More stories

  • in

    Infants outperform AI in 'commonsense psychology'

    Infants outperform artificial intelligence in detecting what motivates other people’s actions, finds a new study by a team of psychology and data science researchers. Its results, which highlight fundamental differences between cognition and computation, point to shortcomings in today’s technologies and where improvements are needed for AI to more fully replicate human behavior.
    “Adults and even infants can easily make reliable inferences about what drives other people’s actions,” explains Moira Dillon, an assistant professor in New York University’s Department of Psychology and the senior author of the paper, which appears in the journal Cognition. “Current AI finds these inferences challenging to make.”
    “The novel idea of putting infants and AI head-to-head on the same tasks is allowing researchers to better describe infants’ intuitive knowledge about other people and suggest ways of integrating that knowledge into AI,” she adds.
    “If AI aims to build flexible, commonsense thinkers like human adults become, then machines should draw upon the same core abilities infants possess in detecting goals and preferences,” says Brenden Lake, an assistant professor in NYU’s Center for Data Science and Department of Psychology and one of the paper’s authors.
    It’s been well-established that infants are fascinated by other people — as evidenced by how long they look at others to observe their actions and to engage with them socially. In addition, previous studies focused on infants’ “commonsense psychology” — their understanding of the intentions, goals, preferences, and rationality underlying others’ actions — have indicated that infants are able to attribute goals to others and expect others to pursue goals rationally and efficiently. The ability to make these predictions is foundational to human social intelligence.
    Conversely, “commonsense AI” — driven by machine-learning algorithms — predicts actions directly. This is why, for example, an ad touting San Francisco as a travel destination pops up on your computer screen after you read a news story on a newly elected city official. However, what AI lacks is flexibility in recognizing different contexts and situations that guide human behavior.
    To develop a foundational understanding of the differences between humans’ and AI’s abilities, the researchers conducted a series of experiments with 11-month-old infants and compared their responses to those yielded by state-of-the-art learning-driven neural-network models.
    To do so, they deployed the previously established “Baby Intuitions Benchmark” (BIB) — six tasks probing commonsense psychology. BIB was designed to allow for testing both infant and machine intelligence, allowing for a comparison of performance between infants and machines and, significantly, providing an empirical foundation for building human-like AI.
    Specifically, infants on Zoom watched a series of videos of simple animated shapes moving around the screen — similar to a video game. The shapes’ actions simulated human behavior and decision-making through the retrieval of objects on the screen and other movements. Similarly, the researchers built and trained learning-driven neural-network models — AI tools that help computers recognize patterns and simulate human intelligence — and tested the models’ responses to the exact same videos.
    Their results showed that infants recognize human-like motivations even in the simplified actions of animated shapes. Infants predict that these actions are driven by hidden but consistent goals — for example, the on-screen retrieval of the same object no matter what location it’s in and the movement of that shape efficiently even when the surrounding environment changes. Infants demonstrate such predictions through their longer looking to such events that violate their predictions — a common and decades-old measurement for gauging the nature of infants’ knowledge. Adopting this “surprise paradigm” to study machine intelligence allows for direct comparisons between an algorithm’s quantitative measure of surprise and a well-established human psychological measure of surprise — infants’ looking time. The models showed no such evidence of understanding the motivations underlying such actions, revealing that they are missing key foundational principles of commonsense psychology that infants possess.
    “A human infant’s foundational knowledge is limited, abstract, and reflects our evolutionary inheritance, yet it can accommodate any context or culture in which that infant might live and learn,” observes Dillon.
    The research was supported by grants from the National Science Foundation (DRL1845924) and the Defense Advanced Projects Research Agency (HR001119S0005). More

  • in

    Solid-state thermal transistor demonstrated

    An effective, stable solid-state electrochemical transistor has been developed, heralding a new era in thermal management technology.
    In modern electronics, a large amount of heat is produced as waste during usage — this is why devices such as laptops and mobile phones become warm during use, and require cooling solutions. In the last decade, the concept of managing this heat using electricity has been tested, leading to the development of electrochemical thermal transistors — devices that can be used to control heat flow with electrical signals. Currently, liquid-state thermal transistors are in use, but have critical limitations: chiefly, any leakage causes the device to stop working.
    A research team at Hokkaido University lead by Professor Hiromichi Ohta at the Research Institute for Electronic science has developed the first solid-state electrochemical thermal transistor. Their invention, described in the journal Advanced Functional Materials, is much more stable than and just as effective as current liquid-state thermal transistors.
    “A thermal transistor consists broadly of two materials, the active material and the switching material,” explains Ohta. “The active material has changeable thermal conductivity, and the switching material is used to control the thermal conductivity of the active material.”
    The team constructed their thermal transistor on a yttrium oxide-stabilized zirconium oxide base, which also functioned as the switching material, and used strontium cobalt oxide as the active material. Platinum electrodes were used to supply the power required to control the transistor.
    The thermal conductivity of the active material in the “on” state was comparable to some liquid-state thermal transistors. In general, thermal conductivity of the active material was four times higher in the “on” state compared to the “off” state. Further, the transistor was stable over 10 use cycles, better than some current liquid-state thermal transistors. This behavior was tested across more than 20 separately fabricated thermal transistors, ensuring the results were reproducible. The only drawback was the operating temperature of around 300°C.
    “Our findings show that solid-state electrochemical thermal transistors have the potential to be just as effective as liquid-state electrochemical thermal transistors, with none of their limitations,” concludes Ohta. “The main hurdle to developing practical thermal transistors is the high resistance of the switching material, and hence a high operating temperature. This will be the focus of our future research.” More

  • in

    First wearable device for vocal fatigue senses when your voice needs a break

    Northwestern University researchers have developed the first smart wearable device to continuously track how much people use their voices, alerting them to overuse before vocal fatigue and potential injury set in.
    The first-of-its-kind, battery-powered, wireless device and accompanying algorithms could be a game-changer for professional singers, teachers, politicians, call-center workers, coaches and anyone who relies on their voices to communicate effectively and make a living. It also could help clinicians remotely and continuously monitor patients with voice disorders throughout their treatment.
    Developed by an interdisciplinary team of materials scientists, biomedical engineers, opera singers and a speech-language pathologist, the research behind the new technology will be published during the week of Feb. 20 in the Proceedings of the National Academy of Sciences.
    The soft, flexible, postage-stamp-sized device comfortably adheres to the upper chest to sense the subtle vibrations associated with talking and singing. From there, the captured data is instantaneously streamed via Bluetooth to the users’ smartphone or tablet, so they can monitor their vocal activities in real time throughout the day and measure cumulative total vocal usage. Custom machine-learning algorithms distinguish the difference between speaking and singing, enabling singers to separately track each activity.
    With the app, users can set their personalized vocal thresholds. When they near that threshold, their smartphone, smartwatch or an accompanying device located on the wrist provides real-time haptic feedback as an alert. Then, they can rest their voices before pushing it too far.
    “The device precisely measures the amplitude and frequency for speaking and singing,” said Northwestern’s John A. Rogers, a bioelectronics pioneer who led the device’s development. “Those two parameters are most important in determining the overall load that’s occurring on the vocal folds. Being aware of those parameters, both at a given instant and cumulatively over time, is essential for managing healthy patterns of vocalization.”
    “It’s easy for people to forget how much they use their voice,” said Northwestern’s Theresa Brancaccio, a voice expert who co-led the study. “Seasoned classical singers tend to be more aware of their vocal usage because they have lived and learned. But some people — especially singers with less training or people, like teachers, politicians and sports coaches, who must speak a lot for their jobs — often don’t realize how much they are pushing it. We want to give them greater awareness to help prevent injury.”

    Rogers is the Louis Simpson and Kimberly Querrey Professor of Materials Science and Engineering, Biomedical Engineering and Neurological Surgery in the McCormick School of Engineering and Northwestern University Feinberg School of Medicine. He also is director of the Querrey Simpson Institute for Bioelectronics. A distinguished operatic performer, mezzo-soprano, Brancaccio is a senior lecturer at Northwestern’s Bienen School of Music, where she teaches voice and vocal pedagogy.
    Unaware of overuse
    For the millions of people in the U.S. who make their livings by speaking or singing, vocal fatigue is a constant, looming threat. The common condition occurs when overused vocal folds swell, making the voice sound raspy and lose endurance. Vocal fatigue negatively affects singers, in particular, altering their abilities to sing clearly or hit the same notes as their healthy voice can. At best, one short period of vocal fatigue can briefly interrupt a singer’s plans. At worst, it can lead to enough damage to derail a career.
    Lack of awareness is the underlying problem. People rarely make the connection between vocal activities and how those activities affect their voices. Although one in 13 U.S. adults have experienced vocal fatigue, most people don’t notice they are overusing their voices until hoarseness already has set in.
    “What leads people into trouble is when events stack up,” Brancaccio said. “They might have rehearsals, teach lessons, talk during class discussions and then go to a loud party, where they have to yell over the background noise. Then, throw a cold or illness into the mix. People have no idea how much they are coughing or clearing their throats. When these events stack up for days, that can put major stress on the voice.”
    Cross-disciplinary connection

    As an advocate for vocal health, Brancaccio has spent decades exploring ways to keep her students mindful of how much they use their voices. In 2009, she challenged her students to keep a paper budget — physically writing down every time they spoke, sang and drank water, among other things. About 10 years later, she converted the system into Singer Savvy, an app that offers a personalized vocal budget for each user and helps users stay within that budget.
    Separately, Rogers, in collaboration with researchers at the Shirley Ryan AbilityLab, had developed a wireless wearable device to track swallowing and speech in stroke patients. The bandage-like sensor measures swallowing abilities and speech patterns to monitor stroke patients’ recovery processes. In the early weeks of the COVID-19 pandemic, Rogers’ team modified the technology to monitor coughing, as a key symptom of the illness.
    “I wanted to gather more data and make our tracking system more precise and more accurate,” Brancaccio said. “So, I reached out to John to see if his sensors could help us gather more information.”
    “I thought it was a great opportunity for us to extend our technologies beyond our very important, but narrowly targeted, uses in health care to something that might capture a broader population of users,” Rogers said. “Anyone who uses their voice extensively could benefit.”
    The pair also partnered with speech pathologist and voice expert Aaron M. Johnson to explore how the devices could be used to evaluate and monitor treatment for patients with vocal disorders. Johnson, who co-directs the NYU Langone’s Voice Center, said the small, wireless device could help track patients’ voices in the real world — outside of a clinical setting.
    “A key part of voice therapy is helping people change how — and how much — they use their voice,” said Johnson, study co-author and associate professor in the department of otolaryngology at NYU Grossman School of Medicine. “This device will enable patients and their clinicians to understand voice use patterns and make adjustments in vocal demand to reduce vocal fatigue and speed recovery from voice disorders. Generalizing vocal techniques and exercises from therapy sessions into daily life is one of the most challenging aspects of voice therapy, and this device could greatly enhance that process.”
    Singer-trained algorithms
    The team modified Rogers’ existing devices to precisely measure vocal load over time. That includes frequency, volume, amplitude, duration and time of day. Like Rogers’ previous devices for COVID-19 and stroke patients, the new device also senses vibrations rather than recording audio. This enables the device to detect vocal activity precisely from the user, rather than the ambient noise surrounding them.
    The biggest challenge was to develop algorithms capable of distinguishing speaking from singing. To overcome this challenge, Brancaccio recruited voice and opera students to undertake a variety of singing exercises to train the machine-learning algorithms. A team of classical singers with different vocal ranges — varying from bass to soprano — wore the devices while humming, singing staccato scales and songs, reading and more. Each singer generated 2,500 one-second-long windows of singing and 2,500 one-second-long windows of speaking.
    The resulting algorithm can separate singing from speaking with more than 95% accuracy. And, when used in a choir setting, the device captures only data from the wearer and not noise from nearby singers.
    “Prolonged talking is one of the most fatiguing activities for people who are training to become professional singers,” Brancaccio said. “By separating singing and speaking, it can help people develop more awareness around how much they are speaking. There is evidence that even brief 15- to 20-minute periods of total silence interspersed throughout the day can help vocal fold tissues recover and repair.”
    How to use it
    To use the device, the wearer simply adheres it to the sternum, below the neck, and syncs the device with the accompanying app. Rogers’ team currently is working on a method to personalize vocal budgets for each user. Here, users will press a button in the app if they experience vocal discomfort at any point during the day, effectively capturing the instantaneous and cumulative vocal load at the time. These data can serve as a personalized threshold for vocal fatigue. When the user nears or exceeds their personalized threshold, a haptic device will vibrate as an alert.
    Similar in size and form to a wristwatch, this haptic device includes multiple motors that can activate in different patterns and with varying levels of intensity to convey different messages. Users also can monitor a graphical display within the app, which splits information into speaking and singing categories.
    “It uses Bluetooth, so it can talk to any device that has a haptic motor embedded,” Rogers said. “So, you don’t have to use our wristband. You could just leverage a standard smart watch for haptic feedback.”
    Although other vocal-monitoring devices do exist, those use big collars, tethering wires and bulky equipment. Some also use embedded microphones to capture audible vocal data, leading to privacy concerns.
    “Those don’t work for continuous monitoring in a real environment,” Brancaccio said. “Instead of wearing cumbersome, wired equipment, I can stick on this soft, wearable device. Once it’s on, I don’t even notice it. It’s super light and easy.”
    What’s next
    Because Rogers’ previous devices capture body temperature, heart rates and respiratory activity, the researchers included those capabilities in the vocal-monitoring device. They believe these extra data will help to explore fundamental research questions concerning vocal fatigue.
    “This is more speculative, but it might be interesting to see how physical activity affects vocal fatigue,” Rogers said. “If someone is dancing while singing, is that more stressful on the vocal folds compared to someone who is not physically exerting themselves? Those are the kinds of questions we can ask and quantitatively answer.”
    In the meantime, Brancaccio is excited for her students to have a tool that can help prevent injury. She hopes others — including non-singers — will see the benefit to keeping their vocal cords healthy.
    “Your voice is part of your identity — whether you are a singer or not,” she said. “It’s integral to daily life, and it’s worth protecting.”
    The study, “Closed-loop network of skin-interfaced wireless devices for quantifying vocal fatigue and providing user feedback,” was supported by the Querrey Simpson Institute for Bioelectronics at Northwestern University. More