It’s possible that I shall make an ass of myself. But in that case one can always get out of it with a little dialectic. I have, of course, so worded my proposition as to be right either way (K.Marx, Letter to F.Engels on the Indian Mutiny)
IMAGE: BY OBSERVING A COMET IN THERMAL INFRARED WAVELENGTHS, THE SAME WAVELENGTHS USED BY NONCONTACT THERMOMETERS, IT IS POSSIBLE TO DETERMINE NOT ONLY ITS CURRENT TEMPERATURE, BUT ALSO THE SURFACE COMPOSITION... view more
CREDIT: KYOTO SANGYO UNIVERSITY
The world's first ground-based observations of the bare nucleus of a comet nearing the end of its active life revealed that the nucleus has a diameter of 800 meters and is covered with large grains of phyllosilicate; on Earth large grains of phyllosilicate are commonly available as talcum powder. This discovery provides clues to piece together the history of how this comet evolved into its current burnt-out state.
Comet nuclei are difficult to observe because when they enter the inner Solar System, where they are easy to observe from Earth, they heat up and release gas and dust which form a coma obscuring the nuclei. When Comet P/2016 BA14 (PANSTARRS) was discovered in January 2016 it was first mistaken for an asteroid, but subsequent observations revealed weak cometary activity. It is believed that after many trips through the inner Solar System, this comet has burnt off almost all of its ice and is now nearing the end of its cometary life.
On March 22, 2016, this comet passed within 3.6 million kilometers of Earth, only 9 times farther than the Moon. A team of astronomers from the National Astronomical Observatory of Japan (NAOJ) and Koyama Astronomical Observatory of Kyoto Sangyo University used this unique opportunity to observe the comet with the Subaru Telescope about 30 hours before its closest approach to Earth. They successfully observed the nucleus with minimal interference from dust grains in the coma. Previously, the surface composition of a cometary nucleus has only been observed by a few "in-situ" observations by space missions.
Because the team observed thermal infrared radiation, the same region of the infrared used by contactless thermometers, they were able to find evidence that the nucleus is 800 meters in diameter and covered with organic molecules and large grains of phyllosilicate. This is the first time hydrous silicate minerals such as talc have been found in a comet. Comparison with laboratory measurements of various minerals revealed that the hydrous silicate minerals on the surface of P/2016 BA14 have been heated to more than about 330 degrees Celsius in the past. Since the surface temperature of P/2016 BA14 cannot reach higher than about 130 degrees Celsius in its current orbit, the comet may have been in an orbit closer to the Sun in the past.
The next question is whether comets are covered with talcum powder from the start or if it develops over time as they burn out. "This result provides us a precious clue to study the evolution of comets." comments Dr. Takafumi Ootsubo, the lead author of this research, "We believe that further observations of the comet nuclei will enable us to learn more about the evolution of comets."
###
The target of this research, P/2016 BA14, is a potential backup target for the Comet Interceptor mission, a comet exploration mission being carried out by the ESA and JAXA.
Are you using antihistamines properly?
Antihistamines can be helpful during spring allergy season and as COVID vaccines roll out, for which rashes are common
IMAGE: DEREK CHU, CLINICAL SCHOLAR, MEDICINE view more
CREDIT: MCMASTER UNIVERSITY
Hamilton, ON (April 6, 2021) - If you are one of the millions of people worldwide suffering from allergies, you may take an antihistamine pill to ward off hives, sneezing and watery eyes.
But you may be taking your medications incorrectly, says Derek Chu, a McMaster University allergy expert and clinical scholar.
"People need to rethink what they stock in their home cabinets as allergy medicines, what hospitals keep on formulary, and what policymakers recommend. The message needs to get out. This publication is on time for the spring allergy season and as COVID vaccines roll out, for which rashes are common and antihistamines can be helpful," said Chu.
Co-author Gordon Sussman, professor of medicine at University of Toronto's Temerty Faculty of Medicine, agrees.
"There are new modern second generation antihistamines that are potent, specific, fast acting and of proven safety which everyone should be using to treat allergic rhinitis and hives," he said.
This team's study is published in the Canadian Medical Association Journal (CMAJ), and list five facts to know:
1. Antihistamines are among the most commonly and incorrectly used medicines worldwide. Antihistamines are best used to relieve symptoms of hay fever and outbreaks of hives, but not for asthma, eczema, coughs or insomnia.
2. First-generation antihistamines are associated with substantial and sometimes fatal side effects. Antihistamines with diphenhydramine, chlorpheniramine or hydroxyzine are first-generation antihistamines that can cause drowsiness and affect cognitive functions such as school performance. Overdosing can cause death and they are potentially dangerous for both young and older people.
3. Newer antihistamines are safer, as affordable and as effective as first-generation antihistamines. Later-generation antihistamines are proven to be more effective and last longer with less side-effects like drowsiness.
4. Antihistamines should not be used instead of epinephrine to treat anaphylaxis. Oral drugs can be used together with epinephrine injections to treat anaphylactic reactions, but they are not a substitute.
5. Most antihistamines are safe to use during pregnancy and breastfeeding. Medical research has shown that antihistamines at standard doses do not harm fetuses during pregnancy and may be used during breastfeeding. They are also safe for children to use.
###
No external funding was received for this CMAJ article.
IMAGE: THERE ARE A TOTAL OF OVER 2,500 DESCRIBED GOLDEN WASP SPECIES, WITH 40 RECORDED IN NORWAY. view more
CREDIT: PHOTO: ARNSTEIN STAVERLØKK, NINA, CC BY3.
Cuckoo wasps - also called emerald wasps - are some of the most beautiful insects we have, with colourful exteriors that shine like jewels. However, these beauties have also created a lot of headaches.
"Normally we distinguish insects from each other by their appearance, but cuckoo wasps are so similar to each other that it makes it difficult," says Frode Ødegaard.
Ødegaard is an insect researcher at the NTNU (Norwegian University of Science and Technology) University Museum and belongs to the European research group that has now described thia recent contribution to species diversity. The new species is very rare, and is only a single specimen has been found on the Lista peninsula in Agder county in Norway.
For more than 200 years, insect researchers have struggled to sort cuckoo wasps into the right "species boxes," and to determine which characteristics are variations within a species and which are species-specific differences.
In the last 10 years, DNA barcoding has brought about a major breakthrough by making it possible to distinguish different species of cuckoo wasps from each other by looking at the differences in their genetic material.
"But it's not always that easy, either. In this case, we had two cuckoo wasps with microscopic differences in appearance and very small differences in DNA," Ødegaard says.
"The next step was to look at the language of each of the wasps to find out if they belonged to different species," he says.
Insects communicate with each other through pheromones - in other words, they have a chemical language. Very closely related species often have completely different languages to prevent them from interbreeding.
The cuckoo wasp is an insect with above-average linguistic abilities. They are parasites, which means that they behave like cuckoos and lay their eggs in the nests of other bees and wasps. The larvae grow quickly and hatch before the host's eggs. Then they eat the eggs, the larvae and the food supply that the host has arranged in the nest.
"When you live as a parasite, it's important not to be discovered, and therefore the cuckoo wasp has also learned the language of its host," says Ødegaard.
By conducting an ever-so-small language study, the researchers were able to discover that the two almost identical cuckoo wasps did indeed belong to different species. They use different hosts - and that means that they also speak completely different languages.
"The evolutionary development associated with sponging off another species happens very fast. That's why you can have two species that are really similar genetically but still belong to different species," says Ødegaard.
When a new species is described it has to be given a name, and Frode Ødegaard had the good fortune to receive the honour of naming the newcomer.
"A naming competition was announced among researchers in Europe who work with cuckoo wasps, and then the proposals that came in were voted on. It turned out my proposal actually got the most votes!" Ødegaard says.
"As mentioned, the new wasp is very similar to another species called Chrysis brevitarsis, so the new species was named Chrysis parabrevitarsis, which means 'the one standing next to brevitarsis'."
Ødegaard was also responsible for giving the species its slightly simpler Norwegian name of sporegullveps. He makes no secret of the fact that he found it great to be able to name a new species.
"In a way, you place yourself in the perspective of eternity, because that species will always have that name. There's something very fundamental about it."
The only known specimen of this cuckoo wasp has been captured and pinned in an insect collection. So it may seem both morally reprehensible and unnecessary that this one lone individual was stuck onto a needle.
"Even with today's advanced methods, using live animals for studies like this isn't possible, but collecting individual specimens fortunately has no impact on the population," Ødegaard says.
"The insects have enormous reproductive potential, and the size and quality of the habitats are what determine the viability of the population, not whether any specimens are eaten by birds or collected by an insect researcher."
He adds that the collected insects are absolutely crucial for researchers to be able to map and describe their diversity and thus take care of viable populations for posterity.
###
You can read more about how Frode combines his role of mass murderer and conservation biologist in his blog (in Norwegian): http://www.beetlebee.me/
Reference: Frode Ødegaard et al: Cuticular Hydrocarbon Profile Analyses Help Clarify the Species Identity of Dry-Mounted Cuckoo Wasps (Hymenoptera: Chrysididae), Including Type Material, and Reveal Evidence for a Cryptic Species Insect Systematics and Diversity, Volume 5, Issue 1, January 2021, 3, https://doi.org/10.1093/isd/ixab002
Careful fieldwork pays off (IMAGE)
CAPTION
When a new species is described, it has to be given a name, and Frode Ødegaard (pictured) was lucky enough to name the new wasp species.
CREDIT
Photo: Jake Bryant
Rising Sika deer populations linked to bovine TB infections
IMAGE: NEW RESEARCH SUGGESTS IRELAND'S INCREASING POPULATIONS OF SIKA DEER MAY BE LINKED TO LOCAL OUTBREAKS OF TB INFECTION IN CATTLE. ALTHOUGH TB INFECTION RATES HAVE DECREASED IN GENERAL IN RECENT... view more
CREDIT: BENJAMIN LECOMTE
New research suggests Ireland's increasing populations of Sika deer may be linked to local outbreaks of TB infection in cattle.
Although TB infection rates have decreased in general in recent decades, county-level data shows a correlation between higher Sika numbers and higher local TB infections - with County Wicklow a particular hotspot.
The research, conducted by researchers from Trinity College Dublin and the National Parks and Wildlife Service and supported by the Department of Agriculture, Food and the Marine, has major implications for controlling TB. It has just been published in the journal, Frontiers in Veterinary Science.
First author of the journal article, Dr David Kelly, from Trinity's School of Natural Sciences, said:
"Irish farmers have been aware of the effects of bovine TB for well over 70 years. Its incidence has steadily diminished in Ireland, from 3% in 1960 to 0.3% some 50 years later. In the mid-1980s it became clear European badgers were a TB wildlife host. Since then, badger populations have been controlled around farms with TB outbreaks whenever those outbreaks cannot be linked to cattle.
In recent years, however, it has become clear that controlling TB in badger populations cannot rely on culling alone. Along with this change in thinking, there has been a steady shift away from culling and towards badger vaccination. Unfortunately, while the management of one TB wildlife host has made great advances, another wildlife TB host has appeared on the radar: deer."
The number of deer in Ireland has been rising steadily during the 21st century, and studies in Europe and the USA have shown that deer, at higher densities, can sustain TB in their herds. Indeed, recent research has identified Sika deer in County Wicklow as one such maintenance host of TB.
The researchers behind the current study used county-level population densities (taken between 2000 and 2018) to track the three known maintenance hosts of TB in Ireland: cattle, badgers, and Sika deer.
They then considered how variations in local densities compared with the variation in bovine TB infection across the country (and also at county-level).
Dr Kelly said:
"Our analyses show that while there have been general reductions in TB infection, there is a correlation between increasing Sika deer density and higher local TB infection. This pattern has emerged in recent years and has major implications for TB control within Ireland.
"Now, when attempting to manage TB in wildlife, Sika deer will need to be considered as well as badgers. Our analyses suggest Sika deer are currently of greatest concern in County Wicklow but if numbers continue to rise in other counties they may also pose problems elsewhere."
###
LSU Health New Orleans study discovers source of Zika neurodevelopmental defects
New Orleans, LA - A study led by Edward Wojcik, PhD, Associate Professor of Biochemistry & Molecular Biology at LSU Health New Orleans School of Medicine, identified how microcephaly (abnormally small heads) and blindness may develop in Zika-infected fetuses, as well as a new way to potentially prevent these neurodevelopmental defects. The results are published online in iScience, available here.
The mechanism by which Zika virus disrupts neuronal development and results in congenital Zika syndrome was unknown. Because of similarities between Zika syndrome and a recognized congenital genetic disease (Kinesin-5) known to cause microcephaly and retinopathies in developing infants, the research team studied both, looking for similarities. They discovered a direct link, the first molecular and cellular evidence supporting a direct connection between the two.
"We had a hunch that the microcephaly and blindness that results from Kinesin-5 genetic disease could be linked to Zika infection, and the hunch paid off," notes Dr. Wojcik. "Our experiments identify a molecular motor as a target for degradation by an encoded Zika virus protein (Zika protease). The molecular motor is Kinesin-5, and it is required for cell division in humans. Our data identify Kinesin-5 as a target for the virus and links the infection to microcephaly."
The researchers observed that Zika protease cuts Kinesin-5 during cell division, disrupting the process and causing a loss of function. They also suggest a way to prevent it.
The Zika protease can degrade only a target protein it can reach. Since the protease is part of the endoplasmic reticulum (ER) membrane, only target proteins that come in direct contact with the ER can be degraded. In this way, the protease acts in a spatially restricted manner in the cell; target proteins are degraded only in certain regions of the cell volume and not in others. So, the research team proposes a drug that would affect only the Zika protease instead of drugs that would affect all target proteins in a cell.
"We predict and hope that potential drugs that inhibit Zika protease may be effective in preventing microcephaly and blindness from developing within Zika-infected fetuses," Dr. Wojcik concludes.
###
The LSU Health New Orleans research team also included Liqiong Liu, Micquel Downs, and Jesse Guidry.
The research was supported by funds from the National Institute of General Medical Sciences of the National Institutes of Health and LSU Health Sciences Center New Orleans.
LSU Health Sciences Center New Orleans educates Louisiana's health care professionals. The state's flagship health sciences university, LSU Health New Orleans includes a School of Medicine with branch campuses in Baton Rouge and Lafayette, the state's only School of Dentistry, Louisiana's only public School of Public Health, and Schools of Allied Health Professions, Nursing, and Graduate Studies. LSU Health New Orleans faculty take care of patients in public and private hospitals and clinics throughout the region. In the vanguard of biosciences research in a number of areas in a worldwide arena, the LSU Health New Orleans research enterprise generates jobs and enormous economic impact. LSU Health New Orleans faculty have made lifesaving discoveries and continue to work to prevent, advance treatment, or cure disease. To learn more, visit http://www.lsuhsc.edu, http://www.twitter.com/LSUHealthNO, or http://www.facebook.com/LSUHSC.
Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system
Competing for high status speeds up aging in male baboons
Study suggests that high social status contributes to accelerated aging in baboons, despite its other advantages
IMAGE: MALE BABOONS IN AMBOSELI NATIONAL PARK, KENYA, ENGAGE IN PHYSICAL COMPETITION FOR HIGH RANK, DEMONSTRATING THE POTENTIAL COSTS OF ATTAINING HIGH STATUS view more
CREDIT: BETH ARCHIE (CC BY 4.0)
Battling other male baboons to achieve high social status comes with physiological costs that accelerate aging, according to study published today in eLife.
The findings suggest that current life circumstances may be more important contributors to premature aging than early life hardship, at least in baboons.
Chemical changes to DNA, also called epigenetic changes, can be used as a kind of 'clock' to measure aging. While these epigenetic changes usually correspond with age, they can also be used to detect signs of premature aging.
"Environmental stressors can make the clock tick faster, so that some individuals appear biologically older than their actual age and experience a higher risk of age-related disease," explains co-first author Jordan Anderson, a PhD student in Evolutionary Anthropology at Duke University, Durham, North Carolina, US. "We sought to answer what social or early life experiences contribute to accelerated aging in baboons."
The team measured aging in 245 wild baboons from a well-studied population in Kenya using the epigenetic clock and other methods. They found that the epigenetic clock was a good predictor of chronological age overall. But contrary to what they expected, early life adversity was not a good predictor of accelerated aging in the animals.
Instead, they found that the highest-ranking males showed signs of accelerated aging. Higher body mass index, which is associated with having more lean muscle mass in baboons, was also associated with accelerated aging, likely because of the physical demands of maintaining high status. The team was also able to show that the epigenetic clock sped up as the animals climbed the social ladder and slowed down as they moved down it.
"Our results argue that achieving high rank for male baboons - the best predictor of reproductive success in these animals - imposes costs that are consistent with a 'live fast, die young,' life history strategy," says co-first author Rachel Johnston, Postdoctoral Associate in Evolutionary Anthropology at Duke University.
"While the findings reveal how social pressures can influence aging for males, we don't see the same effect of rank in female baboons, who are born into their social rank rather than having to fight for it," adds senior author Jenny Tung, Associate Professor in the Departments of Evolutionary Anthropology and Biology at Duke University, and a Faculty Associate of the Duke University Population Research Institute.
"Our results have important implications for research on the social determinants of health in humans and other animals because they show that 'high status' can mean very different things in different contexts. They also highlight the importance of examining the effects of both early life and current life environments on biological aging," Tung concludes.
eLife is a non-profit organisation created by funders and led by researchers. Our mission is to accelerate discovery by operating a platform for research communication that encourages and recognises the most responsible behaviours. We aim to publish work of the highest standards and importance in all areas of biology and medicine, including Evolutionary Biology, and Genetics and Genomics, while exploring creative new ways to improve how research is assessed and published. eLife receives financial support and strategic guidance from the Howard Hughes Medical Institute, the Knut and Alice Wallenberg Foundation, the Max Planck Society and Wellcome. Learn more at https://elifesciences.org/about.
Teeth-baring, glaring confrontations are a normal part of being the boss male baboon. A new study shows that the guys at the top will age faster as a result of constantly having to defend their higher status.
CREDIT
Courtney L. Fitzpatrick
Being top baboon costs males their longevity
Struggle for dominance leaves a mark on genes and speeds up aging
IMAGE: MALE BABOONS MAINTAIN THEIR PECKING ORDER IN THE TROOP WITH PHYSICAL DISPLAYS OF AGGRESSION. A NEW STUDY SHOWS THAT THE GUYS AT THE TOP WILL AGE FASTER AS A RESULT... view more
CREDIT: ELIZABETH ARCHIE
DURHAM, N.C. -- Some guys have it all: the muscle, the power, the high social status, the accelerated aging.
But wait. Faster aging? Who wants that? For male baboons, it's the price they pay to be at the top.
New research appearing April 6 in eLife by Jenny Tung, associate professor of evolutionary anthropology and biology at Duke University, and her colleagues shows that male baboons that climb the social ladder age faster than males with lower social standing. If a male drops in social status, his estimated rate of aging drops as well.
Using blood samples from 245 wild baboons in the Amboseli ecosystem in Kenya, the team analyzed chemical modifications to DNA known as DNA methylation marks.
"These marks change with age in a clock-like fashion," Tung said. "However, environmental stressors can make the clock tick faster." This would make an individual appear older than they really are, and, research in humans suggests, can put them at a higher risk of aging-related disease.
Since this cohort of baboons is one of the most intensively studied wild mammal populations in the world, the researchers already knew not only each baboon's age, but also the environment in which they grew up, their exposure to early life adversity, and a great deal about their adult environment, especially the aspects that predict how long they live and how many offspring they leave behind.
"We used DNA methylation to compare the baboons known ages to their 'biological ages,'" said Jordan Anderson, a graduate student in the Tung lab who co-led the work. These methylation markers are found across the genome, so the team first needed to measure a large number of these sites - about 400,000 of them - and then, through statistical methods and models, whittle the number of sites down to about 500 that best predicted age.
Interestingly, for males, early life adversity didn't affect how fast their biological clocks tick.
Adult social status was the strongest factor that affected aging. "Male baboons who compete successfully for high social status appear to age faster," Tung said. "We repeatedly sampled some of these males and were able to show that the clock can speed up or slow down as males move up or down the social ladder."
This is contrary to what we see in humans. Typically, high social status in humans predicts better health, not worse. The most wealthy and powerful humans have access to and can afford the best houses, schools, healthcare and more. Those who live in poverty and have lower socioeconomic status are at increased risk and have higher rates of disease, cancer and all-cause mortality.
Male baboons, though, have to fight for their social status. Because of this, it's common to see male-male competition on a regular basis, where baboon observers can see a clear winner and a clear loser.
To maintain their social status, males at the top regularly have to hold their ground and defend themselves physically. Because of this, male baboons at the top tend to have more muscle mass and better body condition than lower ranking baboons. But as their physicality starts to diminish with age, a new, younger, stronger male may overcome them for the top spot.
High ranking males also spend a lot of time mate-guarding females. Around ovulation, they follow females closely and ward off other males. Mate-guarding constrains a male's other activities, and Tung and her team think it is likely to be energetically costly -- perhaps helping to explain their accelerated aging result.
So why do these males work so hard to achieve a high stress social status? It's simple: to have offspring.
"If male baboons are going to have babies, they need to achieve high rank," Tung said. "They will have very little chance to leave offspring if they don't achieve high rank, which creates a powerful evolutionary motivation."
This study highlights one way that the social environment can influence aging. "Our research shows that the manner in which social status is attained and maintained is crucial to understanding its consequences," Tung said.
###
This research was supported by the US National Science Foundation and the US National Institutes of Health, Canadian Institute of Advanced Research, North Carolina Biotechnology Center, and the Center for Population Health and Aging. (2018264636, IOS1456832, R01AG053308, R01AG053330, R01HD088558, P01AG031719, F32HD095616, 2016-IDG-1013, P30AG034424)
CITATION: "High Social Status Males Experience Accelerated Epigenetic Aging in Wild Baboons," Jordan A. Anderson, Rachel A. Johnston, Amanda J. Lea, Fernando A. Campos, Tawni N. Voyles, Mercy Y. Akinyi, Susan C. Alberts, Elizabeth A. Archie, Jenny Tung. eLife, April 6, 2021. DOI: 10.7554/eLife.661
Deep learning networks prefer the human voice -- just like us
Columbia engineers demonstrate that AI systems might reach higher performance if programmed with sound files of human language rather than with binary data labels
COLUMBIA UNIVERSITY SCHOOL OF ENGINEERING AND APPLIED SCIENCE
IMAGE: A DEEP NEURAL NETWORK THAT IS TAUGHT TO SPEAK OUT THE ANSWER DEMONSTRATES HIGHER PERFORMANCES OF LEARNING ROBUST AND EFFICIENT FEATURES. THIS STUDY OPENS UP NEW RESEARCH QUESTIONS ON THE... view more
New York, NY--April 6, 2021--The digital revolution is built on a foundation of invisible 1s and 0s called bits. As decades pass, and more and more of the world's information and knowledge morph into streams of 1s and 0s, the notion that computers prefer to "speak" in binary numbers is rarely questioned. According to new research from Columbia Engineering, this could be about to change.
A new study from Mechanical Engineering Professor Hod Lipson and his PhD student Boyuan Chen proves that artificial intelligence systems might actually reach higher levels of performance if they are programmed with sound files of human language rather than with numerical data labels. The researchers discovered that in a side-by-side comparison, a neural network whose "training labels" consisted of sound files reached higher levels of performance in identifying objects in images, compared to another network that had been programmed in a more traditional manner, using simple binary inputs.
"To understand why this finding is significant," said Lipson, James and Sally Scapa Professor of Innovation and a member of Columbia's Data Science Institute, "It's useful to understand how neural networks are usually programmed, and why using the sound of the human voice is a radical experiment."
When used to convey information, the language of binary numbers is compact and precise. In contrast, spoken human language is more tonal and analog, and, when captured in a digital file, non-binary. Because numbers are such an efficient way to digitize data, programmers rarely deviate from a numbers-driven process when they develop a neural network.
Lipson, a highly regarded roboticist, and Chen, a former concert pianist, had a hunch that neural networks might not be reaching their full potential. They speculated that neural networks might learn faster and better if the systems were "trained" to recognize animals, for instance, by using the power of one of the world's most highly evolved sounds--the human voice uttering specific words.
One of the more common exercises AI researchers use to test out the merits of a new machine learning technique is to train a neural network to recognize specific objects and animals in a collection of different photographs. To check their hypothesis, Chen, Lipson and two students, Yu Li and Sunand Raghupathi, set up a controlled experiment. They created two new neural networks with the goal of training both of them to recognize 10 different types of objects in a collection of 50,000 photographs known as "training images."
One AI system was trained the traditional way, by uploading a giant data table containing thousands of rows, each row corresponding to a single training photo. The first column was an image file containing a photo of a particular object or animal; the next 10 columns corresponded to 10 possible object types: cats, dogs, airplanes, etc. A "1" in any column indicates the correct answer, and nine 0s indicate the incorrect answers.
The team set up the experimental neural network in a radically novel way. They fed it a data table whose rows contained a photograph of an animal or object, and the second column contained an audio file of a recorded human voice actually voicing the word for the depicted animal or object out loud. There were no 1s and 0s.
Once both neural networks were ready, Chen, Li, and Raghupathi trained both AI systems for a total of 15 hours and then compared their respective performance. When presented with an image, the original network spat out the answer as a series of ten 1s and 0s--just as it was trained to do. The experimental neural network, however, produced a clearly discernible voice trying to "say" what the object in the image was. Initially the sound was just a garble. Sometimes it was a confusion of multiple categories, like "cog" for cat and dog. Eventually, the voice was mostly correct, albeit with an eerie alien tone (see example on website).
At first, the researchers were somewhat surprised to discover that their hunch had been correct--there was no apparent advantage to 1s and 0s. Both the control neural network and the experimental one performed equally well, correctly identifying the animal or object depicted in a photograph about 92% of the time. To double-check their results, the researchers ran the experiment again and got the same outcome.
What they discovered next, however, was even more surprising. To further explore the limits of using sound as a training tool, the researchers set up another side-by-side comparison, this time using far fewer photographs during the training process. While the first round of training involved feeding both neural networks data tables containing 50,000 training images, both systems in the second experiment were fed far fewer training photographs, just 2,500 apiece.
It is well known in AI research that most neural networks perform poorly when training data is sparse, and in this experiment, the traditional, numerically trained network was no exception. Its ability to identify individual animals that appeared in the photographs plummeted to about 35% accuracy. In contrast, although the experimental neural network was also trained with the same number of photographs, its performance did twice as well, dropping only to 70% accuracy.
Intrigued, Lipson and his students decided to test their voice-driven training method on another classic AI image recognition challenge, that of image ambiguity. This time they set up yet another side-by-side comparison but raised the game a notch by using more difficult photographs that were harder for an AI system to "understand." For example, one training photo depicted a slightly corrupted image of a dog, or a cat with odd colors. When they compared results, even with more challenging photographs, the voice-trained neural network was still correct about 50% of the time, outperforming the numerically-trained network that floundered, achieving only 20% accuracy.
Ironically, the fact their results went directly against the status quo became challenging when the researchers first tried to share their findings with their colleagues in computer science. "Our findings run directly counter to how many experts have been trained to think about computers and numbers; it's a common assumption that binary inputs are a more efficient way to convey information to a machine than audio streams of similar information 'richness,'" explained Boyuan Chen, the lead researcher on the study. "In fact, when we submitted this research to a big AI conference, one anonymous reviewer rejected our paper simply because they felt our results were just 'too surprising and un-intuitive.'"
When considered in the broader context of information theory however, Lipson and Chen's hypothesis actually supports a much older, landmark hypothesis first proposed by the legendary Claude Shannon, the father of information theory. According to Shannon's theory, the most effective communication "signals" are characterized by an optimal number of bits, paired with an optimal amount of useful information, or "surprise."
"If you think about the fact that human language has been going through an optimization process for tens of thousands of years, then it makes perfect sense, that our spoken words have found a good balance between noise and signal;" Lipson observed. "Therefore, when viewed through the lens of Shannon Entropy, it makes sense that a neural network trained with human language would outperform a neural network trained by simple 1s and 0s."
The study, to be presented at the International Conference on Learning Representations conference on May 3, 2021, is part of a broader effort at Lipson's Columbia Creative Machines Lab to create robots that can understand the world around them by interacting with other machines and humans, rather than by being programed directly with carefully preprocessed data.
"We should think about using novel and better ways to train AI systems instead of collecting larger datasets," said Chen. "If we rethink how we present training data to the machine, we could do a better job as teachers."
One of the more refreshing results of computer science research on artificial intelligence has been an unexpected side effect: by probing how machines learn, sometimes researchers stumble upon fresh insight into the grand challenges of other, well-established fields.
"One of the biggest mysteries of human evolution is how our ancestors acquired language, and how children learn to speak so effortlessly," Lipson said. "If human toddlers learn best with repetitive spoken instruction, then perhaps AI systems can, too."
###
About the Study
The study is titled "BEYOND CATEGORICAL LABEL REPRESENTATIONS FOR IMAGE CLASSIFICATION"
Authors are: Boyuan Chen, Yu Li, Sunand Raghupathi, Hod Lipson, Mechanical Engineering and Computer Science, Columbia Engineering.
The study was supported by NSF NRI 1925157 and DARPA MTO grant L2M Program HR0011-18-2-0020.
The authors declare no financial or other conflicts of interest.
Columbia Engineering, based in New York City, is one of the top engineering schools in the U.S. and one of the oldest in the nation. Also known as The Fu Foundation School of Engineering and Applied Science, the School expands knowledge and advances technology through the pioneering research of its more than 220 faculty, while educating undergraduate and graduate students in a collaborative environment to become leaders informed by a firm foundation in engineering. The School's faculty are at the center of the University's cross-disciplinary research, contributing to the Data Science Institute, Earth Institute, Zuckerman Mind Brain Behavior Institute, Precision Medicine Initiative, and the Columbia Nano Initiative. Guided by its strategic vision, "Columbia Engineering for Humanity," the School aims to translate ideas into innovations that foster a sustainable, healthy, secure, connected, and creative humanity.
