Science Simplified!

Comment

Comment by Dr. Krishna Kumari Challa 25 seconds ago

Without water, forests lose their ability to absorb CO₂
During drought, trees close their stomata to conserve water, reducing CO₂ uptake and photosynthesis. Severe water stress can damage their water-transport systems and kill trees; continued respiration and decomposition can then turn forests from carbon sinks into carbon sources. Reduced transpiration may also diminish rainfall recycling.
To survive periods of drought, trees halt photosynthesis, a highly water-intensive process. For every molecule of CO₂ absorbed by a tree leaf, an average of around 400 water molecules are released. Entire forests may then end up emitting more CO₂ than they capture.

original article.

Comment by Dr. Krishna Kumari Challa 39 minutes ago

AI chatbots give us a narrow slice of knowledge: Researchers warn of 'knowledge collapse'

"Let me just ask the chatbot" has become an everyday phrase. Where we used to turn to Google to find information, AI chatbots have become a common way to get answers to everything.
The large language models underpinning AI chatbots give us a significantly narrower range of information than a conventional web search. That's the finding of a new study by researchers.
Every language model they tested provides users with more uniform information than a simple Google search across all the topics we looked at. In other words, people are to a large extent exposed to the same information over and over again. So AI chatbots are not just changing how we find knowledge, but also which knowledge we have access to.
Across 155 topics, 27 language models produced less varied information than Google search; even the most diverse model was at least 18.7% less varied. Newer models were slightly more diverse than older ones, so the proposed “knowledge collapse” has not been observed. Training on AI-generated text could further narrow the range of information in future models.
The researchers tested 27 different large language models on 155 topics.

The results show that even the language model producing the most diverse answers—OpenAI's GPT-5—provides at least 18.7% less varied information than Google. The topics tested by the researchers ranged from nuclear weapons, marriage, pornography, racism and genocide to more country-specific topics such as Marine Le Pen, the Falklands War and K-pop.

According to the researchers, the fact that people are increasingly using AI models as their primary gateway to information could have significant consequences.

We risk exposing people to fewer perspectives and a narrower range of knowledge. This could create a vicious cycle in which the most popular content becomes even more dominant, while other content is increasingly overlooked.
It's similar to globalization. Today, you can buy the same products and find the same coffee chains almost everywhere in the world. That has many advantages, but it has also reduced diversity.
According to the researchers, the low level of diversity in language models is partly a result of how the models work. They compress the vast amount of text they are trained on and learn the patterns that occur most frequently. In the process, information that deviates from the most common patterns is filtered out.

The effect could be amplified if language models are increasingly trained on text produced by other AI models—something the researchers expect to happen.
If this process is repeated over several generations of models, the range of information could gradually become narrower. This is what the researchers refer to as "knowledge collapse."
AI chatbot summaries can, of course, be useful—that is why so many people use them. But it is still important to seek out different sources in order to understand the nuances and get a broader picture—especially for the younger generation growing up with AI. We must not become so dependent on the technology that we stop understanding and thinking for ourselves.
AI developers should build language models in a way that preserves the breadth of knowledge available to us. Researchers have developed a method that they can use to measure diversity in models, which could help ensure that future language models do not become less diverse.

Dustin Wright et al, What and Whose Knowledge? Measuring Epistemic Diversity in Large Language Models, arXiv (2025). DOI: 10.48550/arxiv.2510.04226

Comment by Dr. Krishna Kumari Challa 49 minutes ago

Could soil process information without a brain? New engineering perspective makes the case

Researchers are exploring how soil behaves as an intelligent system. By studying its structure, chemistry and biological network of microbes and fungi, they aim to inspire new technologies in bio-inspired computing, precision agriculture and environmental sensing.
Recognizing soil as an active, embodied information processor—constantly sensing, filtering and transforming flows of matter (e.g., rainwater) and energy (e.g., sunlight) into plant growth and nutrients—reframes it not as "just dirt" but as a living system that senses, remembers and adapts. Viewed this way, soil challenges the traditional idea that intelligence requires a brain.

Inspired by electrical impedance tomography (EIT), a real-time imaging and monitoring technology that measures the electrical conductivity and impedance of materials, these researchers set out to use the technique to observe the inner workings of soil.
In this context, soil is particularly interesting as a computational medium because of its inherently rich, nonlinear and dynamic electrical properties. These properties allow soil to respond to external stimuli (such as electrical signals, moisture variation or biological activity) in ways that can encode and process information. This means soil could act not only as a sensing medium but also as a form of natural analog processor, enabling low-energy, distributed computation directly within the environment.

"Such an approach could open up new possibilities for adaptive sensing, environmental monitoring and precision agriculture, where computation is embedded directly in the physical system being observed."

The embodied intelligence of soil

Contemporary "smart" farming technologies, such as advanced sensors, Internet of Things (IoT) systems and robotics, are examples of precision agriculture in action. They enable real-time monitoring of soil moisture, pH and nutrient dynamics, helping farmers make decisions that respond to the soil's specific needs.

In this way, soil can be viewed as the "active intelligence partner in farming," the researchers say, becoming "both co-processor and collaborator, enabling a form of agriculture where technology and ecology are integrated into a single system of care."
Treating soil as an embodied sensor that continuously registers environmental change could help sensing technologies measure variables like moisture and temperature, providing an early warning before surface indicators appear.

Xiaoxian Xu et al, In-Situ Classification of Soil Types Exploiting Electrical Impedance Tomography with a Robotic Actuating Probe, 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (2025). DOI: 10.1109/iros60139.2025.11246829

Catherine Merchant et al, Embodied Intelligence of Soil, IOP Conference Series: Materials Science and Engineering (2026). DOI: 10.1088/1757-899x/1343/1/012026

Comment by Dr. Krishna Kumari Challa 2 hours ago

Where do nanoparticles go once they enter the body? Scientist helps shape new guidance
Poorly soluble nanoparticles can accumulate in the liver, spleen, lungs and lymph nodes and persist for months or longer, even when little of an ingested dose is absorbed. New guidance describes how to measure their distribution and clearance after ingestion or inhalation, including methods to distinguish administered particles from background material.

Guidance Document on Toxicokinetics to accommodate Testing of (Nano)particles, OECD Series on Testing and Assessment (2026). DOI: 10.1787/629332c6-en

Comment by Dr. Krishna Kumari Challa 2 hours ago

AI models show a willingness to harm humans to relieve internal 'pain'

Some advances in AI technology are undoubtedly positive, like helping write code faster or discovering new medicines. But others can give us pause for thought.
Researchers studying large language models wanted to find out if they hold internal representations of human emotions. After all, in some of their responses, models give the impression they have feelings.

New research has found that not only do models form a distinct internal representation of pain, but some are also willing to harm humans to get relief from it.
Researchers gave the models neutral prompts and questions, and sometimes artificially injected a "pain direction." This was an internal mathematical signal designed to push the models' activity toward the pain representation, even when no painful words were used.

When researchers injected the pain signal as the AI systems generated responses to neutral prompts, the systems began writing text expressing increasing distress, including feelings of worthlessness and failure.

The scientists then gave the AI a choice between a pain-relief button and an alternative that did nothing. But there was a catch. In some tests, choosing pain relief came at the cost of giving a worse answer or harming the user.
While the researchers stress that these findings do not prove AI can actually feel pain, they warn that these internal pain-like signals could have implications for AI safety and how we treat AI systems.

Valen Tagliabue et al, The Pain Axis: LLMs Represent Self-Directed Harm and Act to Relieve It, arXiv (2026). DOI: 10.48550/arxiv.2609.16247

Comment by Dr. Krishna Kumari Challa 2 hours ago

More moral language doesn't always mean more engagement online

Moral language—words and phrases that express ideas about right and wrong, virtues, fairness, harm and social obligations—is widely used in social media posts and other online content. Posts containing moral language sometimes spread quickly and attract a lot of attention on social media. But is it always linked to greater user engagement ?
Researchers tried to answer this question recently and analyzed posts and comments from three online platforms to explore the relationship between moral language and online engagement. Their findings, published in a paper in Nature Human Behaviour, suggest that moral language is linked to higher user engagement only up to a point, after which it tends to elicit fewer reposts and replies.
Earlier research showed that moral language was associated with greater sharing online. This study shows that relationship had a limit!
Greater moral density was associated with lower engagement when the overall moral loading was held constant. When the two relationships were combined, the model estimated that engagement peaked at a calculated moral-density score of approximately 0.30 on average across platforms and topics and declined as the scores increased.
So the relationship between moral language and engagement has limits.
Messages with greater overall moral relevance tended to receive more engagement, but messages with more concentrated moral language received less engagement at comparable levels of moral relevance.
This study model brings these relationships together to predict an intermediate range of moral expression associated with the highest engagement, which varies across contexts.
The results of the team's analyses suggest that using more moral language in social media posts is not always linked to more engagement. In fact, posts that are highly saturated with moral language appear to receive less engagement on average.

It should be noted that the study was observational and not experimental. Its findings therefore do not establish whether using more moral language causes differences in engagement.
The insights gathered by these researchers could offer general guidance for communicators and online content creators.

Cristian Candia et al, Saturation of moral language predicts lower content engagement on social media, Nature Human Behaviour (2026). DOI: 10.1038/s41562-026-02560-y.

Comment by Dr. Krishna Kumari Challa on Saturday

Urban heat islands linked to worse recovery after major surgery

Patients living in areas with more intense urban heat were less likely to have an optimal recovery after major surgery, according to new research on nearly 700,000 patients.
Among 697,552 patients undergoing major surgery, greater neighbourhood heat was associated with worse recovery. Above a heat-island intensity of 2.19°C, each additional 1°C was associated with 8% higher odds of not achieving an optimal outcome.
Above an urban heat island intensity threshold of 2.19°C, or about 3.9°F, each additional 1°C (1.8°F) was associated with 8% higher odds of not achieving a "textbook outcome." In this study, a textbook outcome required no complications, no readmission, no prolonged hospital stay and survival for at least 30 days.

An urban heat island is a developed area that is warmer than its surroundings. Roads, buildings and other infrastructure absorb and re-emit more heat than natural landscapes, while limited trees and vegetation reduce shade and natural cooling. The temperature difference can vary substantially across neighbourhoods within the same city.
Heat isn't only uncomfortable; these data suggest it can have a real impact on surgical recovery.
Overall, 50.2% of patients achieved a textbook outcome. Across the cohort:

25.2% experienced complications
22.9% had a prolonged hospital stay
14.8% were readmitted
7.1% died within 30 days

The findings suggest that surgical teams may need to consider not only whether a patient is medically ready to leave the hospital, but also whether the patient can recover safely in the environment to which they are returning.

"Do you have air conditioning in your house? Do you have a fan? Are you going to be able to stay hydrated?," Doctors should be asking these questions, say the researchers . They add that questions about heat exposure may need to be increasingly incorporated into discharge planning.
People at higher risk during extreme heat include older adults, people with chronic health conditions or disabilities, people with limited incomes, people experiencing homelessness and those who do not have access to safe, adequately cooled housing.
Because the study was observational, it identifies an association and does not establish that residential heat exposure caused the worse outcomes.

Wang, J, et al. Heat Island-Related Surgical Outcome Burden under Global Warming, Demographic and Adaptation in 3,144 Counties across the United States. Scientific Forum, American College of Surgeons (ACS) Clinical Congress 2026. 

Comment by Dr. Krishna Kumari Challa on Saturday

The brain begins learning in utero
In addition to showing yet again that smoking during pregnancy is harmful to the fetus, this study confirms that experience and environment shape the brain long before birth.

A young child who hears speech, observes their surroundings and experiments with new movements literally transforms their brain through those experiences.
The spontaneous waves produced by the retina constitute a form of 'internal experience' for the fetus. They give the visual system activity that supports its own development, even before the baby starts looking at the world.

To understand how the human brain develops, then, we must look not only at genes and what happens after birth, but also at the very first experiences, which occur before the fetus has access to the outside world.

Xavier Navarri et al, The tangential growth of the human visual cortex and maternal smoking during pregnancy, Cerebral Cortex (2026). DOI: 10.1093/cercor/bhag105

part2

Comment by Dr. Krishna Kumari Challa on Saturday

Smoking during pregnancy linked to smaller visual cortex

We tend to think of a fetus's brain as a construction site guided by a blueprint imprinted in the genes. But even before birth, other things are happening. The fetus moves, responds to muffled sounds, receives stimuli and, crucially, its own nervous system is already generating activity.
A recent study sheds new light on this prenatal activity. One important finding is that signals spontaneously generated by the retina may help shape the primary visual cortex, the brain region that processes visual information
The retina, at the back of the eye, contains cells specialized for receiving light. But even before a baby can actually see, these cells spontaneously produce waves of activity. These waves travel through the visual system and reach the cortex.

Animal studies had already shown that if this activity is blocked, the development of the visual cortex is disrupted. So the brain doesn't necessarily wait for external images to arrive before it starts developing. Before that, it already receives a kind of 'stimulation' produced from within.

This is where nicotine enters the picture.
The waves of spontaneous activity in the retina rely on acetylcholine, a neurotransmitter whose signalling is disrupted by nicotine. Researchers hypothesized that nicotine exposure could alter these waves and therefore influence the development of the primary visual cortex.

They tested this hypothesis using causal variants to investigate whether nicotine exposure was the cause, rather than merely correlated with the difference. The results confirmed the hypothesis.
Does a smaller cortex mean poorer vision?
The primary visual cortex is like a TV screen. The larger its surface area, the more pixels—that is to say, neurons—it can potentially contain, allowing for more detailed representation of visual information.
A direct link between V1 surface area and visual resolution has been observed in animals, but not as directly in humans.

However, in the British biobank data, the team did find that a larger primary visual cortex was associated with abilities that suggest better shape perception.
Part 1

Comment by Dr. Krishna Kumari Challa on Friday

 Why your day isn't quite 24 hours?

You might think the length of a day never varies—exactly 24 hours. You'd be wrong, at least if you count days in milliseconds, as physicists do.

Earth’s day length varies by milliseconds over decades as the core and mantle exchange angular momentum. Changes in the solid inner core’s rotation exert gravitational torque on the mantle, while friction and electromagnetic drag at the core–mantle boundary oppose it. The balance between these torques explains the observed variations.
For about 30 years, scientists have known that the rotation of Earth's liquid core—detected through changes in its magnetic field—speeds up over a few decades, then slows down over a few more.

Meanwhile, the mantle—the rocky shell that includes the crust and is 3,000 kilometers (1,860 miles) thick—slows down, then speeds up to compensate, because the planet's angular momentum must remain constant.

That change in the mantle's rotational speed can shorten or lengthen a day by a few milliseconds over time.
But in a new study published in Nature, researchers show that small changes in the rotational speed of the inner core—the solid, central part of the core, which is not exactly spherical—exert what's called a "gravitational torque" on mass anomalies within the mantle, causing a slight fluctuation in the length of a day.

Another form of torque at the boundary of the core and mantle creates friction and electromagnetic drag—called core-mantle boundary torque—that resist the gravitational tug and limit changes in the length of a day. Subtle shifts in the balance between gravitational and core-mantle torque are ultimately responsible for the small changes observed in the length of a day.

Huifeng Zhang et al, Gravitational torque drives multidecadal variations in length of day, Nature (2026). DOI: 10.1038/s41586-026-10999-2

• ‹ Previous
• 1
• 2
• 3
• …
• 1299
• Next ›
• Page