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)
Thursday, June 05, 2025
Multiple extreme climate events at the same time may be the new normal
According to the study, more and more regions will be affected by multiple climate-related extreme events simultaneously. The blue colours show regions where isolated events are most common, the red ones show where co-occurring events are most common.
The left map shows present day, and the right map shows the future with medium-high emissions.
Heatwaves, droughts and forest fires are some of the extreme climate-related events that are expected not only to become more frequent but also to increasingly strike at the same time. This finding emerges from a new study led by Uppsala University, in which researchers have mapped the impact of climate change in different regions of the world.
In a new study published in the journal Earth’s Future, researchers from Uppsala University and Belgian, French and German universities have shown that in the near future several regions of the world will no longer just be affected by isolated climate-related events. Instead, several different events will occur concurrently or in quick succession.
“We have long known, for example, that there will be more heatwaves, forest fires and severe droughts in many regions – that in itself is no surprise. What surprised us is that the increase is so large that we see a clear paradigm shift with multiple coinciding extreme events becoming the new normal,” says Professor Gabriele Messori, the study’s lead author.
Using models to predict the future climate – temperature, rainfall, wind and so on – is a common method in climate research. In this study, the researchers have gone a step further by feeding that data into additional models that deliver information on the concrete impact on society. By calculating the effect of climate change on, say, the risk of forest fires or floods, a clearer picture emerges of how different regions of the world might actually be affected. The analysis examines what will happen between 2050 and 2099. The researchers looked specifically at six types of events: floods, droughts, heatwaves, forest fires, tropical cyclone winds and crop failures.
Heatwaves and forest fires a recurrent feature
The study shows that combinations of heat waves and forest fires will increase sharply in almost all regions of the world, except where there is no vegetation, as in the Sahara. Heatwaves and droughts will become a recurrent feature in areas such as the Mediterranean region and Latin America. Areas that now generally experience isolated events, such as the Nordic countries, will also be more frequently affected by heatwaves and forest fires in combination.
“The summer of 2018 in Northern Europe was characterised by unusually high temperatures and widespread forest fires – which at the time was regarded as an exceptional event. In a few decades, it may not be so unusual,” says Messori.
Poses new challenges for preparedness
The researchers’ analysis covers several possible emission scenarios. However, the main focus is on a medium scenario, which is considered realistic given current emission trends.
“It is important to emphasise that this shift that we see does not only occur if we look at the most extreme case, where we do nothing to reduce our emissions, but also if we consider a less pessimistic scenario. From a societal perspective, we need to broaden our preparedness to deal with these co-occurring extreme events. We are going to face a new climate reality that we have limited experience of today.”
Schematic representation of the ultra-lightweight artificial intelligence model architecture and training process based on a massive-training artificial neural network (MTANN).
Credit: Kenji Suzuki from Institute of Science Tokyo, Japan
Imagine diagnosing cancer not with a supercomputer but on an ordinary laptop instead. Sounds like science fiction? Thanks to a revolutionary artificial intelligence (AI) model developed by Professor Kenji Suzuki and his research team from Institute of Science Tokyo (Science Tokyo), this far-fetched scenario is now a reality.
Unveiled at the prestigious Radiological Society of North America (RSNA) 2024 Annual Meeting, the team introduced an ultra-lightweight deep learning model that assists with lung cancer diagnosis without relying on costly graphics processing unit (GPU) servers or massive datasets. Developed using a unique deep learning approach based on massive-training artificial neural network (MTANN), the model was trained and tested on nothing more than a standard laptop computer, achieving what once required entire data centers.
AI, trained by deep learning models, has gained significant attention in recent years, leading to innovations in multiple fields of research. It has also been reported that if a deep learning model is trained on a large amount of data, such as a million images, it can acquire a performance that can surpass that of conventional technologies and even humans.
Where most models rely on big data, the AI model developed by Suzuki’s team is unique—unlike conventional large-scale AI models, it does not require entire medical image sets. Instead, it learns directly from individual pixels extracted from computed tomography (CT) scan images. This strategy significantly reduced the number of required cases from thousands to just 68!
Despite being trained only on a small set of data, the model outperformed state-of-the-art large-scale AI systems, such as Vision Transformer and 3D ResNet, attaining a discrimination performance corresponding to an area under the curve (AUC) value of 0.92 (against AUC values of 0.53 and 0.59 for the traditional state-of-the-art (SOTA) models, respectively). Once trained, with the full training process only taking 8 minutes and 20 seconds on a standard laptop, it could generate diagnostic predictions at an unprecedented rate of 47 milliseconds per case.
“This technology isn’t just about making AI cheaper or faster,” says Suzuki. “It’s about making powerful diagnostic tools accessible, especially for rare diseases where training data is hard to obtain. Furthermore, it will cut down the power demands for developing and using AI at data centers substantially, and can solve the global power shortage problem we may face due to the rapid growth in AI use.”
In recognition of its significance, the team’s research was conferred the coveted Cum Laude Award at RSNA 2024, an honor received by only 1.45% of the 1,312 presentations. While this innovation is sure to have a transformative impact on cancer diagnosis, it stands as a testament to Suzuki’s deep knowledge and unwavering dedication.
With profound expertise in the field of biomedical AI, Suzuki was the first to invent the MTANN technology (used in the current research) in the early 2000s. It was one of the earliest deep learning models that he had developed and improved on. In his 25 years of research experience, Suzuki has made significant contributions to his field, with more than 400 publications and over 40 patents, most of which have been licensed and commercialized.
Suzuki continues to lead groundbreaking research at the intersection of AI and medical imaging, actively fostering interdisciplinary collaboration that pushes the boundaries of what AI can achieve in clinical practice. His team’s work on compact, high-performance diagnostic models exemplifies how innovative thinking—combined with practical implementation—can bridge gaps between engineering and medicine. With a dynamic research environment and a strong network of collaborators, Suzuki is not only advancing the field of biomedical AI but also helping shape the next generation of translational medical technologies.
Despite being trained on a significantly lower-computational setup (MacBook Air with M1 chip), the 3D Massive-Training Artificial Neural Network (MTANN) achieves superior performance (area under the curve (AUC) = 0.92), faster inference, and drastically reduced training time and parameter count compared to that of 3D ResNet.
Credit
Kenji Suzuki from Institute of Science Tokyo, Japan
***
About Institute of Science Tokyo (Science Tokyo) Institute of Science Tokyo (Science Tokyo) was established on October 1, 2024, following the merger between Tokyo Medical and Dental University (TMDU) and Tokyo Institute of Technology (Tokyo Tech), with the mission of “Advancing science and human wellbeing to create value for and with society.”
Wet cough over a period lasting longer than four weeks? We are supposed to believe that this is not unusual for children, nor anything to be worried about. But Dr. Anne Schlegtendal, senior physician at the Universitätskinderklinik Bochum, Germany, says otherwise. Young children suffering from protracted bacterial bronchitis (PBB) have to be treated accordingly, she says. “If they do not receive antibiotics for at least two weeks, there could be long-term pulmonary sequelae. In the worst case, they could develop irreversible, chronic lung damage.” The team in Bochum behind the study reached this conclusion by examining children who had undergone inpatient treatment for protracted bacterial bronchitis in early childhood. The study was published in the journal Pediatric Pulmonology on April 21, 2025.
Noticeable pulmonary infection years later
“We identified 200 children who had developed PBB in early childhood and invited 63 of them to be re-examined and have their lung function tested at the Kinderklinik five to 14 years after their diagnosis,” says Anne Schlegtendal. “Among them were children who still have a chronic cough.” It was revealed that not only are the children who were not properly treated affected, but that children who had received antibiotics were also at risk of developing long-term damage. “Many of the children who had PBB in early childhood exhibit noticeable pulmonary function later on,” warns the pediatric pulmonologist. “Unfortunately, PBB is underdiagnosed and there is not yet a guideline in Germany for how to treat it, nor any recommendation that children suffering from it should undergo regular follow-up examinations.”
The researchers hope to change this. “We want to raise awareness of this disease to improve health in childhood and adolescence.” The team thus used their idea for a digital decision support tool in pediatric outpatient clinics to apply to the innovation fund of the Federal Joint Committee to improve care for patients in the statutory health insurance system. “Our idea is to use a digital signal system to identify and treat children with chronic cough and risk factors for eventual consequences earlier,” explains Schlegtendal. “Red would mean, go to the hospital right away.”
As Floridians prepare for an active 2025 Atlantic hurricane season, the most serious threat may not come from wind, but from water. Data from the National Hurricane Center shows that 86% of all direct hurricane and tropical storm fatalities in the United States between 2013 and 2023 were caused by water impacts – freshwater flooding, storm surge and rip currents. Of those, more than half resulted from drownings due to inland flooding, highlighting the critical importance of accurate rainfall and flood forecasting.
Florida Atlantic University’s Sensing Institute (I-SENSE) has emerged as a vital contributor to the nation’s weather forecasting capabilities through its leadership of the Southeast Atlantic (SEA) Econet. This academic-led network of atmospheric and hydrological monitoring stations delivers real-time data that directly informs forecasts and warnings issued by the National Weather Service.
The SEA Econet, managed in partnership with Coastal Carolina University, spans from Key West to Waities Island, South Carolina, with the institute managing the entire Florida subnetwork. Operating 160 atmospheric and more than 30 water-level stations across 32 counties in Florida, FAU leads the largest academic mesonet of its kind – a network of automated weather stations – in the Southeast and the fourth largest in the U.S.
Beyond Florida, the SEA Econet includes stations in Oklahoma (one weather station), Texas (two weather stations), Illinois (one weather station), and Georgia (one weather station and one water-level station). In South Carolina alone, the network manages 10 weather-only stations, 11 combined weather and water-level stations, and five water-level-only stations. Additionally, although not under direct management, the SEA Econet re-shares data from 65 other stations in South Carolina, further enhancing regional coverage.
“Every forecast that helps a family seek shelter, every alert that gives emergency crews time to mobilize – it all begins with accurate, real-time data from the ground,” said Jason Hallstrom, Ph.D., executive director of I-SENSE, which oversees the SEA Econet, and a professor in the FAU Department of Electrical Engineering and Computer Science within the College of Engineering and Computer Science. “That’s what we’ve built at Florida Atlantic: a statewide infrastructure that quietly powers some of the most critical decisions made during severe weather events. We’ve designed and deployed a system that delivers immense public value at a fraction of the typical cost.”
The infrastructure that FAU has developed is unique. Unlike other partners in the National Mesonet Program – many of which rely on millions in annual state support – FAU’s system was built entirely without direct state funding. Over the past 15 years, the university designed and engineered the technology in-house, with more than $8 million in federal research support from agencies like the National Science Foundation (NSF), NOAA and the Environmental Protection Agency (EPA). The network’s architecture was intentionally developed to dramatically reduce the cost of operation, enabling broader geographic coverage without expanding budgetary needs.
The result is a highly efficient, deeply integrated system that powers weather alerts, supports emergency response coordination and enhances public safety. The data collected by the institute’s network is used by the National Weather Service, the South Florida Water Management District, the National Park Service, and numerous other local, state and federal entities. Counties like Broward, Palm Beach, Brevard, Orange, Saint Lucie, Martin, Miami-Dade and Monroe all benefit from direct station coverage that enables hyper-local forecasting during critical events.
“FAU’s mesonet would not be possible without an extensive coalition of partners that includes the South Florida Water Management District, the Southeast Coastal Ocean Observing Regional Association, the Florida Fish and Wildlife Conservation Commission, the National Park Service, the Naval Sea Systems Command, SBA Communications, U.S. Sugar, and many others,” said Hallstrom. “These enduring partnerships reflect the broad value and utility of the network across sectors.”
To meet the increasing need and ensure the continued protection of Floridians, I-SENSE is working to expand its network from 160 to 445 stations over the next five years. This expansion would target high-risk and overlooked areas in Central and North Florida, where forecasting gaps remain. A dedicated operational team will also be established to support the growing network and develop new communication tools to provide real-time data access to emergency responders, government agencies and the public.
“There are important challenges here as Florida bears the brunt of damage from tropical storms and hurricanes,” said Hallstrom. “To keep Florida weather-ready, we plan to expand our network, invest in its longevity and ensure that every community – from the coasts to the heart of the state – has the data it needs to stay safe.”
Florida, among the most hurricane-prone states in the country, has already absorbed more than $400 billion in direct weather-related costs since 1980, the second-highest total in the nation. These storms pose a persistent threat to life and property, and significantly affect key industries such as real estate, tourism, insurance, health care, construction and agriculture, which together account for more than half of Florida’s GDP. The devastation of recent hurricanes Helene and Milton, with an estimated combined damage of more than $100 billion and 237 confirmed deaths, underscores the urgency for advanced forecasting infrastructure.
“With the right investment and continued support, Florida Atlantic is uniquely positioned to ensure that Florida not only leads the nation in storm preparedness but sets the global standard for weather forecasting,” said Stella Batalama, Ph.D., dean of the College of Engineering and Computer Science. “Our mesonet network spearheaded by I-SENSE, provides real-time, localized data that directly enhances our ability to predict and respond to hurricanes, flooding and other severe weather events. By expanding this network and continuing to innovate, we can equip communities with the most accurate, timely forecasts available, ultimately saving lives, reducing economic losses, and strengthening Florida’s resilience in the face of increasingly frequent and intense storms.”
- FAU -
A weather and water level station in a canal in Florida.
Nicholas Alberto, a student at FAU I-SENSE, completes a recent installation of an atmospheric station in Florida.
Credit
FAU I-SENSE
About FAU’s Sensing Institute (I-SENSE) Florida Atlantic University’ Sensing Institute (I-SENSE) is a university-wide research institute advancing innovation in sensing, smart systems, and real-time situational awareness technologies. As the hub for FAU’s strategic research emphasis in Sensing and Smart Systems, I-SENSE integrates cutting-edge research in sensing, computing, AI/ML, and wireless communication across disciplines and domains. With a mission to catalyze research excellence and deliver high-impact technological solutions, I-SENSE drives interdisciplinary collaboration across academia, industry, and government. From infrastructure systems and weather forecasting to health, behavior, and connected autonomy, I-SENSE-enabled technologies support improved decision-making, automated control, and fine-grained situational awareness at scale. Learn more at isense.fau.edu.
About FAU’s College of Engineering and Computer Science:
The FAU College of Engineering and Computer Science is internationally recognized for cutting-edge research and education in the areas of computer science and artificial intelligence (AI), computer engineering, electrical engineering, biomedical engineering, civil, environmental and geomatics engineering, mechanical engineering, and ocean engineering. Research conducted by the faculty and their teams expose students to technology innovations that push the current state-of-the art of the disciplines. The College research efforts are supported by the National Science Foundation (NSF), the National Institutes of Health (NIH), the Department of Defense (DOD), the Department of Transportation (DOT), the Department of Education (DOEd), the State of Florida, and industry. The FAU College of Engineering and Computer Science offers degrees with a modern twist that bear specializations in areas of national priority such as AI, cybersecurity, internet-of-things, transportation and supply chain management, and data science. New degree programs include Master of Science in AI (first in Florida), Master of Science and Bachelor in Data Science and Analytics, and the new Professional Master of Science and Ph.D. in computer science for working professionals. For more information about the College, please visit eng.fau.edu.
About Florida Atlantic University: Florida Atlantic University, established in 1961, officially opened its doors in 1964 as the fifth public university in Florida. Today, Florida Atlantic serves more than 30,000 undergraduate and graduate students across six campuses located along the Southeast Florida coast. In recent years, the University has doubled its research expenditures and outpaced its peers in student achievement rates. Through the coexistence of access and excellence, Florida Atlantic embodies an innovative model where traditional achievement gaps vanish. Florida Atlantic is designated as a Hispanic-serving institution, ranked as a top public university by U.S. News & World Report, and holds the designation of “R1: Very High Research Spending and Doctorate Production” by the Carnegie Classification of Institutions of Higher Education. Florida Atlantic shares this status with less than 5% of the nearly 4,000 universities in the United States. For more information, visit www.fau.edu.
New system enables robots to solve manipulation problems in seconds
Researchers developed an algorithm that lets a robot “think ahead” and consider thousands of potential motion plans simultaneously.
The researchers' robot planning approach considers thousands of possible actions simultaneously, enabling it to rapidly determine how to manipulate and tightly pack items without damaging them, like these blocks.
CAMBRIDGE, MA – Ready for that long-awaited summer vacation? First, you’ll need to pack all items required for your trip into a suitcase, making sure everything fits securely without crushing anything fragile.
Because humans possess strong visual and geometric reasoning skills, this is usually a straightforward problem, even if it may take a bit of finagling to squeeze everything in.
To a robot, though, it is an extremely complex planning challenge that requires thinking simultaneously about many actions, constraints, and mechanical capabilities. Finding an effective solution could take the robot a very long time — if it can even come up with one.
Researchers from MIT and NVIDIA Research have developed a novel algorithm that dramatically speeds up the robot’s planning process. Their approach enables a robot to “think ahead” by evaluating thousands of possible solutions in parallel and then refining the best ones to meet the constraints of the robot and its environment.
Instead of testing each potential action one at a time, like many existing approaches, this new method considers thousands of actions simultaneously, solving multistep manipulation problems in a matter of seconds.
The researchers harness the massive computational power of specialized processors called graphics processing units (GPUs) to enable this speedup.
In a factory or warehouse, their technique could enable robots to rapidly determine how to manipulate and tightly pack items that have different shapes and sizes without damaging them, knocking anything over, or colliding with obstacles, even in a narrow space.
“This would be very helpful in industrial settings where time really does matter and you need to find an effective solution as fast as possible. If your algorithm takes minutes to find a plan, as opposed to seconds, that costs the business money,” says MIT graduate student William Shen SM ’23, lead author of the paper on this technique.
The researchers’ algorithm is designed for what is called task and motion planning (TAMP). The goal of a TAMP algorithm is to come up with a task plan for a robot, which is a high-level sequence of actions, along with a motion plan, which includes low-level action parameters, like joint positions and gripper orientation, that complete that high-level plan.
To create a plan for packing items in a box, a robot needs to reason about many variables, such as the final orientation of packed objects so they fit together, as well as how it is going to pick them up and manipulate them using its arm and gripper.
It must do this while determining how to avoid collisions and achieve any user-specified constraints, such as a certain order in which to pack items.
With so many potential sequences of actions, sampling possible solutions at random and trying one at a time could take an extremely long time.
“It is a very large search space, and a lot of actions the robot does in that space don’t actually achieve anything productive,” Garrett adds.
Instead, the researchers’ algorithm, called cuTAMP, which is accelerated using a parallel computing platform called CUDA, simulates and refines thousands of solutions in parallel. It does this by combining two techniques, sampling and optimization.
Sampling involves choosing a solution to try. But rather than sampling solutions randomly, cuTAMP limits the range of potential solutions to those most likely to satisfy the problem’s constraints. This modified sampling procedure allows cuTAMP to broadly explore potential solutions while narrowing down the sampling space.
“Once we combine the outputs of these samples, we get a much better starting point than if we sampled randomly. This ensures we can find solutions more quickly during optimization,” Shen says.
Once cuTAMP has generated that set of samples, it performs a parallelized optimization procedure that computes a cost, which corresponds to how well each sample avoids collisions and satisfies the motion constraints of the robot, as well as any user-defined objectives.
It updates the samples in parallel, chooses the best candidates, and repeats the process until it narrows them down to a successful solution.
Harnessing accelerated computing
The researchers leverage GPUs, specialized processors that are far more powerful for parallel computation and workloads than general-purpose CPUs, to scale up the number of solutions they can sample and optimize simultaneously. This maximized the performance of their algorithm.
“Using GPUs, the computational cost of optimizing one solution is the same as optimizing hundreds or thousands of solutions,” Shen explains.
When they tested their approach on Tetris-like packing challenges in simulation, cuTAMP took only a few seconds to find successful, collision-free plans that might take sequential planning approaches much longer to solve.
And when deployed on a real robotic arm, the algorithm always found a solution in under 30 seconds.
The system works across robots and has been tested on a robotic arm at MIT and a humanoid robot at NVIDIA. Since cuTAMP is not a machine-learning algorithm, it requires no training data, which could enable it to be readily deployed in many situations.
“You can give it a brand-new problem and it will provably solve it,” Garrett says.
The algorithm is generalizable to situations beyond packing, like a robot using tools. A user could incorporate different skill types into the system to expand a robot’s capabilities automatically.
In the future, the researchers want to leverage large language models and vision language models within cuTAMP, enabling a robot to formulate and execute a plan that achieves specific objectives based on voice commands from a user.
###
This work is supported, in part, by the National Science Foundation (NSF), Air Force Office for Scientific Research, Office of Naval Research, MIT Quest for Intelligence, NVIDIA, and the Robotics and Artificial Intelligence Institute.
