Toggle light / dark theme

Asking AI to act like an expert can make it less reliable

To get the best out of AI, some users tell it to provide answers as if it were an expert. Others ask it to adopt a persona, such as a safety monitor, to guide its responses. However, this approach can sometimes hurt performance, according to a study available on the arXiv preprint server.

To see how well large language models (LLMs) behave when they are told to be someone else, researchers from the University of California ran a huge test using 12 different personas across six language models. These included experts in fields like math, coding and STEM (science, technology, engineering and mathematics) as well as general roles such as creative writer or safety monitor.

The team found that adopting a persona was something of a double-edged sword. While it makes AI sound more professional and keeps it safer (more likely to follow rules and less likely to generate harmful content), it sometimes performs worse at recalling facts.

GitHub adds AI-powered bug detection to expand security coverage

GitHub is adopting AI-based scanning for its Code Security tool to expand vulnerability detections beyond the CodeQL static analysis and cover more languages and frameworks.

The developer collaboration platform says that the move is meant to uncover security issues “in areas that are difficult to support with traditional static analysis alone.”

CodeQL will continue to provide deep semantic analysis for supported languages, while AI detections will provide broader coverage for Shell/Bash, Dockerfiles, Terraform, PHP, and other ecosystems.

‘Gray-box’ AI reveals why catalysts work while speeding discovery

Self-driving laboratories (SDLs) powered by artificial intelligence (AI) are rapidly accelerating materials discovery, but can they also explain their results? Researchers from the Theory Department of the Fritz Haber Institute, in collaboration with BASF, and BasCat—UniCat BASF JointLab, show that they can.

Their new AI-driven strategy works hand-in-hand with SDLs to identify better catalysts while revealing the chemistry behind their performance. The approach was validated on the industrially crucial conversion of propane into propylene.

An SDL integrates an AI doing the experiment planning with lab automation and robotics. In the race to develop better materials, AI and SDLs are often celebrated for one main reason: speed.

Fish gill-inspired panels reveal path to efficient thermal mixing

A fascination with fish gills has led researchers at Cornell to develop a bio-inspired approach to mixing heat and molecules in fluids—findings that could inform future biomedical devices, heat exchangers and soft robotics.

Moving heat and mass efficiently through flowing liquids is central to technologies ranging from dialysis machines to industrial cooling systems, yet many of those technologies rely on rigid components to get the job done.

Looking for an alternative, Yicong Fu, a mechanical engineering doctoral student, turned to fish gills—soft, porous tissue that constantly stirs water to keep gases and ions flowing. Working with Sunghwan “Sunny” Jung, professor of biological and environmental engineering in the College of Agriculture and Life Sciences, Fu designed a gill-like thermal dispenser that is providing new insights into fluid-structure interactions.

Novel Octameric Resistosome and Immune Mechanism in Wheat

After activation, these NLRs form multi-protein complexes—called resistosomes—that carry out the immune response. Studies have shown that certain resistosomes are pentameric (e.g., ZAR1 and Sr35), whereas others are hexameric (e.g., NRC2 and NRC4). These complexes initiate immune responses by triggering calcium (Ca2+) influx into the cytoplasm. However, the G10 type of CC-NLR (CCG10-NLR) immune receptors constitutes a unique clade among CC-NLRs and its activation mechanism has remained poorly understood.

Now, in a study published in Cell, a research team has revealed a novel octameric resistosome formed by an activated wheat CCG10-NLR immune receptor, which induces Ca2+ influx and immune responses through a unique channel architecture.

The researchers identified the Wheat Autoimmunity 3 (WAI3) gene, which encodes a CCG10-NLR protein. Subsequent analysis revealed that a gain-of-function (GOF) single amino acid mutation in the leucine-rich repeat (LRR) domain leads to autoactivation, providing an opportunity to study the activation mechanism of CCG10-NLR.

After expressing the WAI3 proteins in Nicotiana benthamiana, the researchers used cryo-electron microscopy to resolve the octameric structure of the activated WAI3 resistosome—marking the first time an octameric resistosome has been identified in plants.

The CCG10-NLR WAI3 resistosome differs from known resistosomes both in the number of monomers and in its conformation, representing a novel assembly mechanism for plant NLR resistosomes.

Using Nicotiana benthamiana and animal cell expression systems, the researchers also demonstrated that the WAI3 resistosome induces Ca2+ influx in plants but is not effective in animal cells. ScienceMission sciencenewshighlights.


NVIDIA CEO Doesn’t Believe in “Succession Plans,” Says He Operates in a Way That Doesn’t Make His Company a One-Man Show

On the Lex Fridman podcast, NVIDIA’s CEO was asked about his mortality and whether he fears dying in his current state. Jensen offered a rather interesting response, saying that his company is currently in the midst of a technological revolution and that, if he died in the meantime, it might not be the best-case scenario for him.

“The most important thing you should do today, if you care about the future of your company, post you, is to pass on knowledge, information, insight, skills, experience as often and continuously as you can. Which is the reason why I continuously reason about everything in front of my team.”


NVIDIA has become the largest business entity and the driving force in the AI world, yet CEO Jensen Huang has no succession plans in sight.

Self-repairing spacecraft could change future missions

Healable spacecraft structures could soon be possible thanks to cutting-edge composite technology. Swiss companies CompPair and CSEM with Belgian company Com&Sens have partnered with the European Space Agency (ESA) to modify their self-healing carbon fiber product for use in space transportation.

Project Cassandra (a loose abbreviation of Composite Autonomous SenSing AnD RepAir) includes sensors and a heating element into a composite carbon-fiber material, allowing spacecraft to autonomously repair initial stages of damage.

Cassandra is part of ESA’s Future Innovation Research in Space Transportation (FIRST!) Initiative which is finding and testing innovative technology that will benefit European space transportation.

/* */