Archives: News

news custom post type.

How to detect covert AI-generated text

Machine-generated text has been fooling humans since the release of GPT-2 in 2019. Large language model (LLM) tools have gotten progressively better at crafting stories, news articles, student essays and more. So humans are often unable to recognize when they are reading text produced by an algorithm.

Misuse and harmful outcomes

These LLMs are being used to save time and even boost creativity in ideating and writing, but their power can lead to misuse and harmful outcomes that are already showing up where we consume information.

One way both academics and companies are trying to improve this detection is by employing machine learning models that can identify subtle patterns of word choice and grammatical constructions. These allow for recognizing LLM-generated text in a way that our human intuition cannot. 

Today, many commercial detectors are claiming to be highly successful at detecting machine-generated text, with up to 99% accuracy—but are these claims verified? Chris Callison-Burch, Professor in Computer and Information Science, and Liam Dugan, a doctoral student in Callison-Burch’s group, aimed to find out in their recent paper, published at the 62nd Annual Meeting of the Association for Computational Linguistics.

“As the technology to detect machine-generated text advances, so does the technology used to evade detectors,” said Callison-Burch in a statement. “It’s an arms race, and while the goal to develop robust detectors is one we should strive to achieve, there are many limitations and vulnerabilities in detectors that are available now.”   

Testing AI detector ability

To investigate those limitations and provide a path forward for developing robust detectors, the research team created Robust AI Detector (RAID), a data set of more than 10 million documents across recipes, news articles, blog posts and more, including AI-generated text and human-generated text.

RAID is the first standardized benchmark to test detection ability in current and future detectors. The team also created a leaderboard, which publicly ranks the performance of all detectors that have been evaluated using RAID in an unbiased way.

“The concept of a leaderboard has been key to success in many aspects of machine learning like computer vision,” said Dugan in a statement. “The RAID benchmark is the first leaderboard for robust detection of AI-generated text. We hope that our leaderboard will encourage transparency and high-quality research in this quickly evolving field.”

Dugan has already seen the influence this recemt paper is having in companies that develop detectors. “Originality.ai is a prominent company that develops detectors for AI-generated text,” he says. “They shared our work in a blog post, ranked their detector in our leaderboard and are using RAID to identify previously hidden vulnerabilities and improve their detection tool.”

How detectors fail

So do the current detectors hold up to the work at hand? RAID shows that not many do as well as they claim. 

“Detectors trained on ChatGPT were mostly useless in detecting machine-generated text outputs from other LLMs such as Llama and vice versa,” says Callison-Burch. “Detectors trained on news stories don’t hold up when reviewing machine-generated recipes or creative writing. What we found is that there are a myriad of detectors that only work well when applied to very specific use cases and when reviewing text similar to the text they were trained on.” 

Detectors are able to detect AI-generated text when it contains no edits or “disguises,” but when manipulated, current detectors are not reliably able to detect AI-generated text. Faulty detectors are not only an issue because they don’t work well; they can be as dangerous as the AI tool used to produce the text in the first place. 

“If universities or schools were relying on a narrowly trained detector to catch students’ use of ChatGPT to write assignments, they could be falsely accusing students of cheating when they are not,” says Callison-Burch. “They could also miss students who were cheating by using other LLMs to generate their homework.”   

Adversarial attacks and tricks

The team also looked into how adversarial attacks (such as replacing letters with look-alike symbols) can easily derail a detector and allow machine-generated text to fly under the radar.

“It turns out, there are a variety of edits a user can make to evade detection by the detectors we evaluated in this study,” says Dugan. “Something as simple as inserting extra spaces, swapping letters for symbols, or using alternative spelling or synonyms for a few words can cause a detector to be rendered useless.”

While current detectors are not robust enough to be of significant use in society just yet, openly evaluating detectors on large, diverse, shared resources is critical to accelerating progress and trust in detection, and that transparency will lead to the development of detectors that do hold up in a variety of use cases, the study concludes. 

Reducing harms

“My work is focused on reducing the harms that LLMs can inadvertently cause, and, at the very least, making people aware of the harms so that they can be better informed when interacting with information,” Dugan continues. “In the realm of information distribution and consumption, it will become increasingly important to understand where and how text is generated, and this paper is just one way I am working towards bridging those gaps in both the scientific and public communities.”

This study was funded by the Intelligence Advanced Research Activity (IARPA), a directive of the Office of the Director of National Intelligence and within the Human Interpretable Attribution of Text Using Underlying Structure (HIATUS) program.

Citation: Dugan, L., Hwang, A., Trhlik, F., Ludan, J. M., Zhu, A., Xu, H., & Ippolito, D. (2024). RAID: A Shared Benchmark for Robust Evaluation of Machine-Generated Text Detectors. arXiv. https://arxiv.org/abs/2405.07940 (open access)

Let us know your thoughts! Sign up for a Mindplex account now, join our Telegram, or follow us on Twitter. 

Smart fabric for sensing and energy harvesting

University of Waterloo researchers have developed a smart fabric that can convert body heat and solar energy into electricity, replacing batteries and other power sources for energy harvesting, health monitoring, and movement tracking applications. 

The fabric sensors can monitor temperature, pressure, chemical composite, and more, which can be integrated into the material. For example, a smart face mask that can track breath temperature and rate and detect chemicals in breath to help identify viruses, lung cancer, and other conditions.

The researchers say the material is more stable, durable, and cost-effective than other fabrics on the market, integrating advanced materials such as MXene and conductive polymers with cutting-edge textile technologies to advance smart fabrics for wearable technology.

The study is published in the Journal of Materials Science & Technology.

The next phase of research will focus on further enhancing the fabric’s performance and integrating it with electronic components, in collaboration with electrical and computer engineers. Future developments may include a smartphone app to track and transmit data from the fabric to healthcare professionals, enabling real-time, non-invasive health monitoring and everyday use.

Citation: Peng, J., Ge, F., Han, W., Wu, T., Tang, J., Li, Y., & Wang, C. (2024). MXene-based thermoelectric fabric integrated with temperature and strain sensing for health monitoring. Journal of Materials Science & Technology, 212, 272-280. https://doi.org/10.1016/j.jmst.2024.06.011

Let us know your thoughts! Sign up for a Mindplex account now, join our Telegram, or follow us on Twitter. 

Scientists find oceans of water on Mars

Geophysicists have found evidence for a large underground reservoir of liquid water—enough to fill oceans on the planet’s surface, using seismic activity.

The Scripps Oceanography scientists estimate that the amount of groundwater could cover the entire planet to a depth of between 1 and 2 kilometers (about a mile), based on data from NASA’s Insight lander.

However, the water is located in tiny cracks and pores in rock in the middle of the Martian crust, between 11.5 and 20 kilometers below the surface, a challenge to reach by drilling.

Water on the planet’s surface

Manga noted that lots of evidence—river channels, deltas and lake deposits, as well as water-altered rock—supports the hypothesis that water once flowed on the planet’s surface more than 3 billion years ago, after Mars lost its atmosphere.

The researchers note that understanding the water situation on Mars will help us get closer to knowing if life exists there. The research appears this week in the journal Proceedings of the National Academy of Sciences.

The Canadian Institute for Advanced Research, the National Science Foundation and the U.S. Office of Naval Research supported the work.

Citation: Wright, V., Morzfeld, M., & Manga, M. (2024). Liquid water in the Martian mid-crust. Proceedings of the National Academy of Sciences, 121(35), e2409983121. https://doi.org/10.1073/pnas.2409983121 (open access)

Let us know your thoughts! Sign up for a Mindplex account now, join our Telegram, or follow us on Twitter. 

New, more accurate way to deliver medicine to the brain

Houston Methodist researchers have discovered a more accurate and timely way to deliver life-saving drug therapies to the brain, laying the groundwork for more effective treatment of brain tumors and other neurological diseases than with the current convection-enhanced delivery (CED).

Delivering the correct dosage of drugs to the right place in the brain has long been a challenge. The natural blood-brain barrier that protects our brains from toxins and pathogens can also block the delivery of important medical treatments. And CED also follows the path of least resistance, so therapeutics don’t always hit the target.

Electric field infuses medicine accurately

So the researchers developed a new process called “electrokinetic convection-enhanced delivery” (ECED) that allows surgeons to design the appropriate delivery path and potentially reach brain lesions and tumors more accurately. ECED uses an electric field to infuse medicine from a reservoir outside the brain to specific targets inside the brain. It applies continuous pressure over time to inject a fluid containing therapeutics into the brain.

In the Springer Nature journal Communications Biology, Houston Methodist research scientist and co-author Jesus G. Cruz-Garza explains how ECED infuses macromolecules into the brain from a hydrogel reservoir placed at the brain’s surface. “The brain acts as a charged porous scaffold that in the presence of an electric field allows for electroosmosis—bulk fluid flow in a porous media.” From the hydrogel reservoir, Cruz-Garza explains, this bulk flow of fluid enables the delivery of therapeutic agents.

This invention improves on the 30-year-old process of injecting therapeutics into the brain via, say the researchers.

“Delivering therapeutics by way of ECED has many applications,” explained Dr. Amir Faraji, principal investigator and Houston Methodist neurosurgeon, in a statement. “It has the potential to “improve gene therapy and as a treatment for traumatic brain injury and degenerative diseases, and in a more targeted manner.”

The research was conducted at Houston Methodist Department of Neurosurgery, Houston Methodist Research Institute, Center for Neural Systems Restoration, Center for Neuroregeneration, in conjunction with Texas A&M University College of Medicine and School of Engineering.

Further research is needed before Faraji and team can bring this investigational therapy to humans, the researchers advise.

Citations:

Cruz-Garza, J.G., Bhenderu, L.S., Taghlabi, K.M. et al. (2024) Electrokinetic convection-enhanced delivery for infusion into the brain from a hydrogel reservoir. Commun Biol 7, 869. https://www.nature.com/articles/s42003-024-06404-1 (open access)

Eid, F., Chen, A. T., Chan, K. Y., Huang, Q., Zheng, Q., Tobey, I. G., Pacouret, S., Brauer, P. P., Keyes, C., Powell, M., Johnston, J., Zhao, B., Lage, K., Tarantal, A. F., Chan, Y. A., & Deverman, B. E. (2024). Systematic multi-trait AAV capsid engineering for efficient gene delivery. Nature Communications, 15(1), 1-14. https://www.nature.com/articles/s41467-024-50555-y (open-access)

Let us know your thoughts! Sign up for a Mindplex account now, join our Telegram, or follow us on Twitter. 

Tracking how psychedelics affect neurons

Researchers at the University of California, Davis have developed a rapid, noninvasive tool to track the neurons and biomolecules activated in the brain by psychedelic drugs.

The protein-based tool, called Ca2+-activated Split-TurboID, or CaST, is described in research published in Nature Methods. 

The new tool could help scientists unlock the benefits of psychedelic treatments for patients with brain disorders. 

There has been mounting interest in the value of psychedelic-inspired compounds as treatments for brain disorders, including depression, post-traumatic stress and substance use. Psychedelic compounds like LSD, DMT and psilocybin promote the growth and strengthening of neurons and their connections in the brain’s prefrontal cortex.

“It’s important to think about the cellular mechanisms that these psychedelics act upon,” said Christina Kim, an assistant professor of neurology at the UC Davis Center for Neuroscience and School of Medicine, and an affiliate of the UC Davis Institute for Psychedelics and Neurotherapeutics, in a statement.

The new technique could be used to track step-by-step the molecular signaling processes that are responsible for these compounds’ beneficial neuroplastic effects, taking 10 to 30 minutes, rather than the hours typical of other tagging methods.   

“We designed these proteins in the lab that can be packaged into DNA and then put into harmless adeno-associated viruses,” Kim said. “Once we deliver the CaST tool and these proteins into neurons, they incubate inside the cells and start expressing.”

A snapshot of the brain

The CaST tool measures changes in intracellular calcium concentrations, a nearly universal marker to track activity in a neuron. When neurons exhibit high activity, they exhibit high calcium levels. CaST uses this cue to tag the cell with a small biomolecule called biotin.

In the study, Kim and her colleagues dosed mice with the psychedelic psilocybin. They then used CaST in tandem with biotin to identify neurons with increased calcium in the prefrontal cortex (an area affected by many brain disorders and where psychedelics promote neuronal growth and strengthening).

The researchers also monitor the mouse head-twitch responses of a freely behaving animal. These are the primary behavioral correlates for hallucinations caused by psychedelics.

Next steps

Kim and her colleagues are now working on methods to enable brain-wide cellular labeling with the CaST tool and ways to enrich the signature of individual proteins produced by neurons affected by psychedelics. 

“We can send those samples to the UC Davis Proteomics Core Facility and they can give us an unbiased picture of all the proteins we identified,” Kim said.

The goal is to identify how psychedelics benefit the cellular profiles of those with brain disorders, elucidating the step-by-step cellular process of their therapeutic effects. 

The work was supported by grants from the Brain and Behavior Research Foundation, Kinship Foundation, Arnold and Mabel Beckman Foundation, NIH, NSF and the Boone Family Foundation. 

Citation: Zhang, R., Anguiano, M., Aarrestad, I. K., Lin, S., Chandra, J., Vadde, S. S., Olson, D. E., & Kim, C. K. (2024). Rapid, biochemical tagging of cellular activity history in vivo. Nature Methods, 1-11. https://doi.org/10.1038/s41592-024-02375-7 (open access)

Let us know your thoughts! Sign up for a Mindplex account now, join our Telegram, or follow us on Twitter. 

Sustainable AI for healthcare

High-speed development of AI has made its way into healthcare—particularly in the radiology field.

AI-based diagnostic systems are flourishing, with hospitals quickly adopting the technology to assist radiologists with AI-based diagnostic systems, for example. However, there are concerns about the environmental impact of increasingly complex AI models.

Energy consumption of medical AI systems

So Associate Professor Daiju Ueda of Osaka Metropolitan University’s Graduate School of Medicine led a research team to investigate energy consumption of AI systems in the medical field, carbon emissions of data centers, and electronic waste issues. The goal: developing energy-efficient AI models, implementing green computing and the use of renewable energy.

In the study, the researchers proposed specific guidelines for the sustainable deployment of AI in the medical field in an environmentally responsible manner.

“The challenge for the future will be to verify and further elaborate these recommendations in actual medical practice,” said Ueda in a statement. “They are also expected to contribute to the standardization of methods for assessing AI’s environmental impact and the development of an international regulatory framework.”

The results were published in Diagnostic and Interventional Imaging.

Citation: Ueda, D., Walston, S. L., Fujita, S., Fushimi, Y., Tsuboyama, T., Kamagata, K., Yamada, A., Yanagawa, M., Ito, R., Fujima, N., Kawamura, M., Nakaura, T., Matsui, Y., Tatsugami, F., Fujioka, T., Nozaki, T., Hirata, K., & Naganawa, S. (2024). Climate change and artificial intelligence in healthcare: Review and recommendations towards a sustainable future. Diagnostic and Interventional Imaging. https://doi.org/10.1016/j.diii.2024.06.002 (open access)

Let us know your thoughts! Sign up for a Mindplex account now, join our Telegram, or follow us on Twitter. 

Mapping the body’s internal sensory communication highway

Running from the brain to the large intestine, the vagus nerve is the body’s longest cranial nerve, encoding sensory information from the visceral organs. It plays a key role in respiratory, gastrointestinal, cardiovascular, endocrine and immune system functions.

Interoception—the body’s ability to sense its internal state in a timely and precise manner—is facilitated by the vagus sensory neurons, which independently code the three critical features of a body signal: the involved visceral organ, the tissue layer where the signal originates, and the type of sensory modality.

This was just discovered by Qiancheng Zhao, assistant professor in medicine-endocrinology and McNair Scholar at Baylor University School of Medicine, and Department of Medicine-Endocrinology, Baylor College of Medicine, and Department of Neuroscience, School of Medicine, Yale University.

Clinical applications

Zhao said his findings could provide potential new vagal targets, so future researchers might use this genetic information to help them access different visceral organs precisely.

The more detailed map also could help clarify the vagus nerve’s role in interception (the body’s ability to sense its internal state in a timely and precise manner) and find whether there might be neural modulatory applications for treating interoceptive disorders in respiratory, gastrointestinal, cardiovascular, endocrine and immune system functions.

Encoding sensory information

However, it has not been clear how the body’s longest cranial nerve, running from the brain to the large intestine, encodes sensory information from the visceral organs.

Applying a variety of techniques, Zhao discovered that vagal sensory neurons independently code the three critical features of a body signal: the involved visceral organ; the tissue layer where the signal originates; and the type of sensory modality.

Decoding vagal sensory nerve traffic

The vagal highway is crowded with traffic, with sensory and motor pathways intermingling in the nerve bundle. One of the potential challenges of expanding vagal nerve stimulation (VNS) is finding ways to identify and target specific vagal signals, instead of broadly stimulating the nerve—potentially creating unwanted side effects.

“We know that vagal sensory neurons can project to the visceral organs,” said Zhao. “So our question was: what signals from those visceral organs need to be sensed by vagal sensory neurons?”

To decipher the complexity of VNS traffic, Zhao and his colleagues focused on three aspects: the visceral organ sending the signal, the tissue layer in the organ where the signal originates and the kind of sensory stimulus.

To identify organ-projecting neurons, they have combined a viral tracing approach with single-cell RNA sequencing The analysis revealed that vagal sensory neurons use different gene modules to code specific visceral organs. They also traced neuronal projections in transparent, whole-mounted mouse organs to determine how vagal sensory neurons innervate the layers of specific tissues.

Zhao and his colleague Chuyue Yu at Yale University also developed an in vivo calcium imaging technique. It allowed them to identify the molecular features of vagal sensory neurons responding to different types of stimuli, such as mechanical inflation in the lung and chemical stimuli of nutrients from a protein shake in the mouse GI tract.

“When we have the anatomical map together with the molecular information and the inputs from functional imaging, then we can really have a full picture to understand the sensory logic,” Zhao said.

Clinical applications

Zhao’s research could inform future therapies that stimulate the vagus nerve to treat a variety of physical and psychiatric disorders, using VNS through an implantable electrical pulse generator (which has been approved by the U.S. Food and Drug Administration to treat drug-resistant epilepsy and depression).

Zhao said his findings could provide potential new vagal targets, so future researchers might use this genetic information to help them access different visceral organs precisely. The more detailed map also could help clarify the vagus nerve’s role in interoception and whether there might be neural modulatory applications for treating interoceptive disorders.

Aging

Zhao is also interested in discovering more about how vagal sensory neurons might behave differently across the lifespan and under different disease conditions. “We know that our physiological parameters, such as heart rate and blood pressure will change during aging so we might want to look, for example, at how the aging state changes the interaction between the sensory nerve and different organs.”

For his work investigating and mapping this internal information highway, Zhao is the 2024 grand prize winner of the Science & PINS Prize for Neuromodulation.

Citation: Zhao, Q. (2024). Navigating internal senses: A road map for the vagal interoceptive system. Science. https://www.science.org/doi/10.1126/science.adq8578 (open access)

Let us know your thoughts! Sign up for a Mindplex account now, join our Telegram, or follow us on Twitter. 

Smithsonian scientists devise method to secure Earth’s biodiversity on the moon

New research led by scientists at the Smithsonian proposes a plan to safeguard Earth’s imperiled biodiversity by cryogenically preserving biological material on the moon.

The moon’s permanently shadowed craters are cold enough for cryogenic preservation without the need for electricity or liquid nitrogen, according to the researchers. 

The paper, published today in BioScience and written in collaboration with researchers from the Smithsonian’s National Zoo and Conservation Biology Institute (NZCBI), Smithsonian’s National Museum of Natural History, Smithsonian’s National Air and Space Museum and others, outlines a roadmap to create a lunar biorepository.

The paper includes ideas for governance, the types of biological material to be stored and a plan for experiments to understand and address challenges such as radiation and microgravity. The study also demonstrates the successful cryopreservation of skin samples from a fish, which are now stored at the National Museum of Natural History. 

“Initially, a lunar biorepository would target the most at-risk species on Earth today, but our ultimate goal would be to cryopreserve most species on Earth,” said Mary Hagedorn, a research cryobiologist at NZCBI and lead author of the paper in a statement,

Global Seed Vault model

The proposal was inspired by the Global Seed Vault in Svalbard, Norway, which contains more than 1 million frozen seed varieties 400 feet underground and functions as a backup for the world’s crop biodiversity in case of global disaster. However, in 2017, thawing permafrost threatened the collection with a flood of meltwater. The seed vault has since been waterproofed, but the incident showed that even an Arctic subterranean bunker could be vulnerable to climate change. 

Cryopreservation storage temperatures

Unlike seeds, animal cells require much lower storage temperatures for preservation (-320 degrees Fahrenheit or -196 degrees Celsius). On Earth, cryopreservation of animal cells requires a supply of liquid nitrogen, electricity and human staff potentially vulnerable to disruptions that could destroy an entire collection, Hagedorn said.

To reduce these vulnerabilities, scientists needed a way to passively maintain cryopreservation storage temperatures. Since such cold temperatures do not naturally exist on Earth, Hagedorn and her co-authors looked to the moon. 

Moon’s polar regions: ideal for cryopreservation storage

The moon’s polar regions feature numerous craters that never receive sunlight due to their orientation and depth. These “permanently shadowed regions” can be −410 degrees Fahrenheit (−246 degrees Celsius)—more than cold enough for passive cryopreservation storage. To block out the DNA-damaging radiation present in space, samples could be stored underground or inside a structure with thick walls made of moon rocks. 

Fibroblasts cryopreserved easily

At the Hawaiʻi Institute of Marine Biology, the research team cryopreserved skin samples from a reef fish called the starry goby. The fins contain a type of skin cell called fibroblasts, the primary material to be stored in the National Museum of Natural History’s biorepository.

The team says fibroblasts have several advantages over other types of commonly cryopreserved cells such as sperm, eggs and embryos. Science cannot yet reliably preserve the sperm, eggs and embryos of most wildlife species. However, for many species, fibroblasts can be cryopreserved easily. In addition, fibroblasts can be collected from an animal’s skin, which is simpler than harvesting eggs or sperm. For species that do not have skin per se, such as invertebrates, Hagedorn said the team may use a diversity of types of samples depending on the species, including larvae and other reproductive materials.   

The next steps are to begin a series of radiation exposure tests for the cryopreserved fibroblasts on Earth to help design packaging that could safely deliver samples to the moon. The team is actively seeking partners and support to conduct additional experiments on Earth and aboard the International Space Station. Such experiments would provide robust testing for the prototype packaging’s ability to withstand the radiation and microgravity associated with space travel and storage on the moon. 

“We aren’t saying what if the Earth fails—if the Earth is biologically destroyed this biorepository won’t matter,” Hagedorn said. “This is meant to help offset natural disasters and, potentially, to augment space travel. Life is precious and, as far as we know, rare in the universe. This biorepository provides another, parallel approach to conserving Earth’s precious biodiversity.”    

Citation: Mary Hagedorn, Smithsonian National Zoological Park Library, Lynne Parenti, Smithsonian National Museum of Natural History, Robert A. Craddock, Smithsonian Institution, Pierre Comizzoli, Smithsonian National Zoological Park Library, Paula Mabee, National Ecological Observatory Network, Bonnie Meinke, University Corporation for Atmospheric Research, Susan M. Wolf, University of Minnesota, John Bischof, University of Minnesota. Rebecca Sandlin, Harvard Medical School, Shannon N. Tessier, Harvard Medical School, Mehmet Toner, Harvard-MIT Division of Health Sciences and Technology. Safeguarding Earth’s Biodiversity By Creating a Lunar Biorepository. BioScience, Oxford University Press. DOI: 10.1093/biosci/biae058

Let us know your thoughts! Sign up for a Mindplex account now, join our Telegram, or follow us on Twitter. 

Wearable blood pressure monitor uses ultrasound to capture a continuous record of blood pressure

Continuous, noninvasive blood pressure monitoring has been a longtime goal of medicine because of blood pressure’s utility as a metric for clinicians.

However, options have been limited to internally placed arterial catheters or inflatable pressure cuffs, requiring frequent calibration with an inflatable cuff.

Ultrasound-based measurement

Raymond Jimenez and colleagues propose a new method based on “resonance sonomanometry,” in which the artery is stimulated by an acoustic transducer (similar to a medical ultrasound scanner) and the resonant response and dimensions are measured using ultrasound.

Similar to how a guitar string changes pitch as its tension is manipulated, changing circumferential tension (by blood pulses) of the arterial wall changes its resonant frequency through the continuous phases of the cardiac cycle.

The method was tested on humans on the carotid artery in the neck and the axillary, brachial, and femoral arteries. Measurements were compared with those from a blood pressure cuff.

Not limited to an arm

All four sites produced measurements in a single subject that were broadly in line with those obtained from a cuff. Additional testing on the carotid arteries of six volunteers showed promising results, albeit with lower systolic values (predictable, given the carotid being closer to the heart than the brachial artery measured by the cuff).

According to the authors, their proposed device could be worn over any ultrasound-accessible artery, even including the radial (back of arm near hand), where the device could be worn like a watch.

Citation: Jimenez, R., Yurk, D., Dell, S., Rutledge, A. C., Fu, M. K., Dempsey, W. P., Rajagopal, A., & Brinley Rajagopal, A. (2024). Resonance sonomanometry for noninvasive, continuous monitoring of blood pressure. PNAS Nexus, 3(7). https://academic.oup.com/pnasnexus/article/3/7/pgae252/7717708 (open access)

Let us know your thoughts! Sign up for a Mindplex account now, join our Telegram, or follow us on Twitter. 

How to make AI more energy-efficient

The International Energy Agency (IEA) issued a global energy use forecast in March 2024 warning that global energy consumption for AI is likely to double from 460 terawatt-hours (TWh) in 2022 to 1,000 TWh in 2026—roughly equivalent to the electricity consumption of the entire country of Japan. 

The reason: machine or AI processes transfer data between logic (where information is processed within a system) and memory (where the data is stored)—consuming a large amount of power and energy in the process. 

Significant energy savings

So engineering researchers at the University of Minnesota Twin Cities have developed a state-of-the-art hardware device in which data never leaves computational random-access memory (CRAM). Compared to traditional methods, the new CRAM-based machine-learning inference accelerator can achieve 1,700 to 2,500 times energy savings, the researchers estimate.

The research is published in npj Unconventional Computing, a peer-reviewed scientific journal published by Nature. The work was supported by grants from the U.S. Defense Advanced Research Projects Agency (DARPA), the National Institute of Standards and Technology (NIST), the National Science Foundation (NSF) and Cisco Inc.

Citation: Lv, Y., Zink, B. R., Bloom, R. P., Cılasun, H., Khanal, P., Resch, S., Chowdhury, Z., Habiboglu, A., Wang, W., Sapatnekar, S. S., Karpuzcu, U., & Wang, J. (2024). Experimental demonstration of magnetic tunnel junction-based computational random-access memory. Npj Unconventional Computing, 1(1), 1-10. https://doi.org/10.1038/s44335-024-00003-3 (open access)

Let us know your thoughts! Sign up for a Mindplex account now, join our Telegram, or follow us on Twitter.