More stories

  • in

    Common approach to demystify black box AI not ready for prime time

    Artificial intelligence models that interpret medical images hold the promise to enhance clinicians’ ability to make accurate and timely diagnoses, while also lessening workload by allowing busy physicians to focus on critical cases and delegate rote tasks to AI.
    But AI models that lack transparency about how and why a diagnosis is made can be problematic. This opaque reasoning — also known “black box” AI — can diminish clinician trust in the reliability of the AI tool and thus discourage its use. This lack of transparency could also mislead clinicians into over-trusting the tool’s interpretation.
    In the realm of medical imaging, one way to create more understandable AI models and to demystify AI decision-making have been saliency assessments — an approach that uses heat maps to pinpoint whether the tool is correctly focusing only on the relevant pieces of a given image or homing in on irrelevant parts of it.
    Heat maps work by highlighting areas on an image that influenced the AI model’s interpretation. This could help human physicians see whether the AI model focuses on the same areas as they do or is mistakenly focusing on irrelevant spots on an image.
    But a new study, published in Nature Machine Intelligence on Oct. 10, shows that for all their promise, saliency heat maps may not be yet ready for prime time.
    The analysis, led by Harvard Medical School investigator Pranav Rajpurkar, Matthew Lungren of Stanford, and Adriel Saporta of New York University, quantified the validity of seven widely used saliency methods to determine how reliably and accurately they could identify pathologies associated with 10 conditions commonly diagnosed on X-ray, such as lung lesions, pleural effusion, edema, or enlarged heart structures. To ascertain performance, the researchers compared the tools’ performance against human expert judgment. More

  • in

    Claims AI can boost workplace diversity are 'spurious and dangerous'

    New research highlights a growing market in AI-powered recruitment tools, used to process high volumes of job applicants, that claim to bypass human bias and remove discrimination from hiring. •These AI tools reduce race and gender to trivial data points, and often rely on personality analysis that is “automated pseudoscience,” according to Cambridge researchers. Academics have also teamed up with computing students to debunk use of AI in recruitment by building a version of the kinds of software increasingly used by HR teams. It demonstrates how random changes in clothing or lighting give radically different personality readings that could prove make-or-break for a generation of job seekers.
    Recent years have seen the emergence of AI tools marketed as an answer to lack of diversity in the workforce, from use of chatbots and CV scrapers to line up prospective candidates, through to analysis software for video interviews.
    Those behind the technology claim it cancels out human biases against gender and ethnicity during recruitment, instead using algorithms that read vocabulary, speech patterns and even facial micro-expressions to assess huge pools of job applicants for the right personality type and “culture fit.”
    However, in a new report published in Philosophy and Technology, researchers from Cambridge’s Centre for Gender Studies argue these claims make some uses of AI in hiring little better than an “automated pseudoscience” reminiscent of physiognomy or phrenology: the discredited beliefs that personality can be deduced from facial features and skull shape.
    They say it is a dangerous example of “technosolutionism”: turning to technology to provide quick fixes for deep-rooted discrimination issues that require investment and changes to company culture.
    In fact, the researchers have worked with a team of Cambridge computer science undergraduates to debunk these new hiring techniques by building an AI tool modelled on the technology, available at: https://personal-ambiguator-frontend.vercel.app/
    The ‘Personality Machine’ demonstrates how arbitrary changes in facial expression, clothing, lighting and background can give radically different personality readings — and so could make the difference between rejection and progression for a generation of job seekers vying for graduate positions. More

  • in

    Self-teaching AI uses pathology images to find similar cases, diagnose rare diseases

    Rare diseases are often difficult to diagnose and predicting the best course of treatment can be challenging for clinicians. Investigators from the Mahmood Lab at Brigham and Women’s Hospital, a founding member of the Mass General Brigham healthcare system, have developed a deep learning algorithm that can teach itself to learn features which can then be used to find similar cases in large pathology image repositories. Known as SISH (Self-Supervised Image search for Histology), the new tool acts like a search engine for pathology images and has many potential applications, including identifying rare diseases and helping clinicians determine which patients are likely to respond to similar therapies. A paper introducing the self-teaching algorithm is published in Nature Biomedical Engineering.
    “We show that our system can assist with the diagnosis of rare diseases and find cases with similar morphologic patterns without the need for manual annotations, and large datasets for supervised training,” said senior author Faisal Mahmood, PhD, in the Brigham’s Department of Pathology. “This system has the potential to improve pathology training, disease subtyping, tumor identification, and rare morphology identification.”
    Modern electronic databases can store an immense amount of digital records and reference images, particularly in pathology through whole slide images (WSIs). However, the gigapixel size of each individual WSI and the ever-increasing number of images in large repositories, means that search and retrieval of WSIs can be slow and complicated. As a result, scalability remains a pertinent roadblock for efficient use.
    To solve this issue, researchers at the Brigham developed SISH, which teaches itself to learn feature representations which can be used to find cases with analogous features in pathology at a constant speed regardless of the size of the database.
    In their study, the researchers tested the speed and ability of SISH to retrieve interpretable disease subtype information for common and rare cancers. The algorithm successfully retrieved images with speed and accuracy from a database of tens of thousands of whole slide images from over 22,000 patient cases, with over 50 different disease types and over a dozen anatomical sites. The speed of retrieval outperformed other methods in many scenarios, including disease subtype retrieval, particularly as the image database size scaled into the thousands of images. Even while the repositories expanded in size, SISH was still able to maintain a constant search speed.
    The algorithm, however, has some limitations including a large memory requirement, limited context awareness within large tissue slides and the fact that it is limited to a single imaging modality.
    Overall, the algorithm demonstrated the ability to efficiently retrieve images independent of repository size and in diverse datasets. It also demonstrated proficiency in diagnosis of rare disease types and the ability to serve as a search engine to recognize certain regions of images that may be relevant for diagnosis. This work may greatly inform future disease diagnosis, prognosis, and analysis.
    “As the sizes of image databases continue to grow, we hope that SISH will be useful in making identification of diseases easier,” said Mahmood. “We believe one important future direction in this area is multimodal case retrieval which involves jointly using pathology, radiology, genomic and electronic medical record data to find similar patient cases.”
    Story Source:
    Materials provided by Brigham and Women’s Hospital. Note: Content may be edited for style and length. More

  • in

    Novel navigation strategies for microscopic swimmers

    Autonomous optimal navigation of microswimmers is in fact possible, as researchers from the Max Planck Institute for Dynamics and Self-Organization (MPI-DS) showed. In contrast to the targeted navigation of boats, the motion of swimmers at the microscale is strongly disturbed by fluctuations. The researchers now described a navigation strategy for microswimmers that does not need an external interpreter. Their findings may contribute to the understanding of transport mechanisms in the microcosm as well as to applications such as targeted drug delivery.
    Whereas the shortest way between two points is a straight connection, it might not be the most efficient path to follow. Complex currents often affect the motion of microswimmers and make it difficult for them to reach their destination. At the same time, making use of these currents to navigate as fast as possible is a certain evolutionary advantage. Whereas such strategies allow biological microswimmers to better access food or escape a predator, microrobots could this way be directed to perform specific tasks.
    The optimal path in a given current can readily be determined mathematically, yet fluctuations perturb the motion of microswimmers and deviate them from the optimal route. Thus, they have to readjust their motion in order to account for environmental changes. This typically requires the help of an external interpreter and takes away their autonomy.
    “Thanks to evolution, some microorganisms have developed autonomous strategies that enable directed motion towards larger concentration of nutrients or light,” first author of the study Lorenzo Piro explains. Inspired by this idea, the researchers from the Department of Living Matter Physics at the MPI-DS designed strategies that allow microswimmers to navigate optimally in a nearly autonomous way.
    Light as a guide for autonomous navigation
    When an external interpreter defines the navigation pattern, microswimmers on average follow a well-defined path. Thus, it is a good approach to guide the microswimmer along that path within the current. This can be achieved autonomously via external stimuli, despite the presence of fluctuations. This principle could be applied to swimmers that respond to variation of light, such as certain algae, in which case the optimal path can simply be illuminated. Remarkably, the resulting performances are comparable to externally supervised navigation. “These new strategies can moreover conveniently be applied to more complex scenario such as navigation on curved surfaces or in presence of random currents,” concludes Ramin Golestanian, director at MPI-DS.
    Possible applications of the study thus range from targeted drug delivery at the microscale to the optimal design of autonomous micromachines.
    Story Source:
    Materials provided by Max Planck Institute for Dynamics and Self-Organization. Note: Content may be edited for style and length. More

  • in

    Optical foundations illuminated by quantum light

    Optics, the study of light, is one of the oldest fields in physics and has never ceased to surprise researchers. Although the classical description of light as a wave phenomenon is rarely questioned, the physical origins of some optical effects are. A team of researchers at Tampere University have brought the discussion on one fundamental wave effect, i.e., the debate around the anomalous behaviour of focused light waves, to the quantum domain.
    The researchers have been able to show that quantum waves behave significantly differently from their classical counterparts and can be used to increase the precision of distance measurements. Their findings also add to the discussion on physical origin of the anomalous focusing behaviour. The results are now published in the journal of Nature Photonics.
    “Interestingly, we started with an idea based on our earlier results and set out to structure quantum light for enhanced measurement precision. However, we then realised that the underlying physics of this application also contributes to the long debate about the origins of the Gouy phase anomaly of focused light fields.,” explains Robert Fickler, group leader of the Experimental Quantum Optics group at Tampere University.
    Quantum waves behave differently but point to the same origin
    Over the last decades, methods for structuring light fields down on the single photon level have vastly matured and led to a myriad of novel findings. In addition, a better of optics’ foundations has been achieved. However, the physical origin of why light behaves in such an unexpected way when going through a focus, the so-called Gouy phase anomaly, is still often debated. This is despite its widespread use and importance in optical systems. The novelty of the current study is now to put the effect into the quantum domain.
    “When developing the theory to describe our experimental results, we realised (after a long debate) that the Gouy phase for quantum light is not only different than the standard one, but its origin can be linked to another quantum effect. This is just like what was speculated in an earlier work,” adds Doctoral researcher Markus Hiekkamäki, leading author of the study.
    In the quantum domain, the anomalous behaviour is sped up when compared to classical light. As the Gouy phase behaviour can be used to determine the distance a beam of light has propagated, the speed up of the quantum Gouy phase could allow for an improvement in the precision of measuring distances.
    With this new understanding at hand, the researchers are planning to develop novel techniques to enhance their measurement abilities such that it will be possible to measure more complex beams of structured photons. The team expects that this will help them push forward the application of the observed effect, and potentially bring to light more differences between quantum and classical light fields.
    Story Source:
    Materials provided by Tampere University. Note: Content may be edited for style and length. More

  • in

    Sleep mode makes Energy Internet more energy efficient

    A group of scientists in Nagoya University, Japan, have developed a possible solution to one of the biggest problems of the Internet of Energy, energy efficiency. They did so by creating a controller that has a sleep mode and only procures energy when needed.
    Widespread generation of electricity based on renewable energy has become necessary to combat the climate crisis. One solution to realize society’s electrification needs is the Internet of Energy, which would operate like the information Internet, except that it would consist of energy linked by smart power generation, smart power consumption, smart interconnection, and cloud sharing.
    When information is sent over the Internet, it is divided into transmittable units called ‘packets’, which are tagged with their destination. The energy Internet is based on a similar concept. Information tags are added to power pulses to create units called ‘power packets’. On the basis of requests from terminals, these are then distributed over networks to where they are needed. However, one problem is that since the packets are sent sporadically, the energy supply is intermittent. Current solutions, such as storage batteries or capacitors, complicate the system and reduce its efficiency.
    An alternative solution is what is known as ‘sparse control’, where the terminal’s actuators are active part of the time and are in sleep mode for the rest of the time. In sleep mode, they do not consume fuel or electricity, leading to efficient energy saving and reducing environmental and noise pollution. Although sparse control has been used with a single actuator, it does not necessarily provide good performance when multiple actuators are used. The problem of determining how to do this for multiple actuators is called the ‘maximum turn-off control problem’.
    Now, a Nagoya University research group, led by Professor Shun-ichi Azuma and Doctoral student Takumi Iwata of the Graduate School of Engineering, has developed a model control scheme for multiple actuators. The model has an awake mode, during which it procures and controls the necessary power packets for when they are needed, and a sleep mode. The research was published in the International Journal of Robust and Nonlinear Control.
    “We can see our research being useful in the motor control of production equipment,” explains Professor Azuma. “This research provides a control system configuration method based on the assumption that the energy supply is intermittent. It has the advantage of eliminating the need for storage batteries and capacitors. It is expected to accelerate the practical application of the power packet type energy Internet.”
    This research was supported by Japan Science and Technology Agency Emergent Research Support Program and Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan.
    Story Source:
    Materials provided by Nagoya University. Note: Content may be edited for style and length. More

  • in

    Tree-climbing carnivores called fishers are back in Washington’s forests

    Holding an antenna above his head, Jeff Lewis crept through an evergreen forest in the Cascade mountains, southeast of Seattle. As he navigated fallen fir logs and dripping ferns, he heard it: a faint “beep” from a radio transmitter implanted in an animal code-named F023.

    F023 is a fisher (Pekania pennanti), an elusive member of the weasel family that Lewis fondly describes as a “tree wolverine.” Resembling a cross between a cat and an otter, these sleek carnivores hunt in forests in Canada and parts of the northern United States. But fur trapping and habitat loss had wiped out Washington’s population by the mid-1900s.

    Back in 2017 when Lewis was keeping tabs on F023, he tracked her radio signal from a plane two or three times a month, along with dozens of other recently released fishers. Come spring, he noticed that F023’s behavior was different from the others.

    Her locations had been clustered close together for a few weeks, a sign that she might be “busy with babies,” says Lewis, a conservation biologist with the Washington Department of Fish and Wildlife. He and colleagues trekked into the woods to see if she had indeed given birth. If so, it would be the first wild-born fisher documented in the Cascades in at least half a century.

    As the faint beeps grew louder, the biologists found a clump of fur snagged on a branch, scratch marks in the bark and — the best clue of all — fisher scat. The team rigged motion-detecting cameras to surrounding trees. A few days later, after sifting through hundreds of images of squirrels and deer, the team hit the jackpot: a grainy photo of F023 ferrying a kit down from her den high in a hemlock tree. The scientists were ecstatic.

    “We’re all a bunch of little kids when it comes to getting photos like that,” Lewis says.

    Sign Up For the Latest from Science News

    Headlines and summaries of the latest Science News articles, delivered to your inbox

    Thank you for signing up!

    There was a problem signing you up.

    Chasing babies

    This notable birth came during the second phase of a 14-year fisher reintroduction effort. After 90 fishers were released in Olympic National Park from 2008 to 2010, the project turned its focus east of Seattle, relocating 81 fishers in the South Cascades (home to Mount Rainier National Park) from 2015 to 2020, and then 89 fishers in the North Cascades from 2018 to 2020. The animals were brought in from British Columbia and Alberta. The project concluded last year, when researchers let loose the final batch of fishers.

    Baby animals are the key measure of success for a wildlife reintroduction project. As part of Washington’s Fisher Recovery Plan, biologists set out to document newborn kits as an indicator of how fishers were faring in the three relocation regions.

    Before F023’s kit was caught on camera in May 2017, biologists had already confirmed births by seven relocated females on the Olympic Peninsula, where the whole project began. Two of the seven females had four kits, “the largest litter size ever documented on the West Coast,” says Patti Happe, wildlife branch chief at Olympic National Park. Most females have one to three kits.

    Lewis is often asked, why put all of this effort into restoring a critter many people have never heard of? His answer: A full array of carnivores makes the ecosystem more resilient.

    Happe admits to another motive: “They’re freaking adorable — that’s partly why we’re saving them.”

    This agile member of the weasel family is a fearsome predator. Fishers are one of the few carnivores that can hunt and kill quill-covered porcupines.EMILY BROUWER/NPS (CC BY 2.0)

    The missing piece

    Contrary to their name, fishers don’t hunt fish, though they’ll happily munch on a dead one if it’s handy. They mainly prey on small mammals, but they also eat reptiles, amphibians, insects, fruit and carrion. About a meter long, males weigh up to six kilograms, about twice as much as females. Fun facts: Females raise young high above the forest floor in hollowed-out spaces in tree trunks. Fishers can travel face-first down tree trunks by turning their hind feet 180 degrees. They have wickedly sharp teeth and partially retractable claws. And they’re incredibly agile, leaping up to two meters between branches and traveling as much as 30 kilometers in a day.

    Fishers’ stubby legs and unique climbing skills make them a threat to tree-climbing porcupines. It isn’t pretty: A fisher will force the quill-covered animal down a tree and attack its face until it dies from blood loss or shock. Then the fisher neatly skins the prickly prey, eating most everything except the quills and bones.

    These camera trap photos, taken in April 2021, show female fisher F105 carrying one of her four kits down from her tree den near Lake Wenatchee in the North Cascades.NPS

    But these fearsome predators were no match for humans. In the 1800s, trappers began targeting fishers for their fur. Soft and luxuriant, the glossy brown-gold pelts were coveted fashion accessories, selling for as much as $345 each in the 1920s. This demand meant fishers disappeared not only from Washington, but from more than a dozen states across the northern United States. Once fisher populations plummeted, porcupines ran rampant across the Great Lakes region and New England. This wreaked havoc on forests because the porcupines gobbled up tree seedlings.

    Hoping to keep porcupine populations in check, private timber companies partnered with state agencies to bring fishers back to several states in the 1950s and 1960s. Thanks to these efforts and stricter trapping regulations, fishers are once again abundant in Michigan, Wisconsin, New York and Massachusetts.

    But in Washington, like most of the West, fisher numbers were still slim. By the turn of the 21st century, no fisher had been sighted in the state for over three decades.

    As in the Midwest and New England, private timber companies in Washington supported bringing back fishers. Although porcupines are uncommon in Washington, mountain beavers — a large, primitive rodent endemic to the Pacific Northwest — fill a similar role in Washington’s evergreen forests: They eat tree seedlings. And fishers eat them.

    By 2006, the state hatched a plan to bring the animals in from Canada. “It was a big opportunity to restore a species,” Lewis says. “We can fix this.”

    [embedded content]
    This 2009 camera trap vídeo from Olympic Peninsula shows fisher F007 scaling a cedar tree and carrying her four kits to the forest floor, one at a time

    A new home

    Like the other Canadian fishers moved to Washington, F023’s relocation story began when she walked into a box trap in British Columbia, lured by a tasty morsel of meat. The bait had been set by local trappers hired by Conservation Northwest, a nonprofit that is one of the recovery project’s three main partners, along with Washington Fish and Wildlife and the National Park Service. After veterinarians checked her health and administered vaccines and antiparasitics to help her survive in her new home, F023 received a surgically implanted radio transmitter and was driven across the border.

    She was met by members of the fisher recovery team, who released her just south of Mount Rainier National Park. The forest’s towering Douglas fir, western red cedar and western hemlock trees were full of cubby holes and cavities to hide in, and the undergrowth held plenty of small mammals to eat. At the release, upward of 150 people gathered around F023’s box, part of the team’s effort to engage the public in championing fisher recovery. Everyone cheered as a child opened the door and the furry female bounded into the snowy woods, out of sight in a flash.

    The team monitored each relocated fisher for up to two years to see if the project met key benchmarks of success in each of the three regions: more than 50 percent of the fishers surviving their first year, at least half establishing a home range near the release site, and a confirmed kit born to at least one female.

    “We met those marks,” says Dave Werntz, science and conservation director at Conservation Northwest.

    The effort may have been aided by a series of bypasses built over and under a roughly 25-kilometer stretch of Interstate 90 east of Seattle. One of these structures is the largest wildlife bridge in North America, an overpass “paved” with forest. In 2020, a remote camera caught an image of what looks like a fisher moving through one of the underpasses.

    Speeding vehicles on busy highways pose a threat to fishers and other migrating wildlife. This new bridge east of Seattle is “paved” with trees and plants to let animals safely cross I-90 to find habitat, food or mates on the other side.WASHINGTON STATE DEPT. OF TRANSPORTATION

    “Male fishers go on these huge walkabouts to find females,” Werntz says. While biologists assumed fishers would cross the freeway to search for mates, having photographic proof “is pretty wonderful,” he says.

    Happe and others hope to also see wildlife crossings along Interstate 5 one day. The freeway, which runs north-south near the coast, is the main obstacle keeping the Olympic and Cascade populations apart, she says. “We’re all working on wildlife travel corridors and connectivity in hopes the two populations hook up.”

    Learning curve

    The majority of the initial 90 fishers relocated to the Olympic Peninsula settled nicely into their new homes, according to radio tracking. In the year following release in that location, the fisher survival rate averaged 73 percent, but varied based on the year and season they were released, as well as sex and age of the fishers.

    Males fared better than females: Seventy-four percent of recorded deaths were of females, partly because they are smaller and more vulnerable to predators, such as bobcats and coyotes. Of 24 recovered carcasses where cause of death could be determined, 14 were killed by predators, seven were struck by vehicles, two drowned and one died in a leg-hold trap, Lewis, Happe and colleagues reported in the April 2022 Journal of Wildlife Management.

    Because the first fishers relocated to the Olympic Peninsula were released in several locations, the animals had trouble finding mates. As a result, only a few parents sired the subsequent generations.

    The researchers became concerned when they looked at the genetic diversity of fishers on the Olympic Peninsula six years post-relocation. Happe and colleagues set up 788 remote cameras and hair-snare stations: triangular cubbies open on either end with a chicken leg as bait in the middle and wire brushes protruding from either side to grab strands of fur. DNA analysis of the fur raised red flags about inbreeding, Happe and Lewis say.

    “Models showed we were going to lose up to 50 percent of genetic diversity, and the population would wink out in something like 100 years,” Happe says. To expand the gene pool, the team brought 20 more fishers to the Olympic Peninsula in 2021. These animals came from Alberta whereas the founding population had hailed from British Columbia.

    [embedded content]
    Two fishers from Canada are released from wooden crates, quickly disappearing into Olympic National Park in November 2021. Both wear radio tracking devices so that researchers can monitor their well-being.

    As the reintroduction effort moved into the Cascades, the team adapted, based on lessons learned from the Olympic Peninsula. For instance, to increase the likelihood of fishers finding each other more quickly, the animals were released at fewer sites that were closer together. The team also released the animals before January, giving females ample time to settle into a home range before the spring mating and birthing season.

    Finding their food

    As the experiment went on, more unanticipated findings popped up. Fishers released in the southern part of the Cascades were more likely to survive the first year (76 percent) than those relocated north of I-90 (40 percent), according to the final project report, released in June. Remote-camera data suggest that’s because there are less prey and slightly more predators in the North Cascades, says Tanner Humphries, community wildlife monitoring program lead for Conservation Northwest.

    And in both the Cascades and the Olympic Peninsula, fishers are using different types of habitat than biologists had predicted, Happe says. The mammals — once assumed to be old-growth specialists — are using a mosaic of young and old forests. Fishers require large, old trees with cavities for denning and resting. But in younger managed forests where trees are thinned or cut, prey may be easier to come by.

    Live traps in the South Cascades support that idea. Fishers’ preferred prey — snowshoe hares and mountain beavers — were most abundant in young regenerating forests. In older forests, traps detected mainly mice, voles and chipmunks, which are not substantial meals for fishers, Mitchell Parsons, a wildlife ecologist at Utah State University in Logan, reported with Lewis, Werntz and others in 2020 in Forest Ecology and Management.

    North America’s fisher populations are blossoming, helping to rebalance forest ecosystems.Emily Brouwer/NPS (CC BY 2.0)

    The future is re-wild

    After F023’s baby was caught on camera five years ago, the mother’s tracking chip degraded as designed — the hardware lasts less than two years. Since then, many more fisher kits have been born in Washington.

    In fact, these furry carnivores are one of the most successfully translocated mammals in North America. According to Lewis, 41 different translocation efforts across the continent have helped fisher populations blossom. The animals now occupy 68 percent of their historical range, up from 43 percent in the mid-1900s.

    With the last batch of fishers delivered to Washington in 2021, the relocation phase of the project has ended. Lewis, Happe and their partners plan to continue monitoring how these sleek tree-climbing carnivores are faring — and how the ecosystem is responding. For instance, fishers are indeed feasting on seedling-eating mountain beavers, according to research reported by Happe, Lewis and others in 2021 in Northwestern Naturalist.

    Given climate change, species loss and ecosystem degradation, animals worldwide face difficult challenges. The fact that fishers are thriving once again in Washington offers hope, Lewis says.

    “It’s a hard time, it’s a hard world, and this feels like something we’re doing right,” he says. “Instead of losing something, we’re getting it back.” More

  • in

    Superconducting hardware could scale up brain-inspired computing

    Scientists have long looked to the brain as an inspiration for designing computing systems. Some researchers have recently gone even further by making computer hardware with a brainlike structure. These “neuromorphic chips” have already shown great promise, but they have used conventional digital electronics, limiting their complexity and speed. As the chips become larger and more complex, the signals between their individual components become backed up like cars on a gridlocked highway and reduce computation to a crawl.
    Now, a team at the National Institute of Standards and Technology (NIST) has demonstrated a solution to these communication challenges that may someday allow artificial neural systems to operate 100,000 times faster than the human brain.
    The human brain is a network of about 86 billion cells called neurons, each of which can have thousands of connections (known as synapses) with its neighbors. The neurons communicate with each other using short electrical pulses called spikes to create rich, time-varying activity patterns that form the basis of cognition. In neuromorphic chips, electronic components act as artificial neurons, routing spiking signals through a brainlike network.
    Doing away with conventional electronic communication infrastructure, researchers have designed networks with tiny light sources at each neuron that broadcast optical signals to thousands of connections. This scheme can be especially energy-efficient if superconducting devices are used to detect single particles of light known as photons — the smallest possible optical signal that could be used to represent a spike.
    In a new Nature Electronics paper, NIST researchers have achieved for the first time a circuit that behaves much like a biological synapse yet uses just single photons to transmit and receive signals. Such a feat is possible using superconducting single-photon detectors. The computation in the NIST circuit occurs where a single-photon detector meets a superconducting circuit element called a Josephson junction. A Josephson junction is a sandwich of superconducting materials separated by a thin insulating film. If the current through the sandwich exceeds a certain threshold value, the Josephson junction begins to produce small voltage pulses called fluxons. Upon detecting a photon, the single-photon detector pushes the Josephson junction over this threshold and fluxons are accumulated as current in a superconducting loop. Researchers can tune the amount of current added to the loop per photon by applying a bias (an external current source powering the circuits) to one of the junctions. This is called the synaptic weight.
    This behavior is similar to that of biological synapses. The stored current serves as a form of short-term memory, as it provides a record of how many times the neuron produced a spike in the near past. The duration of this memory is set by the time it takes for the electric current to decay in the superconducting loops, which the NIST team demonstrated can vary from hundreds of nanoseconds to milliseconds, and likely beyond. This means the hardware could be matched to problems occurring at many different time scales — from high-speed industrial control systems to more leisurely conversations with humans. The ability to set different weights by changing the bias to the Josephson junctions permits a longer-term memory that can be used to make the networks programmable so that the same network could solve many different problems.
    Synapses are a crucial computational component of the brain, so this demonstration of superconducting single-photon synapses is an important milestone on the path to realizing the team’s full vision of superconducting optoelectronic networks. Yet the pursuit is far from complete. The team’s next milestone will be to combine these synapses with on-chip sources of light to demonstrate full superconducting optoelectronic neurons.
    “We could use what we’ve demonstrated here to solve computational problems, but the scale would be limited,” NIST project leader Jeff Shainline said. “Our next goal is to combine this advance in superconducting electronics with semiconductor light sources. That will allow us to achieve communication between many more elements and solve large, consequential problems.”
    The team has already demonstrated light sources that could be used in a full system, but further work is required to integrate all the components on a single chip. The synapses themselves could be improved by using detector materials that operate at higher temperatures than the present system, and the team is also exploring techniques to implement synaptic weighting in larger-scale neuromorphic chips.
    The work was funded in part by the Defense Advanced Research Projects Agency.
    Story Source:
    Materials provided by National Institute of Standards and Technology (NIST). Note: Content may be edited for style and length. More