Comment

Comment by Dr. Krishna Kumari Challa yesterday

Schizophrenia-spectrum disorders may originate in specific brain regions that show early structural damage

 

Schizophrenia-spectrum disorders are linked to early structural damage in specific brain regions, particularly the temporal, cingulate, and insular lobes, with reduced morphological similarity indicating network disconnection. These changes are more pronounced in severe cases and correlate with altered neurobiology, including increased astrocytes, neurotransmitters, and reduced metabolism.

Natalia García-San-Martín et al, Reduced brain structural similarity is associated with maturation, neurobiological features, and clinical status in schizophrenia, Nature Communications (2025). DOI: 10.1038/s41467-025-63792-6

Comment by Dr. Krishna Kumari Challa yesterday

Interestingly, Ash1l belongs to a family of proteins called histone methyltransferases that retain memory in other biological systems as well. "In the immune system, these molecules help the body remember past infections; during development, those same molecules help cells remember that they've become a neuron or muscle and maintain that identity long-term. The brain may be repurposing these ubiquitous forms of cellular memory to support cognitive memories.

The findings may have implications for memory-related diseases.

By identifying the gene programs that preserve memory, researchers may eventually find ways to route memory through alternate circuits and around damaged parts of the brain in conditions such as Alzheimer's.

If we know the second and third areas that are important for memory consolidation, and we have neurons dying in the first area, perhaps we can bypass the damaged region and let healthy parts of the brain take over.

Part3

Comment by Dr. Krishna Kumari Challa yesterday

For decades, memory research focused on two brain regions: the hippocampus, home of short-term memory, and the cortex, which was thought to house long-term memories. The latter, scientists imagined, lie gated behind biological on-and-off switches.

Existing models of memory in the brain involved transistor-like memory molecules that act as on/off switches

In other words, in this model, if a short-term memory was tagged for long-term storage, it would remain so indefinitely. But, even as investigations in this vein led to numerous insights, researchers understood that this model was ultimately too simple—for instance, it didn't account for why some long-term memories last weeks while others last a lifetime.

Then, in 2023, the same researchers published a paper that identified a brain pathway that links short- and long-term memories. An important component of which is a region in the center of the brain called the thalamus, which not only helps select which memories should be remembered, but routes them to the cortex for long-term stabilization.

The findings set the stage for tackling some of the most fundamental questions in the field of memory research: What happens to memories beyond short-term storage in the hippocampus—and what molecular mechanisms are behind the sorting process that promotes important memories to the cortex and demotes unimportant ones to be forgotten?

To answer these questions, the team developed a behavioral model using a virtual reality system where mice formed specific memories.

The results suggest that long-term memory is not maintained by a single molecular on/off switch, but by a cascade of gene-regulating programs that unfold over time and across brain regions like a series of molecular timers.

Initial timers turn on quickly and fade just as fast, allowing for rapid forgetting; later timers act more slowly but create more durable memories. This stepwise process allows the brain to promote important experiences for long-term storage, while others fade.

In this study, the researchers used repetition as a proxy for importance, comparing memories of frequently repeated contexts to those encountered less often. The team identified three transcriptional regulators: Camta1 and Tcf4 in the thalamus, and Ash1l in the anterior cingulate cortex, which are not necessary for initially forming memories, but are crucial for maintaining them. Disrupting Camta1 and Tcf4 impaired functional connections between the thalamus and cortex, leading to memory loss.

The model suggests that, after the basic memory is formed in the hippocampus, Camta1 and its targets ensure the initial persistence of the memory. With time, Tc4 and its targets are activated, providing cell adhesion and structural support to further maintain the memory. Finally, Ash1l recruits chromatin remodeling programs that make the memory more persistent.

Unless you promote memories onto these timers, the researchers think you're primed to forget it quickly.

Part2

Comment by Dr. Krishna Kumari Challa yesterday

How the brain decides what to remember: Study reveals sequentially operating molecular 'timers'

Every day, our brains transform quick impressions, flashes of inspiration, and painful moments into enduring memories that underpin our sense of self and inform how we navigate the world. But how does the brain decide which bits of information are worth keeping—and how long to hold on?

Now, new research demonstrates that long-term memory is formed by a cascade of molecular "timers" unfolding across brain regions. With a virtual reality-based behavioral model in mice, the scientists discovered that long-term memory is orchestrated by key regulators that either promote memories into progressively more lasting forms or demote them until they are forgotten. 

The findings, published in Nature, highlight the roles of multiple brain regions in gradually reorganizing memories into more enduring forms, with gates along the way to assess salience and promote durability.

This is a key revelation because it explains how we adjust the durability of memories.  What we choose to remember is a continuously evolving process rather than a one-time flipping of a switch.

Initial timers turn on quickly and fade just as fast, allowing for rapid forgetting; later timers act more slowly but create more durable memories. This stepwise process allows the brain to promote important experiences for long-term storage, while others fade. 

Unless you promote memories onto these timers, the researchers think you're primed to forget it quickly.

Priya Rajasethupathy, Thalamocortical transcriptional gates coordinate memory stabilization, Nature (2025). DOI: 10.1038/s41586-025-09774-6. www.nature.com/articles/s41586-025-09774-6

Part1

Comment by Dr. Krishna Kumari Challa yesterday

High levels of forever chemicals found in dolphins and whales

New research has revealed that marine mammals who live far below the ocean's surface are not immune from the burden of toxic forever chemicals, with whales and dolphins showing unprecedented levels of PFAS contamination. 

The findings challenge the assumption that a deep-sea habitat offers protection from human-made per- and polyfluoroalkyl substances, otherwise known as PFAS. 

Published in Science of the Total Environment, the findings raise concerns about the long-term health of marine species and the invisible legacy that forever chemicals are leaving in the environment. PFAS are human-made chemicals that accumulate through the food chain and can disrupt immune, endocrine and reproductive systems, raising concerns for both individual and population health in humans and animals, including cetaceans. 

The scientists analyzed tissues from 127 animals across 16 species of toothed whales and dolphins .

The researchers looked at how the acquisition of forever chemicals varied according to species, sex, age and the habitat in which they predominantly live and feed. 

The results showed that  even offshore and deep-diving species are exposed to similar levels of PFAS, highlighting how widespread pollution, compounded by climate-driven stressors, poses a growing threat to marine biodiversity.

Karen A. Stockin et al, No place to hide: Marine habitat does not determine per- and polyfluoroalkyl substances (PFAS) in odontocetes, Science of The Total Environment (2025). DOI: 10.1016/j.scitotenv.2025.180701

Comment by Dr. Krishna Kumari Challa yesterday

The researchers found that the three tested LLMs failed to generate plausible passwords for each user profile, seldom guessing the correct password. Other computational tools that are currently used to guess passwords, known as rule-based and combinator-based techniques, performed significantly better. 

They  also carried out additional analyses aimed at shedding more light on why LLMs perform so poorly when asked to generate plausible passwords. The results of these analyses suggest that these models lack some of the skills necessary to complete this task, such as the ability to recall specific examples encountered during training and to apply learned password patterns in new scenarios.

"Through detailed analysis and visualization, we identify key limitations in the generative reasoning of LLMs when applied to the domain-specific task of password guessing," wrote the authors. 

The  findings suggest that, despite their linguistic prowess, current LLMs lack the domain adaptation and memorization capabilities required for effective password inference, especially in the absence of supervised fine-tuning on leaked password datasets.

Overall, this recent study suggests that LLMs are not currently suitable for the inference of passwords. 

Mohammad Abdul Rehman et al, When Intelligence Fails: An Empirical Study on Why LLMs Struggle with Password Cracking, arXiv (2025). DOI: 10.48550/arxiv.2510.17884

Part2

Comment by Dr. Krishna Kumari Challa yesterday

Why LLMs are not great at cracking passwords

Large language models (LLMs), such as the model underpinning the functioning of OpenAI's conversational platform ChatGPT, have proved to perform well on various language-related and coding tasks. Some computer scientists have recently been exploring the possibility that these models could also be used by malicious users and hackers to plan cyber-attacks or access people's personal data. 

LLMs can generate texts or code tailored for specific purposes and that meet user-specified requirements. In principle, when given information about people, they could also be able to generate passwords that they might use, which include names or dates that are meaningful to them. 

Researchers  recently carried out a study exploring this possibility. Their findings, published on the arXiv preprint server, suggest that most existing LLMs cannot reliably generate plausible passwords for specific users, while also providing an explanation for why they perform poorly on this task. 

XXXX  The remarkable capabilities of large language models (LLMs) in natural language understanding and generation have sparked interest in their potential for cyber security applications, including password guessing.XXXX

The researchers conducted an empirical investigation into the efficacy of pre-trained LLMs for password cracking using synthetic user profiles. 

To carry out their study, the researchers first created synthetic (fake) profiles for non-existing users, which included their names, birthdays and hobbies. They then asked three different LLMs to generate a list of passwords that each of these fictional users might choose when protecting their accounts. 

They evaluated the performance of state-of-the-art open-source LLMs such as TinyLLaMA, Falcon-RW-1B, and Flan-T5 by prompting them to generate plausible passwords based on structured user attributes (e.g., name, birthdate, hobbies).

To assess the performance of the models' on the task, the researchers used metrics commonly used in studies that focus on information retrieval and password-guessing. These metrics, called Hit@1, Hit@5 and Hit@10, specifically measure how good a model is guessing passwords correctly, or specifically, ranking correct passwords as the most plausible.

Their results, measured using Hit@1, Hit@5, and Hit@10 metrics under both plaintext and SHA-256 hash comparisons, reveal consistently poor performance, with all models achieving less than 1.5% accuracy at Hit@10.

In contrast, traditional rule-based and combinator-based cracking methods demonstrate significantly higher success rates.

 Part1

Comment by Dr. Krishna Kumari Challa yesterday

Thousands of genomes reveal the wild wolf genes in most dogs' DNA

Analysis of nearly 2,700 ancient and modern canid genomes shows that most dogs retain small but detectable segments of wild wolf DNA, indicating ongoing gene flow after domestication. About two-thirds of dog breeds and all village dogs carry traces of wolf ancestry, often in genes related to olfaction. Dog genes are also present in about half of wild wolf genomes.

https://www.pnas.org/doi/10.1073/pnas.2421768122

Comment by Dr. Krishna Kumari Challa yesterday

Here is something to make you more lazy: A human washing machine!

A human washing machine is now on sale in Japan.  

Users lie down in the pod, close the lid and get cleaned like clothes in a washing machine—but without the spin—while music plays.

A prototype of the device, called the human washer of the future, attracted long lines at the six-month Expo that wrapped up in Osaka in October after welcoming over 27 million people.

Made by Japanese firm Science, the device is an update of a product displayed the last time Osaka hosted the event, in 1970.

https://www.youtube.com/watch?v=LB6LmIWAJLw&t=61s

https://www.youtube.com/watch?v=AfNrAIITDi0

Comment by Dr. Krishna Kumari Challa yesterday

Climate friendly metals can come from deep sea ores

Deep-sea manganese nodules offer a potential alternative to land-based mining for critical metals like copper, nickel, and cobalt, with significantly less deforestation, waste, and social harm. A hydrogen-based extraction process reduces CO₂ emissions by over 90% and uses less energy compared to carbon-based methods, making deep-sea mining more sustainable if managed responsibly.

Ubaid Manzoor et al, Low-waste, single-step, sustainable extraction of critical metals from deep-sea polymetallic nodules, Science Advances (2025). DOI: 10.1126/sciadv.aea1223

• ‹ Previous
• 1
• 2
• 3
• 4
• …
• 1152
• Next ›
• Page