探花直播 of Cambridge - machine learning

Fully AI driven weather prediction system could start revolution in forecasting

sc604 — Thu, 20 Mar 2025 15:56:54 +0000

探花直播system, Aardvark Weather, has been supported by the Alan Turing Institute, Microsoft Research and the European Centre for Medium Range Weather Forecasts. It provides a blueprint for a new approach to weather forecasting with the potential to transform current practices. The results are reported in the journal Nature.

鈥淎ardvark reimagines current weather prediction methods offering the potential to make weather forecasts faster, cheaper, more flexible and more accurate than ever before, helping to transform weather prediction in both developed and developing countries,鈥� said Professor Richard Turner from Cambridge鈥檚 Department of Engineering, who led the research. 鈥淎ardvark is thousands of times faster than all previous weather forecasting methods.鈥�

Current weather forecasts are generated through a complex set of stages, each taking several hours to run on powerful supercomputers. Aside from daily usage, the development, maintenance and use of these systems require significant time and large teams of experts.

More recently, research by Huawei, Google, and Microsoft has shown that one component of the weather forecasting pipeline, the numerical solver (which calculates how weather evolves over time), can be replaced with AI, resulting in faster and more accurate predictions. This combination of AI and traditional approaches is now being used by the European Centre for Medium Range Weather Forecasts (ECMWF).

But with Aardvark, researchers have replaced the entire weather prediction pipeline with a single, simple machine learning model. 探花直播new model takes in observations from satellites, weather stations and other sensors and outputs both global and local forecasts.

This fully AI driven approach means predictions that were once produced using many models 鈥� each requiring a supercomputer and a large support team to run 鈥� can now be produced in minutes on a desktop computer.

When using just 10% of the input data of existing systems, Aardvark already outperforms the United States national GFS forecasting system on many variables. It is also competitive with United States Weather Service forecasts that use input from dozens of weather models and analysis by expert human forecasters.

鈥淭hese results are just the beginning of what Aardvark can achieve,鈥� said first author Anna Allen, from Cambridge鈥檚 Department of Computer Science and Technology. 鈥淭his end-to-end learning approach can be easily applied to other weather forecasting problems, for example hurricanes, wildfires, and tornadoes. Beyond weather, its applications extend to broader Earth system forecasting, including air quality, ocean dynamics, and sea ice prediction.鈥�

探花直播researchers say that one of the most exciting aspects of Aardvark is its flexibility and simple design. Because it learns directly from data it can be quickly adapted to produce bespoke forecasts for specific industries or locations, whether that's predicting temperatures for African agriculture or wind speeds for a renewable energy company in Europe.

This contrasts to traditional weather prediction systems where creating a customised system takes years of work by large teams of researchers.

鈥� 探花直播weather forecasting systems we all rely on have been developed over decades, but in just 18 months, we鈥檝e been able to build something that鈥檚 competitive with the best of these systems, using just a tenth of the data on a desktop computer,鈥� said Turner, who is also Lead Researcher for Weather Prediction at the Alan Turing Institute.

This capability has the potential to transform weather prediction in developing countries where access to the expertise and computational resources required to develop conventional systems is not typically available.

鈥淯nleashing AI鈥檚 potential will transform decision-making for everyone from policymakers and emergency planners to industries that rely on accurate weather forecasts,鈥� said Dr Scott Hosking from 探花直播Alan Turing Institute. 鈥淎ardvark鈥檚 breakthrough is not just about speed, it鈥檚 about access. By shifting weather prediction from supercomputers to desktop computers, we can democratise forecasting, making these powerful technologies available to developing nations and data-sparse regions around the world.鈥�

鈥淎ardvark would not have been possible without decades of physical-model development by the community, and we are particularly indebted to ECMWF for their ERA5 dataset which is essential for training Aardvark,鈥� said Turner.

鈥淚t is essential that academia and industry work together to address technological challenges and leverage new opportunities that AI offers,鈥� said Matthew Chantry from ECMWF. 鈥淎ardvark鈥檚 approach combines both modularity with end-to-end forecasting optimisation, ensuring effective use of the available datasets."

鈥淎ardvark represents not only an important achievement in AI weather prediction but it also reflects the power of collaboration and bringing the research community together to improve and apply AI technology in meaningful ways,鈥� said Dr Chris Bishop, from Microsoft Research.

探花直播next steps for Aardvark include developing a new team within the Alan Turing Institute led by Turner, who will explore the potential to deploy Aardvark in the global south and integrate the technology into the Institute鈥檚 wider work to develop high-precision environmental forecasting for weather, oceans and sea ice.

Reference:
Anna Allen, Stratis Markou et al. 鈥�End-to-end data-driven weather prediction.鈥� Nature (2025). DOI: 10.1038/s41586-025-08897-0

Adapted from a media release by 探花直播Alan Turing Institute

A new AI weather prediction system, developed by researchers from the 探花直播 of Cambridge, can deliver accurate forecasts tens of times faster and using thousands of times less computing power than current AI and physics-based forecasting systems.

探花直播Alan Turing Institute

Professor Richard Turner using Aardvark Weather

探花直播text in this work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. Images, including our videos, are Copyright 漏探花直播 of Cambridge and licensors/contributors as identified. All rights reserved. We make our image and video content available in a number of ways 鈥� on our main website under its Terms and conditions, and on a range of channels including social media that permit your use and sharing of our content under their respective Terms.

Yes

Scientists develop 鈥榮mart pyjamas鈥� to monitor sleep disorders

sc604 — Tue, 18 Feb 2025 11:06:44 +0000

探花直播team, led by the 探花直播 of Cambridge, developed printed fabric sensors that can monitor breathing by detecting tiny movements in the skin, even when the pyjamas are worn loosely around the neck and chest.

探花直播sensors embedded in the smart pyjamas were trained using a 鈥榣ightweight鈥� AI algorithm and can identify six different sleep states with 98.6% accuracy, while ignoring regular sleep movements such as tossing and turning. 探花直播energy-efficient sensors only require a handful of examples of sleep patterns to successfully identify the difference between regular and disordered sleep.

探花直播researchers say that their smart pyjamas could be useful for the millions of people in the UK who struggle with disordered sleep to monitor their sleep, and how it might be affected by lifestyle changes. 探花直播results are reported in the Proceedings of the National Academy of Sciences (PNAS).

Sleep is vital for human health, yet more than 60% of adults experience poor sleep quality, leading to the loss of between 44 and 54 annual working days, and an estimated one percent reduction in global GDP. Sleep behaviours such as mouth breathing, sleep apnoea and snoring are major contributors to poor sleep quality, and can lead to chronic conditions such as cardiovascular disease, diabetes and depression.

鈥淧oor sleep has huge effects on our physical and mental health, which is why proper sleep monitoring is vital,鈥� said Professor Luigi Occhipinti from the Cambridge Graphene Centre, who led the research. 鈥淗owever, the current gold standard for sleep monitoring, polysomnography or PSG, is expensive, complicated and isn鈥檛 suitable for long-term use at home.鈥�

Home devices that are simpler than PSG, such as home sleep tests, typically focus on a single condition and are bulky or uncomfortable. Wearable devices such as smartwatches, while more comfortable to wear, can only infer sleep quality, and are not effective for accurately monitoring disordered sleep.

鈥淲e need something that is comfortable and easy to use every night, but is accurate enough to provide meaningful information about sleep quality,鈥� said Occhipinti.

To develop the smart pyjamas, Occhipinti and his colleagues built on their earlier work on a smart choker for people with speech impairments. 探花直播team re-designed the graphene-based sensors for breath analysis during sleep, and made several design improvements to increase sensitivity.

鈥淭hanks to the design changes we made, the sensors are able to detect different sleep states, while ignoring regular tossing and turning,鈥� said Occhinpinti. 鈥� 探花直播improved sensitivity also means that the smart garment does not need to be worn tightly around the neck, which many people would find uncomfortable. As long as the sensors are in contact with the skin, they provide highly accurate readings.鈥�

探花直播researchers designed a machine learning model, called SleepNet, that uses the signals captured by the sensors to identify sleep states including nasal breathing, mouth breathing, snoring, teeth grinding, central sleep apnoea (CSA), and obstructive sleep apnoea (OSA). SleepNet is a 鈥榣ightweight鈥� AI network, that reduces computational complexity to the point where it can be run on portable devices, without the need to connect to computers or servers.

鈥淲e pruned the AI model to the point where we could get the lowest computational cost with the highest degree of accuracy,鈥� said Occhinpinti. 鈥淭his way we are able to embed the main data processors in the sensors directly.鈥�

探花直播smart pyjamas were tested on healthy patients and those with sleep apnoea, and were able to detect a range of sleep states with an accuracy of 98.6%. By treating the smart pyjamas with a special starching step, they were able to improve the durability of the sensors so they can be run through a regular washing machine.

探花直播most recent version of the smart pyjamas are also capable of wireless data transfer, meaning the sleep data can be securely transferred to a smartphone or computer.

鈥淪leep is so important to health, and reliable sleep monitoring can be key in preventative care,鈥� said Occhipinti. 鈥淪ince this garment can be used at home, rather than in a hospital or clinic, it can alert users to changes in their sleep that they can then discuss with their doctor. Sleep behaviours such as nasal versus mouth breathing are not typically picked up in an NHS sleep analysis, but it can be an indicator of disordered sleep.鈥�

探花直播researchers are hoping to adapt the sensors for a range of health conditions or home uses, such as baby monitoring, and have been in discussions with different patient groups. They are also working to improve the durability of the sensors for long-term use.

探花直播research was supported in part by the EU Graphene Flagship, Haleon, and the Engineering and Physical Sciences Research Council (EPSRC), part of UK Research and Innovation (UKRI).

Reference:
Chenyu Tang, Wentian Yi et al. 鈥�A deep learning-enabled smart garment for accurate and versatile monitoring of sleep conditions in daily life.鈥� PNAS (2025). DOI: 10.1073/pnas.2420498122

Researchers have developed comfortable, washable 鈥榮mart pyjamas鈥� that can monitor sleep disorders such as sleep apnoea at home, without the need for sticky patches, cumbersome equipment or a visit to a specialist sleep clinic.

We need something that is comfortable and easy to use every night, but is accurate enough to provide meaningful information about sleep quality

Luigi Occhipinti

Illustration and photograph of 'smart pyjamas'

Yes

New datasets will train AI models to think like scientists

sc604 — Mon, 02 Dec 2024 15:59:08 +0000

探花直播initiative, called Polymathic AI, uses technology like that powering large language models such as OpenAI鈥檚 ChatGPT or Google鈥檚 Gemini. But instead of ingesting text, the project鈥檚 models learn using scientific datasets from across astrophysics, biology, acoustics, chemistry, fluid dynamics and more, essentially giving the models cross-disciplinary scientific knowledge.

鈥淭hese datasets are by far the most diverse large-scale collections of high-quality data for machine learning training ever assembled for these fields,鈥� said team member Michael McCabe from the Flatiron Institute in New York City. 鈥淐urating these datasets is a critical step in creating multidisciplinary AI models that will enable new discoveries about our universe.鈥�

On 2 December, the Polymathic AI team released two of its open-source training dataset collections to the public 鈥� a colossal 115 terabytes, from dozens of sources 鈥� for the scientific community to use to train AI models and enable new scientific discoveries. For comparison, GPT-3 used 45 terabytes of uncompressed, unformatted text for training, which ended up being around 0.5 terabytes after filtering.

探花直播full datasets are available to download for free on HuggingFace, a platform hosting AI models and datasets. 探花直播Polymathic AI team provides further information about the datasets in two papers accepted for presentation at the NeurIPS machine learning conference, to be held later this month in Vancouver, Canada.

鈥淛ust as LLMs such as ChatGPT learn to use common grammatical structure across languages, these new scientific foundation models might reveal deep connections across disciplines that we鈥檝e never noticed before,鈥� said Cambridge team lead聽Dr Miles Cranmer from Cambridge鈥檚 Institute of Astronomy. 鈥淲e might uncover patterns that no human can see, simply because no one has ever had both this breadth of scientific knowledge and the ability to compress it into a single framework.鈥�

AI tools such as machine learning are increasingly common in scientific research, and were recognised in two of this year鈥檚 Nobel Prizes. Still, such tools are typically purpose-built for a specific application and trained using data from that field. 探花直播Polymathic AI project instead aims to develop models that are truly polymathic, like people whose expert knowledge spans multiple areas. 探花直播project鈥檚 team reflects intellectual diversity, with physicists, astrophysicists, mathematicians, computer scientists and neuroscientists.

探花直播first of the two new training dataset collections focuses on astrophysics. Dubbed the Multimodal Universe, the dataset contains hundreds of millions of astronomical observations and measurements, such as portraits of galaxies taken by NASA鈥檚 James Webb Space Telescope and measurements of our galaxy鈥檚 stars made by the European Space Agency鈥檚 Gaia spacecraft.

探花直播other collection 鈥� called the Well 鈥� comprises over 15 terabytes of data from 16 diverse datasets. These datasets contain numerical simulations of biological systems, fluid dynamics, acoustic scattering, supernova explosions and other complicated processes.聽Cambridge researchers played a major role in developing both dataset collections, working alongside PolymathicAI and other international collaborators.

While these diverse datasets may seem disconnected at first, they all require the modelling of mathematical equations called partial differential equations. Such equations pop up in problems related to everything from quantum mechanics to embryo development and can be incredibly difficult to solve, even for supercomputers. One of the goals of the Well is to enable AI models to churn out approximate solutions to these equations quickly and accurately.

鈥淏y uniting these rich datasets, we can drive advancements in artificial intelligence not only for scientific discovery, but also for addressing similar problems in everyday life,鈥� said Ben Boyd, PhD student in the Institute of Astronomy.

Gathering the data for those datasets posed a challenge, said team member Ruben Ohana from the Flatiron Institute. 探花直播team collaborated with scientists to gather and create data for the project. 鈥� 探花直播creators of numerical simulations are sometimes sceptical of machine learning because of all the hype, but they鈥檙e curious about it and how it can benefit their research and accelerate scientific discovery,鈥� he said.

探花直播Polymathic AI team is now using the datasets to train AI models. In the coming months, they will deploy these models on various tasks to see how successful these well-rounded, well-trained AIs are at tackling complex scientific problems.

鈥淚t will be exciting to see if the complexity of these datasets can push AI models to go beyond merely recognising patterns, encouraging them to reason and generalise across scientific domains,鈥� said Dr Payel Mukhopadhyay from the Institute of Astronomy. 鈥淪uch generalisation is essential if we ever want to build AI models that can truly assist in conducting meaningful science.鈥�

鈥淯ntil now, haven鈥檛 had a curated scientific-quality dataset cover such a wide variety of fields,鈥� said Cranmer, who is also a member of Cambridge鈥檚 Department of Applied Mathematics and Theoretical Physics. 鈥淭hese datasets are opening the door to true generalist scientific foundation models for the first time. What new scientific principles might we discover? We're about to find out, and that's incredibly exciting.鈥�

探花直播Polymathic AI project is run by researchers from the Simons Foundation and its Flatiron Institute, New York 探花直播, the 探花直播 of Cambridge, Princeton 探花直播, the French Centre National de la Recherche Scientifique and the Lawrence Berkeley National Laboratory.

Members of the Polymathic AI team from the 探花直播 of Cambridge include PhD students, postdoctoral researchers and faculty across four departments: the Department of Applied Mathematics and Theoretical Physics, the Department of Pure Mathematics and Mathematical Statistics, the Institute of Astronomy and the Kavli Institute for Cosmology.

What can exploding stars teach us about how blood flows through an artery? Or swimming bacteria about how the ocean鈥檚 layers mix? A collaboration of researchers, including from the 探花直播 of Cambridge, has reached a milestone toward training artificial intelligence models to find and use transferable knowledge between fields to drive scientific discovery.

Alex Meng, Aaron Watters and the Well Collaboration

A mosaic of simulations included in the Well collection of datasets

Yes

AI algorithm accurately detects heart disease in dogs

sc604 — Tue, 29 Oct 2024 02:20:22 +0000

探花直播research team, led by the 探花直播 of Cambridge, adapted an algorithm originally designed for humans and found it could automatically detect and grade heart murmurs in dogs, based on audio recordings from digital stethoscopes. In tests, the algorithm detected heart murmurs with a sensitivity of 90%, a similar accuracy to expert cardiologists.

Heart murmurs are a key indicator of mitral valve disease, the most common heart condition in adult dogs. Roughly one in 30 dogs seen by a veterinarian has a heart murmur, although the prevalence is higher in small breed dogs and older dogs.

Since mitral valve disease and other heart conditions are so common in dogs, early detection is crucial as timely medication can extend their lives. 探花直播technology developed by the Cambridge team could offer an affordable and effective screening tool for primary care veterinarians, and improve quality of life for dogs. 探花直播results are reported in the Journal of Veterinary Internal Medicine.

鈥淗eart disease in humans is a huge health issue, but in dogs it鈥檚 an even bigger problem,鈥� said first author Dr Andrew McDonald from Cambridge鈥檚 Department of Engineering. 鈥淢ost smaller dog breeds will have heart disease when they get older, but obviously dogs can鈥檛 communicate in the same way that humans can, so it鈥檚 up to primary care vets to detect heart disease early enough so it can be treated.鈥�

Professor Anurag Agarwal, who led the research, is a specialist in acoustics and bioengineering. 鈥淎s far as we鈥檙e aware, there are no existing databases of heart sounds in dogs, which is why we started out with a database of heart sounds in humans,鈥� he said. 鈥淢ammalian hearts are fairly similar, and when things go wrong, they tend to go wrong in similar ways.鈥�

探花直播researchers started with a database of heart sounds from about 1000 human patients and developed a machine learning algorithm to replicate whether a heart murmur had been detected by a cardiologist. They then adapted the algorithm so it could be used with heart sounds from dogs.

探花直播researchers gathered data from almost 800 dogs who were undergoing routine heart examination at four veterinary specialist centres in the UK. All dogs received a full physical examination and heart scan (echocardiogram) by a cardiologist to grade any heart murmurs and identify cardiac disease, and heart sounds were recorded using an electronic stethoscope. By an order of magnitude, this is the largest dataset of dog heart sounds ever created.

鈥淢itral valve disease mainly affects smaller dogs, but to test and improve our algorithm, we wanted to get data from dogs of all shapes, sizes and ages,鈥� said co-author Professor Jose Novo Matos from Cambridge鈥檚 Department of Veterinary Medicine, a specialist in small animal cardiology. 鈥� 探花直播more data we have to train it, the more useful our algorithm will be, both for vets and for dog owners.鈥�

探花直播researchers fine-tuned the algorithm so it could both detect and grade heart murmurs based on the audio recordings, and differentiate between murmurs associated with mild disease and those reflecting advanced heart disease that required further treatment. 聽

鈥淕rading a heart murmur and determining whether the heart disease needs treatment requires a lot of experience, referral to a veterinary cardiologist, and expensive specialised heart scans,鈥� said Novo Matos. 鈥淲e want to empower general practitioners to detect heart disease and assess its severity to help owners make the best decisions for their dogs.鈥�

Analysis of the algorithm鈥檚 performance found it agreed with the cardiologist鈥檚 assessment in over half of cases, and in 90% of cases, it was within a single grade of the cardiologist鈥檚 assessment. 探花直播researchers say this is a promising result, as it is common for there to be significant variability in how different vets grade heart murmurs.

鈥� 探花直播grade of heart murmur is a useful differentiator for determining next steps and treatments, and we鈥檝e automated that process,鈥� said McDonald. 鈥淔or vets and nurses without as much stethoscope skill, and even those who are incredibly skilled with a stethoscope, we believe this algorithm could be a highly valuable tool.鈥�

In humans with valve disease, the only treatment is surgery, but for dogs, effective medication is available. 鈥淜nowing when to medicate is so important, in order to give dogs the best quality of life possible for as long as possible,鈥� said Agarwal. 鈥淲e want to empower vets to help make those decisions.鈥�

鈥淪o many people talk about AI as a threat to jobs, but for me, I see it as a tool that will make me a better cardiologist,鈥� said Novo Matos. 鈥淲e can鈥檛 perform heart scans on every dog in this country聽鈥� we just don鈥檛 have enough time or specialists to screen every dog with a murmur. But tools like these could help vets and owners, so we can quickly identify those dogs who are most in need of treatment.鈥�

探花直播research was supported in part by the Kennel Club Charitable Trust, the Medical Research Council, and Emmanuel College Cambridge.

Reference:
Andrew McDonald et al. 鈥�A machine learning algorithm to grade canine heart murmurs and stage preclinical myxomatous mitral valve disease.鈥� Journal of Veterinary Internal Medicine (2024). DOI: 10.1111/jvim.17224

Researchers have developed a machine learning algorithm to accurately detect heart murmurs in dogs, one of the main indicators of cardiac disease, which affects a large proportion of some smaller breeds such as King Charles Spaniels.

Jacqueline Garget

Huxley, a healthy volunteer Havanese, undergoes a physical examination at the Queen's Veterinary School Hospital, Cambridge.

Yes

探花直播 of Cambridge alumnus awarded 2024 Nobel Prize in Physics

Anonymous — Tue, 08 Oct 2024 10:11:25 +0000

Hinton (King鈥檚 1967) and Hopfield were awarded the prize 鈥榝or foundational discoveries and inventions that enable machine learning with artificial neural networks.鈥� Hinton, who is known as the 鈥楪odfather of AI鈥� is Emeritus Professor of Computer Science at the 探花直播 of Toronto.聽

This year鈥檚 two Nobel Laureates in Physics have used tools from physics to develop methods that are the foundation of today鈥檚 powerful machine learning. John Hopfield, a Guggenheim Fellow at the 探花直播 of Cambridge in 1968-1969, created an associative memory that can store and reconstruct images and other types of patterns in data. Geoffrey Hinton invented a method that can autonomously find properties in data, and perform tasks such as identifying specific elements in pictures.

When we talk about artificial intelligence, we often mean machine learning using artificial neural networks. This technology was originally inspired by the structure of the brain. In an artificial neural network, the brain鈥檚 neurons are represented by nodes that have different values. These nodes influence each other through con颅nections that can be likened to synapses and which can be made stronger or weaker. 探花直播network is trained, for example by developing stronger connections between nodes with simultaneously high values. This year鈥檚 laureates have conducted important work with artificial neural networks from the 1980s onward.

Geoffrey Hinton used a network invented by John Hopfield聽as the foundation for a new network: the Boltzmann machine. This can learn to recognise characteristic elements in a given type of data. Hinton used tools from statistical physics, the science of systems built from many similar components. 探花直播machine is trained by feeding it examples that are very likely to arise when the machine is run. 探花直播Boltzmann machine can be used to classify images or create new examples of the type of pattern on which it was trained. Hinton has built upon this work, helping initiate the current explosive development of machine learning.

Vice-Chancellor Professor Deborah Prentice said:

鈥淢any congratulations to Professor Hinton on receiving the Nobel Prize. Our alumni are a vital part of the Cambridge community, and many of them, like Professor Hinton, have made discoveries and advances that have genuinely changed our world. On behalf of the 探花直播 of Cambridge, I congratulate him on this enormous accomplishment.鈥�

鈥� 探花直播laureates鈥� work has already been of the greatest benefit. In physics we use artificial neural networks in a vast range of areas, such as developing new materials with specific properties,鈥� says Ellen Moons, Chair of the Nobel Committee for Physics. Hinton and Hopfield are the 122nd and 123rd Members of the 探花直播 of Cambridge to be awarded the Nobel Prize.聽

From 1980 to 1982, Hinton was a Scientific Officer at the MRC Applied Psychology Unit (as the MRC Cognition and Brain Sciences Unit was then known), before taking up a position at Carnegie Mellon 探花直播 in Pittsburgh.

In May 2023, Hinton gave a public lecture at the 探花直播's Centre for the Study of Existential Risk entitled 'Two Paths to Intelligence', in which he argued that "large scale digital computation is probably far better at acquiring knowledge than biological computation and may soon be much more intelligent than us".聽

Geoffrey Hinton, an alumnus of the 探花直播 of Cambridge, was awarded the 2024 Nobel Prize in Physics, jointly with John Hopfield of Princeton 探花直播.

Left: King's College Cambridge Right: Ill. Niklas Elmehed 漏 Nobel Prize Outreach

Right: Geoffrey Hinton (circled) at his Matriculation at King's College. Right: Illustration of Geoffrey Hinton

Yes

鈥楽mart choker鈥� uses AI to help people with speech impairment to communicate

sc604 — Fri, 13 Sep 2024 13:40:19 +0000

探花直播smart choker, developed by researchers at the 探花直播 of Cambridge, incorporates electronic sensors in a soft, stretchable fabric, and is comfortable to wear. 探花直播device could be useful for people who have temporary or permanent speech impairments, whether due to laryngeal surgery, or conditions such as Parkinson鈥檚, stroke or cerebral palsy.

By incorporating machine learning techniques, the smart choker can also successfully recognise differences in pronunciation, accent and vocabulary between users, reducing the amount of training required.

探花直播choker is a type of technology known as a silent speech interface, which analyses non-vocal signals to decode speech in silent conditions 鈥� the user only needs to mouth the words in order for them to be captured. 探花直播captured speech signals can then be transferred to a computer or speaker to facilitate conversation.

Tests of the smart choker showed it could recognise words with over 95% accuracy, while using 90% less computational energy than existing state-of-the art technologies. 探花直播results are reported in the journal npj Flexible Electronics.

鈥淐urrent solutions for people with speech impairments often fail to capture words and require a lot of training,鈥� said Dr Luigi Occhipinti from the Cambridge Graphene Centre, who led the research. 鈥淭hey are also rigid, bulky and sometimes require invasive surgery to the throat.鈥�

探花直播smart choker developed by Occhipinti and his colleagues outperforms current technologies on accuracy, requires less computing power, is comfortable for users to wear, and can be removed whenever it鈥檚 not needed. 探花直播choker is made from a sustainable bamboo-based textile, with strain sensors based on graphene ink incorporated in the fabric. When the sensors detect any strain, tiny, controllable cracks form in the graphene. 探花直播sensitivity of the sensors is more than four times higher than existing state of the art.

鈥淭hese sensors can detect tiny vibrations, such as those formed in the throat when whispering or even silently mouthing words, which makes them ideal for speech detection,鈥� said Occhipinti. 鈥淏y combining the ultra-high sensitivity of the sensors with highly efficient machine learning, we鈥檝e come up with a device we think could help a lot of people who struggle with their speech.鈥�

Vocal signals are incredibly complex, so associating a specific signal with a specific word requires a high level of computational processing. 鈥淥n top of that, every person is different in terms of the way they speak, and machine learning gives us the tools we need to learn and adapt the interpretation of signals from person to person,鈥� said Occhipinti.

探花直播researchers trained their machine learning model on a database of the most frequently used words in English, and selected words which are frequently confused with each other, such as 鈥榖ook鈥� and 鈥榣ook鈥�. 探花直播model was trained with a variety of users, including different genders, native and non-native English speakers, as well as people with different accents and different speaking speeds.

Thanks to the device鈥檚 ability to capture rich dynamic signal characteristics, the researchers found it possible to use lightweight neural network architectures with simplified depth and signal dimensions to extract and enhance the speech information features. This resulted in a machine learning model with high computational and energy efficiency, ideal for integration in battery-operated wearable devices with real-time AI processing capabilities.

鈥淲e chose to train the model with lots of different English speakers, so we could show it was capable of learning,鈥� said Occhipinti. 鈥淢achine learning has the capability to learn quickly and efficiently from one user to the next, so the retraining process is quick.鈥�

Tests of the smart choker showed it was 95.25% accurate in decoding speech. 鈥淚 was surprised at just how sensitive the device is,鈥� said Occhipinti. 鈥淲e couldn鈥檛 capture all the signals and complexity of human speech before, but now that we can, it unlocks a whole new set of potential applications.鈥�

Although the choker will have to undergo extensive testing and clinical trials before it is approved for use in patients with speech impairments, the researchers say that their smart choker could also be used in other health monitoring applications, or for improving communication in noisy or secure environments.

探花直播research was supported in part by the EU Graphene Flagship and the Engineering and Physical Sciences Research Council (EPSRC), part of UK Research and Innovation (UKRI).

Reference:
Chenyu Tang et al. 鈥�Ultrasensitive textile strain sensors redefine wearable silent speech interfaces with high machine learning efficiency.鈥� npj Flexible Electronics (2024). DOI: 10.1038/s41528-024-00315-1

Researchers have developed a wearable 鈥榮mart choker鈥� that uses a combination of flexible electronics and artificial intelligence techniques to allow people with speech impairments to communicate by detecting tiny movements in the throat.

Luigi Occhipinti

Smart Choker

Yes

New open-source platform allows users to evaluate performance of AI-powered chatbots

sc604 — Tue, 04 Jun 2024 10:34:36 +0000

A team of computer scientists, engineers, mathematicians and cognitive scientists, led by the 探花直播 of Cambridge, developed an open-source evaluation platform called CheckMate, which allows human users to interact with and evaluate the performance of large language models (LLMs).

探花直播researchers tested CheckMate in an experiment where human participants used three LLMs 鈥� InstructGPT, ChatGPT and GPT-4 鈥� as assistants for solving undergraduate-level mathematics problems.

探花直播team studied how well LLMs can assist participants in solving problems. Despite a generally positive correlation between a chatbot鈥檚 correctness and perceived helpfulness, the researchers also found instances where the LLMs were incorrect, but still useful for the participants. However, certain incorrect LLM outputs were thought to be correct by participants. This was most notable in LLMs optimised for chat.

探花直播researchers suggest models that communicate uncertainty, respond well to user corrections, and can provide a concise rationale for their recommendations, make better assistants. Human users of LLMs should verify their outputs carefully, given their current shortcomings.

探花直播results, reported in the Proceedings of the National Academy of Sciences (PNAS), could be useful in both informing AI literacy training, and help developers improve LLMs for a wider range of uses.

While LLMs are becoming increasingly powerful, they can also make mistakes and provide incorrect information, which could have negative consequences as these systems become more integrated into our everyday lives.

鈥淟LMs have become wildly popular, and evaluating their performance in a quantitative way is important, but we also need to evaluate how well these systems work with and can support people,鈥� said co-first author Albert Jiang, from Cambridge鈥檚 Department of Computer Science and Technology. 鈥淲e don鈥檛 yet have comprehensive ways of evaluating an LLM鈥檚 performance when interacting with humans.鈥�

探花直播standard way to evaluate LLMs relies on static pairs of inputs and outputs, which disregards the interactive nature of chatbots, and how that changes their usefulness in different scenarios. 探花直播researchers developed CheckMate to help answer these questions, designed for but not limited to applications in mathematics.

鈥淲hen talking to mathematicians about LLMs, many of them fall into one of two main camps: either they think that LLMs can produce complex mathematical proofs on their own, or that LLMs are incapable of simple arithmetic,鈥� said co-first author Katie Collins from the Department of Engineering. 鈥淥f course, the truth is probably somewhere in between, but we wanted to find a way of evaluating which tasks LLMs are suitable for and which they aren鈥檛.鈥�

探花直播researchers recruited 25 mathematicians, from undergraduate students to senior professors, to interact with three different LLMs (InstructGPT, ChatGPT, and GPT-4) and evaluate their performance using CheckMate. Participants worked through undergraduate-level mathematical theorems with the assistance of an LLM and were asked to rate each individual LLM response for correctness and helpfulness. Participants did not know which LLM they were interacting with.

探花直播researchers recorded the sorts of questions asked by participants, how participants reacted when they were presented with a fully or partially incorrect answer, whether and how they attempted to correct the LLM, or if they asked for clarification. Participants had varying levels of experience with writing effective prompts for LLMs, and this often affected the quality of responses that the LLMs provided.

An example of an effective prompt is 鈥渨hat is the definition of X鈥� (X being a concept in the problem) as chatbots can be very good at retrieving concepts they know of and explaining it to the user.

鈥淥ne of the things we found is the surprising fallibility of these models,鈥� said Collins. 鈥淪ometimes, these LLMs will be really good at higher-level mathematics, and then they鈥檒l fail at something far simpler. It shows that it鈥檚 vital to think carefully about how to use LLMs effectively and appropriately.鈥�

However, like the LLMs, the human participants also made mistakes. 探花直播researchers asked participants to rate how confident they were in their own ability to solve the problem they were using the LLM for. In cases where the participant was less confident in their own abilities, they were more likely to rate incorrect generations by LLM as correct.

鈥淭his kind of gets to a big challenge of evaluating LLMs, because they鈥檙e getting so good at generating nice, seemingly correct natural language, that it鈥檚 easy to be fooled by their responses,鈥� said Jiang. 鈥淚t also shows that while human evaluation is useful and important, it鈥檚 nuanced, and sometimes it鈥檚 wrong. Anyone using an LLM, for any application, should always pay attention to the output and verify it themselves.鈥�

Based on the results from CheckMate, the researchers say that newer generations of LLMs are increasingly able to collaborate helpfully and correctly with human users on undergraduate-level maths problems, as long as the user can assess the correctness of LLM-generated responses. Even if the answers may be memorised and can be found somewhere on the internet, LLMs have the advantage of being flexible in their inputs and outputs over traditional search engines (though should not replace search engines in their current form).

While CheckMate was tested on mathematical problems, the researchers say their platform could be adapted to a wide range of fields. In the future, this type of feedback could be incorporated into the LLMs themselves, although none of the CheckMate feedback from the current study has been fed back into the models.

鈥淭hese kinds of tools can help the research community to have a better understanding of the strengths and weaknesses of these models,鈥� said Collins. 鈥淲e wouldn鈥檛 use them as tools to solve complex mathematical problems on their own, but they can be useful assistants if the users know how to take advantage of them.鈥�

探花直播research was supported in part by the Marshall Commission, the Cambridge Trust, Peterhouse, Cambridge, 探花直播Alan Turing Institute, the European Research Council, and the Engineering and Physical Sciences Research Council (EPSRC), part of UK Research and Innovation (UKRI).

聽

Reference:
Katherine M聽Collins, Albert Q聽Jiang, et al. 鈥�Evaluating Language Models for Mathematics through Interactions.鈥� Proceedings of the National Academy of Sciences (2024). DOI: 10.1073/pnas.2318124121

聽

Researchers have developed a platform for the interactive evaluation of AI-powered chatbots such as ChatGPT.聽

Anyone using an LLM, for any application, should always pay attention to the output and verify it themselves

Albert Jiang

da-kuk via Getty Images

Chatbot

Yes

Training AI models to answer 鈥榳hat if?鈥� questions could improve medical treatments

sc604 — Fri, 19 Apr 2024 08:02:54 +0000

Artificial intelligence techniques can be helpful for multiple medical applications, such as radiology or oncology, where the ability to recognise patterns in large volumes of data is vital. For these types of applications, the AI compares information against learned examples, draws conclusions, and makes extrapolations.

Now, an international team led by researchers from Ludwig-Maximilians-Universit盲t M眉nchen (LMU) and including researchers from the 探花直播 of Cambridge, is exploring the potential of a comparatively new branch of AI for diagnostics and therapy.

探花直播researchers found that causal machine learning (ML) can estimate treatment outcomes 鈥� and do so better than the machine learning methods generally used to date. Causal machine learning makes it easier for clinicians to personalise treatment strategies, which individually improves the health of patients.

探花直播results, reported in the journal Nature Medicine, suggest how causal machine learning could improve the effectiveness and safety of a variety of medical treatments.

Classical machine learning recognises patterns and discovers correlations. However, the principle of cause and effect remains closed to machines as a rule; they cannot address the question of why. When making therapy decisions for a patient, the 鈥榳hy鈥� is vital to achieve the best outcomes.

鈥淒eveloping machine learning tools to address why and what if questions is empowering for clinicians, because it can strengthen their decision-making processes,鈥� said senior author Professor Mihaela van der Schaar, Director of the Cambridge Centre for AI in Medicine. 鈥淏ut this sort of machine learning is far more complex than assessing personalised risk.鈥�

For example, when attempting to determine therapy decisions for someone at risk of developing diabetes, classical ML would aim to predict how probable it is for a given patient with a range of risk factors to develop the disease. With causal ML, it would be possible to answer how the risk changes if the patient receives an anti-diabetes drug; that is, gauge the effect of a cause. It would also be possible to estimate whether metformin, the commonly-prescribed medication, would be the best treatment, or whether another treatment plan would be better.

To be able to estimate the effect of a hypothetical treatment, the AI models must learn to answer 鈥榳hat if?鈥� questions. 鈥淲e give the machine rules for recognising the causal structure and correctly formalising the problem,鈥� said Professor Stefan Feuerriegel from LMU, who led the research. 鈥淭hen the machine has to learn to recognise the effects of interventions and understand, so to speak, how real-life consequences are mirrored in the data that has been fed into the computers.鈥�

Even in situations for which reliable treatment standards do not yet exist or where randomised studies are not possible for ethical reasons because they always contain a placebo group, machines could still gauge potential treatment outcomes from the available patient data and form hypotheses for possible treatment plans, so the researchers hope.

With such real-world data, it should generally be possible to describe the patient cohorts with ever greater precision in the estimates, bringing individualised therapy decisions that much closer. Naturally, there would still be the challenge of ensuring the reliability and robustness of the methods.

鈥� 探花直播software we need for causal ML methods in medicine doesn鈥檛 exist out of the box,鈥� says Feuerriegel. 鈥淩ather, complex modelling of the respective problem is required, involving close collaboration between AI experts and doctors.鈥�

In other fields, such as marketing, explains Feuerriegel, the work with causal ML has already been in the testing phase for some years now. 鈥淥ur goal is to bring the methods a step closer to practice,鈥� he said. 探花直播paper describes the direction in which things could move over the coming years.鈥�

鈥淚 have worked in this area for almost 10 years, working relentlessly in our lab with generations of students to crack this problem,鈥� said van der Schaar, who is affiliated with the Departments of Applied Mathematics and Theoretical Physics, Engineering and Medicine. 鈥淚t鈥檚 an extremely challenging area of machine learning, and seeing it come closer to clinical use, where it will empower clinicians and patients alike, is very satisfying.鈥�

Van der Schaar is continuing to work closely with clinicians to validate these tools in diverse clinical settings, including transplantation, cancer and cardiovascular disease.

Reference:
Stefan Feuerriegel et al. 鈥�Causal machine learning for predicting treatments.鈥� Nature Medicine (2024). DOI: 10.1038/s41591-024-02902-1

Adapted from an LMU media release.

Machines can learn not only to make predictions, but to handle causal relationships. An international research team shows how this could make medical treatments safer, more efficient, and more personalised.

Yuichiro Chino via Getty Images

Computer-generated image of human brain

Yes