‘Raw’ data show AI signals mirror how the brain listens and learns

Researchers found strikingly similar signals between the brain and artificial neural networks. The blue line is brain wave when humans listen to a vowel. Red is the artificial neural network’s response to the exact same vowel. The two signals are raw, meaning no transformations were needed.
Illustration Credit: Courtesy Gasper Begus

New research from the University of California, Berkeley, shows that artificial intelligence (AI) systems can process signals in a way that is remarkably similar to how the brain interprets speech, a finding scientists say might help explain the black box of how AI systems operate.

Using a system of electrodes placed on participants’ heads, scientists with the Berkeley Speech and Computation Lab measured brain waves as participants listened to a single syllable — “bah.” They then compared that brain activity to the signals produced by an AI system trained to learn English.

“The shapes are remarkably similar,” said Gasper Begus, assistant professor of linguistics at UC Berkeley and lead author on the study published recently in the journal Scientific Reports. “That tells you similar things get encoded, that processing is similar. “

A side-by-side comparison graph of the two signals shows that similarity strikingly.

Scientists discover anatomical changes in the brains of the newly sighted

MIT neuroscientists discovered anatomical changes that occur in the white matter pathways linking visual-processing areas of the brain in children who have congenital cataracts surgically removed. This image shows the late-visual pathways in the brain.
Illustration Credit: Courtesy of the researchers, edited by MIT News

For many decades, neuroscientists believed there was a “critical period” in which the brain could learn to make sense of visual input, and that this window closed around the age of 6 or 7.

Recent work from MIT Professor Pawan Sinha has shown that the picture is more nuanced than that. In many studies of children in India who had surgery to remove congenital cataracts beyond the age of 7, he has found that older children can learn visual tasks such as recognizing faces, distinguishing objects from a background, and discerning motion.

In a new study, Sinha and his colleagues have now discovered anatomical changes that occur in the brains of these patients after their sight is restored. These changes, seen in the structure and organization of the brain’s white matter, appear to underlie some of the visual improvements that the researchers also observed in these patients.

The findings further support the idea that the window of brain plasticity, for at least some visual tasks, extends much further than previously thought.

How does climate change affect global bird reproduction?

Avian ecologist Jeff Hoover and his colleagues explored the potential effects of global warming on bird reproductive output across the world. 
Photo Credit: Fred Zwicky

A new study reported in the Proceedings of the National Academy of Sciences assessed changes in the reproductive output of 104 bird species between 1970 and 2019. Illinois Natural History Survey avian ecologist Jeff Hoover, a co-author of the paper, spoke to News Bureau life sciences editor Diana Yates about the findings and how climate change is altering bird ecology and health around the world. 

What is unique about the study?  

This study explored potential effects of global climate change – in particular, warming – on offspring production for over 100 species from more than 200 bird populations across all continents. We looked at data for each of these bird populations over 15 to 49 breeding seasons to consider if changes in local temperatures and precipitation were associated with changes in the average number of offspring produced per female per year.

Beyond effects of a warming climate on individual species’ reproductive output, the study also considered whether climate change may affect offspring production by interacting with other attributes of the birds. Such traits include body mass, migration status, habitat needs, human impacts to local landscapes, the protection/conservation status of sites and whether the birds can produce two broods in a single breeding season. The temporal and spatial scales of this work and the number of species and populations studied were monumental.  

Fish thought to help reefs have poop that’s deadly to corals

A coral-eating butterflyfish on a Moorea reef in July 2019.
Photo Credit: Carsten Grupstra

Feces from fish that are typically thought to promote healthy reefs can damage and, in some cases, kill corals, according to a recent study by Rice University marine biologists.

Until recently, fish that consume algae and detritus — grazers — were thought to keep reefs healthy, and fish that eat coral — corallivores — were thought to weaken reef structures. The researchers found high levels of coral pathogens in grazer feces and high levels of beneficial bacteria in corallivore feces, which they say could act like a “coral probiotic.”

“Corallivorous fish are generally regarded as harmful because they bite the corals,” said Carsten Grupstra, a former graduate student at Rice and lead author of the open-access study in Frontiers in Marine Science. “But it turns out that this doesn’t tell the whole story.”

Researchers develop technique for rapid detection of neurodegenerative diseases like Parkinson’s and CWD

Illustration Credit: Sang-Hyun Oh Research Group, University of Minnesota

University of Minnesota researchers have developed a groundbreaking new diagnostic technique that will allow for faster and more accurate detection of neurodegenerative diseases. The method will likely open a door for earlier treatment and mitigation of various diseases that affect humans, such as Alzheimer's and Parkinson's, and similar diseases that affect animals, such as chronic wasting disease (CWD).

Their new study is published in Nano Letters.

“This research mainly focuses on chronic wasting disease in deer, but ultimately our goal is to expand the technology for a broad spectrum of neurodegenerative diseases, Alzheimer’s and Parkinson’s being the two main targets,” said Sang-Hyun Oh, senior co-author of the paper and a professor in the College of Science and Engineering. “Our vision is to develop ultra-sensitive, powerful diagnostic techniques for a variety of neurodegenerative diseases so that we can detect biomarkers early on, perhaps allowing more time for the deployment of therapeutic agents that can slow down the disease progression. We want to help improve the lives of millions of people affected by neurodegenerative diseases.”

Researchers develop clever algorithm to improve our understanding of particle beams in accelerators

A representation of a particle beam traveling through an accelerator.
Illustration Credit: Greg Stewart/SLAC National Laboratory

The algorithm pairs machine-learning techniques with beam physics equations to avoid massive data crunching.

Whenever SLAC National Accelerator Laboratory’s linear accelerator is on, packs of around a billion electrons each travel together at nearly the speed of light through metal piping. These electron bunches form the accelerator’s particle beam, which is used to study the atomic behavior of molecules, novel materials and many other subjects. But trying to estimate what a particle beam actually looks like as it travels through an accelerator is difficult, often leaving scientists with only a rough approximation of how a beam will behave during an experiment.

Now, researchers at the Department of Energy’s SLAC National Accelerator Laboratory, the DOE’s Argonne National Laboratory and the University of Chicago have developed an algorithm that more precisely predicts a beam’s distribution of particle positions and velocities as it zips through an accelerator. This detailed beam information will help scientists perform their experiments more reliably – a need that is becoming increasingly important as accelerator facilities operate at higher and higher energies and generate more complex beam profiles. The researchers detailed their algorithm and method in April in Physical Review Letters.

Webb Finds Water Vapor, But from a Rocky Planet or Its Star?

This artist concept represents the rocky exoplanet GJ 486 b, which orbits a red dwarf star that is only 26 light-years away in the constellation Virgo. By observing GJ 486 b transit in front of its star, astronomers sought signs of an atmosphere. They detected hints of water vapor. However, they caution that while this might be a sign of a planetary atmosphere, the water could be on the star itself – specifically, in cool starspots – and not from the planet at all.  GJ 486 b is about 30% larger than the Earth and weighs three times as much. It orbits its star closely in just under 1.5 days.
Illustration Credit: NASA, ESA, CSA, Joseph Olmsted (STScI)

The most common stars in the universe are red dwarf stars, which means that rocky exoplanets are most likely to be found orbiting such a star. Red dwarf stars are cool, so a planet has to hug it in a tight orbit to stay warm enough to potentially host liquid water (meaning it lies in the habitable zone). Such stars are also active, particularly when they are young, releasing ultraviolet and X-ray radiation that could destroy planetary atmospheres. As a result, one important open question in astronomy is whether a rocky planet could maintain, or reestablish, an atmosphere in such a harsh environment.

To help answer that question, astronomers used NASA’s James Webb Space Telescope to study a rocky exoplanet known as GJ 486 b. It is too close to its star to be within the habitable zone, with a surface temperature of about 800 degrees Fahrenheit (430 degrees Celsius). And yet, their observations using Webb’s Near-Infrared Spectrograph (NIRSpec) show hints of water vapor. If the water vapor is associated with the planet, that would indicate that it has an atmosphere despite its scorching temperature and close proximity to its star. Water vapor has been seen on gaseous exoplanets before, but to date no atmosphere has been definitively detected around a rocky exoplanet. However, the team cautions that the water vapor could be on the star itself – specifically, in cool starspots – and not from the planet at all.

“Eat, Sleep, Console” reduces hospital stay and need for medication among opioid-exposed infants

Photo Credit: Fujikama

Researchers have found the “Eat, Sleep, Console” (ESC) care approach to be more effective than using the Finnegan Neonatal Abstinence Scoring Tool (FNAST) to assess and manage opioid-exposed newborns, according to a national, randomized controlled clinical trial funded by the National Institutes of Health. Newborns cared for with ESC were medically ready for discharge approximately 6.7 days earlier and 63% less likely to receive medication as part of their treatment, compared to newborns cared for with FNAST. ESC prioritizes non-pharmacologic approaches to care, such as a low-stimulation environment, swaddling, skin-to-skin contact and breastfeeding. ESC also encourages parental involvement in the care and assessment of their infants. These findings are based on the hospital outcomes of a large and geographically diverse group of opioid-exposed infants. A two-year follow-up study of a subset of infants is ongoing. The current findings are published in the New England Journal of Medicine.

“Medical care for newborns who were exposed to opioids during pregnancy varies widely across hospitals,” said Diana W. Bianchi, M.D., director of NIH’s Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), which co-led the study with the NIH Environmental Influences on Child Health Outcomes (ECHO) Program. “These findings are an important step toward standard, evidence-based guidance for the care of these infants.”

iDNA from flies to track native species across Western Australian wheatbelt

Carrion Fly
Photo Credit: Ian Lindsay

Researchers from Curtin University have collected iDNA from flies to track the movements of Australia’s native species across the Western Australian wheatbelt, with hopes to improve future conservation efforts in the region.

Published in the Journal Conservation Biology, the research team found that native animals, such as the echidna, numbat, woylie and chuditch, were predominantly located in conservation reserves and not across the wider wheatbelt landscape, compared to invasive species like foxes and feral cats which were found across all areas.

Senior researcher and co-author Associate Professor Bill Bateman, from Curtin’s School of Molecular and Life Sciences, said native mammal populations were declining at alarming rates and there was an urgent need to monitor and protect their wellbeing.

“It is essential to monitor the distribution and movements of animals so we can identify which populations are most at risk, which ones are declining, and which ones are on the brink of extinction. Tracking wildlife through alternative techniques, such as camera trapping and audio recording, can be difficult, costly and take several weeks to gather data,” Associate Professor Bateman said.

The Trumpet biocomputing platform heralds a new path for medicine

A biocomputing chip made of bacteria.
Image Credit: College of Biological Sciences / University of Minnesota

From early detection and internal treatment of diseases to futuristic applications like augmenting human memory, biological computing, or biocomputing, it has the potential to revolutionize medicine and computers. Traditional computer hardware is limited in its ability to interface with living organs, which has constrained the development of medical devices. Computerized implants require a constant supply of electricity, they can cause scarring in soft tissue that makes them unusable and they cannot heal themselves the way organisms can. Through the use of biological molecules such as DNA or proteins, biocomputing has the potential to overcome these limitations.

Biocomputing is typically done either with live cells or with non-living, enzyme-free molecules. Live cells can feed themselves and can heal, but it can be difficult to redirect cells from their ordinary functions towards computation. Non-living molecules solve some of the problems of live cells, but have weak output signals and are difficult to fine-tune and regulate.