out medical or psychological conditions that might affect the results. Psychopathic traits were assessed using a questionnaire designed to capture both interpersonal-affective characteristics (like manipulation and lack of empathy) and behavioral traits (like impulsivity and rule-breaking). Externalizing behaviors were also measured with a separate questionnaire that included items on aggression, defiance, and similar tendencies.

Each participant underwent high-resolution brain imaging using diffusion MRI, a technique that maps the white matter tracts—essentially the brain’s wiring—connecting different regions. The researchers used a method called connectome-based predictive modeling, which relies on machine learning to identify patterns in the brain’s structural connectivity that relate to individual differences in behavior.

This method allowed them to identify two kinds of networks: positive networks, where stronger connections were linked to higher psychopathy scores, and negative networks, where weaker connections were related to those same scores. They also tested whether specific connections within these networks helped explain the relationship between psychopathic traits and externalizing behaviors.

The results showed that psychopathic traits were significantly associated with both stronger and weaker connections in different parts of the brain. The positive network—made up of connections that increased with psychopathy—was better at predicting psychopathic traits than the negative network alone. But when both networks were combined, the prediction became even more accurate.

Many of the connections in the positive network were located within the brain’s frontal and parietal lobes, which are involved in decision-making, emotional processing, and attention. These connections included pathways like the uncinate fasciculus, which links the frontal cortex with areas involved in emotion, and the arcuate fasciculus, which supports language and auditory processing. Other connections involved the cingulum bundle, associated with emotional regulation and social behavior, and the posterior corticostriatal pathway, which plays a role in reward processing and learning.