Computers Math

Most artificially intelligent systems are based on neural networks, algorithms inspired by biological neurons found in the brain. These networks can consist of multiple layers, with inputs coming in one side and outputs going out of the other. The outputs can be used to make automatic decisions, for example, in driverless cars. Attacks to mislead a neural network can involve exploiting vulnerabilities in the input layers, but typically only the initial input layer is considered when engineering a defense. For the first time, researchers augmented a neural network’s inner layers with a process involving random noise to improve its resilience.
Artificial intelligence (AI) has become a relatively common thing; chances are you have a smartphone with an AI assistant or you use a search engine powered by AI. While it’s a broad term that can include many different ways to essentially process information and sometimes make decisions, AI systems are often built using artificial neural networks (ANN) analogous to those of the brain. And like the brain, ANNs can sometimes get confused, either by accident or by the deliberate actions of a third party. Think of something like an optical illusion — it might make you feel like you are looking at one thing when you are really looking at another.
The difference between things that confuse an ANN and things that might confuse us, however, is that some visual input could appear perfectly normal, or at least might be understandable to us, but may nevertheless be interpreted as something completely different by an ANN. A trivial example might be an image-classifying system mistaking a cat for a dog, but a more serious example could be a driverless car mistaking a stop signal for a right-of-way sign. And it’s not just the already controversial example of driverless cars; there are medical diagnostic systems, and many other sensitive applications that take inputs and inform, or even make, decisions that can affect people.
As inputs aren’t necessarily visual, it’s not always easy to analyze why a system might have made a mistake at a glance. Attackers trying to disrupt a system based on ANNs can take advantage of this, subtly altering an anticipated input pattern so that it will be misinterpreted, and the system will behave wrongly, perhaps even problematically. There are some defense techniques for attacks like these, but they have limitations. Recent graduate Jumpei Ukita and Professor Kenichi Ohki from the Department of Physiology at the University of Tokyo Graduate School of Medicine devised and tested a new way to improve ANN defense.
“Neural networks typically comprise layers of virtual neurons. The first layers will often be responsible for analyzing inputs by identifying the elements that correspond to a certain input,” said Ohki. “An attacker might supply an image with artifacts that trick the network into misclassifying it. A typical defense for such an attack might be to deliberately introduce some noise into this first layer. This sounds counterintuitive that it might help, but by doing so, it allows for greater adaptations to a visual scene or other set of inputs. However, this method is not always so effective and we thought we could improve the matter by looking beyond the input layer to further inside the network.”
Ukita and Ohki aren’t just computer scientists. They have also studied the human brain, and this inspired them to use a phenomenon they knew about there in an ANN. This was to add noise not only to the input layer, but to deeper layers as well. This is typically avoided as it’s feared that it will impact the effectiveness of the network under normal conditions. But the duo found this not to be the case, and instead the noise promoted greater adaptability in their test ANN, which reduced its susceptibility to simulated adversarial attacks.
“Our first step was to devise a hypothetical method of attack that strikes deeper than the input layer. Such an attack would need to withstand the resilience of a network with a standard noise defense on its input layer. We call these feature-space adversarial examples,” said Ukita. “These attacks work by supplying an input intentionally far from, rather than near to, the input that an ANN can correctly classify. But the trick is to present subtly misleading artifacts to the deeper layers instead. Once we demonstrated the danger from such an attack, we injected random noise into the deeper hidden layers of the network to boost their adaptability and therefore defensive capability. We are happy to report it works.”
While the new idea does prove robust, the team wishes to develop it further to make it even more effective against anticipated attacks, as well as other kinds of attacks they have not yet tested it against. At present, the defense only works on this specific kind of attack.
“Future attackers might try to consider attacks that can escape the feature-space noise we considered in this research,” said Ukita. “Indeed, attack and defense are two sides of the same coin; it’s an arms race that neither side will back down from, so we need to continually iterate, improve and innovate new ideas in order to protect the systems we use every day.” More

The feel of a cat’s fur can reveal some information, but seeing the feline provides critical details: is it a housecat or a lion? While the sound of fire crackling may be ambiguous, its scent confirms the burning wood. Our senses synergize to give a comprehensive understanding, particularly when individual signals are subtle. The collective sum of biological inputs can be greater than their individual contributions. Robots tend to follow more straightforward addition, but Penn State researchers have now harnessed the biological concept for application in artificial intelligence (AI) to develop the first artificial, multisensory integrated neuron.
Led by Saptarshi Das, associate professor of engineering science and mechanics at Penn State, the team published their work on September 15 in Nature Communication.
“Robots make decisions based on the environment they are in, but their sensors do not generally talk to each other,” said Das, who also has joint appointments in electrical engineering and in materials science and engineering. “A collective decision can be made through a sensor processing unit, but is that the most efficient or effective method? In the human brain, one sense can influence another and allow the person to better judge a situation.”
For instance, a car might have one sensor scanning for obstacles, while another senses darkness to modulate the intensity of the headlights. Individually, these sensors relay information to a central unit which then instructs the car to brake or adjust the headlights. According to Das, this process consumes more energy. Allowing sensors to communicate directly with each other can be more efficient in terms of energy and speed — particularly when the inputs from both are faint.
“Biology enables small organisms to thrive in environments with limited resources, minimizing energy consumption in the process,” said Das, who is also affiliated with the Materials Research Institute. “The requirements for different sensors are based on the context — in a dark forest, you’d rely more on listening than seeing, but we don’t make decisions based on just one sense. We have a complete sense of our surroundings, and our decision making is based on the integration of what we’re seeing, hearing, touching, smelling, etcetera. The senses evolved together in biology, but separately in AI. In this work, we’re looking to combine sensors and mimic how our brains actually work.”
The team focused on integrating a tactile sensor and a visual sensor so that the output of one sensor modifies the other, with the help of visual memory. According to Muhtasim Ul Karim Sadaf, a third-year doctoral student in engineering science and mechanics, even a short-lived flash of light can significantly enhance the chance of successful movement through a dark room.
“This is because visual memory can subsequently influence and aid the tactile responses for navigation,” Sadaf said. “This would not be possible if our visual and tactile cortex were to respond to their respective unimodal cues alone. We have a photo memory effect, where light shines and we can remember. We incorporated that ability into a device through a transistor that provides the same response.”
The researchers fabricated the multisensory neuron by connecting a tactile sensor to a phototransistor based on a monolayer of molybdenum disulfide, a compound that exhibits unique electrical and optical characteristics useful for detecting light and supporting transistors. The sensor generates electrical spikes in a manner reminiscent of neurons processing information, allowing it to integrate both visual and tactile cues. More

Soft inflatable robots have emerged as a promising paradigm for applications that require inherent safety and adaptability. However, the integration of sensing and control systems in these robots has posed significant challenges without compromising their softness, form factor, or capabilities. Addressing this obstacle, a research team jointly led by Professor Jiyun Kim (Department of New Material Engineering, UNIST) and Professor Jonbum Bae (Department of Mechanical Engineering, UNIST) has developed groundbreaking “soft valve” technology — an all-in-one solution that integrates sensors and control valves while maintaining complete softness.
Traditionally, soft robot bodies coexisted with rigid electronic components for perception purposes. The study conducted by this research team introduces a novel approach to overcome this limitation by creating soft analogs of sensors and control valves that operate without electricity. The resulting tube-shaped part serves dual functions: detecting external stimuli and precisely controlling driving motion using only air pressure. By eliminating the need for electricity-dependent components, these all-soft valves enable safe operation underwater or in environments where sparks may pose risks — while simultaneously reducing weight burdens on robotic systems. Moreover, each component is inexpensive at approximately 800 Won.
“Previous soft robots had flexible bodies but relied on hard electronic parts for stimulus detection sensors and drive control units,” explained Professor Kim. “Our study focuses on making both sensors and drive control parts using soft materials.”
The research team showcased various applications utilizing this groundbreaking technology. They created universal tongs capable of delicately picking up fragile items such as potato chips — preventing breakage caused by excessive force exerted by conventional rigid robot hands. Additionally, they successfully employed these all-soft components to develop wearable elbow assist robots designed to reduce muscle burden caused by repetitive tasks or strenuous activities involving arm movements. The elbow support automatically adjusts according to the angle at which an individual’s arm is bent — a breakthrough contributing to a 63% average decrease in the force exerted on the elbow when wearing the robot.
The soft valve operates by utilizing air flow within a tube-shaped structure. When tension is applied to one end of the tube, a helically wound thread inside compresses it, controlling inflow and outflow of air. This accordion-like motion allows for precise and flexible movements without relying on electrical power.
Furthermore, the research team confirmed that by programming different structures or numbers of threads within the tube, they could accurately control airflow variations. This programmability enables customized adjustments to suit specific situations and requirements — providing flexibility in driving unit response even with consistent external forces applied to the end of the tube.
“These newly developed components can be easily employed using material programming alone, eliminating electronic devices,” expressed Professor Bae with excitement about this development. “This breakthrough will significantly contribute to advancements in various wearable systems.”
This groundbreaking soft valve technology marks a significant step toward fully soft, electronics-free robots capable of autonomous operation — a crucial milestone for enhancing safety and adaptability across numerous industries.
Support for this work was provided by various organizations including Korea’s National Research Foundation (NRF), Korea Institute of Materials Science (KIMS), and Korea Evaluation Institute of Industrial Technology (KEIT). More

A major new study has revealed that American teenagers are more likely than any other nationality to brag about their math ability.
Research using data from 40,000 15-year-olds from nine English-speaking nations internationally found those in North America were the most likely to exaggerate their mathematical knowledge, while those in Ireland and Scotland were least likely to do so.
The study, published in the peer-reviewed journal Assessment in Education: Principles, Policy & Practice, used responses from the OECD Programme for International Student Assessment (PISA), in which participants took a two-hour maths test alongside a 30-minute background questionnaire.
They were asked how familiar they were with each of 16 mathematical terms — but three of the terms were fake.
Further questions revealed those who claimed familiarity with non-existent mathematical concepts were also more likely to display overconfidence in their academic prowess, problem-solving skills and perseverance.
For instance, they claimed higher levels of competence in calculating a discount on a television and in finding their way to a destination. Two thirds of those most likely to overestimate their mathematical ability were confident they could work out the petrol consumption of a car, compared to just 40 per cent of those least likely to do so.
Those likely to over-claim were also more likely to say if their mobile phone stopped sending texts they would consult a manual (41 per cent versus 30 per cent) while those less likely to do so tended to say they would react by pressing all the buttons (56 per cent versus 49 per cent). More

As 3D printers have become cheaper and more widely accessible, a rapidly growing community of novice makers are fabricating their own objects. To do this, many of these amateur artisans access free, open-source repositories of user-generated 3D models that they download and fabricate on their 3D printer.
But adding custom design elements to these models poses a steep challenge for many makers, since it requires the use of complex and expensive computer-aided design (CAD) software, and is especially difficult if the original representation of the model is not available online. Plus, even if a user is able to add personalized elements to an object, ensuring those customizations don’t hurt the object’s functionality requires an additional level of domain expertise that many novice makers lack.
To help makers overcome these challenges, MIT researchers developed a generative-AI-driven tool that enables the user to add custom design elements to 3D models without compromising the functionality of the fabricated objects. A designer could utilize this tool, called Style2Fab, to personalize 3D models of objects using only natural language prompts to describe their desired design. The user could then fabricate the objects with a 3D printer.
“For someone with less experience, the essential problem they faced has been: Now that they have downloaded a model, as soon as they want to make any changes to it, they are at a loss and don’t know what to do. Style2Fab would make it very easy to stylize and print a 3D model, but also experiment and learn while doing it,” says Faraz Faruqi, a computer science graduate student and lead author of a paper introducing Style2Fab.
Style2Fab is driven by deep-learning algorithms that automatically partition the model into aesthetic and functional segments, streamlining the design process.
In addition to empowering novice designers and making 3D printing more accessible, Style2Fab could also be utilized in the emerging area of medical making. Research has shown that considering both the aesthetic and functional features of an assistive device increases the likelihood a patient will use it, but clinicians and patients may not have the expertise to personalize 3D-printable models.
With Style2Fab, a user could customize the appearance of a thumb splint so it blends in with her clothing without altering the functionality of the medical device, for instance. Providing a user-friendly tool for the growing area of DIY assistive technology was a major motivation for this work, adds Faruqi. More

The era of artificial-intelligence chatbots that seem to understand and use language the way we humans do has begun. Under the hood, these chatbots use large language models, a particular kind of neural network. But a new study shows that large language models remain vulnerable to mistaking nonsense for natural language. To a team of researchers at Columbia University, it’s a flaw that might point toward ways to improve chatbot performance and help reveal how humans process language.
In a paper published online today in Nature Machine Intelligence, the scientists describe how they challenged nine different language models with hundreds of pairs of sentences. For each pair, people who participated in the study picked which of the two sentences they thought was more natural, meaning that it was more likely to be read or heard in everyday life. The researchers then tested the models to see if they would rate each sentence pair the same way the humans had.
In head-to-head tests, more sophisticated AIs based on what researchers refer to as transformer neural networks tended to perform better than simpler recurrent neural network models and statistical models that just tally the frequency of word pairs found on the internet or in online databases. But all the models made mistakes, sometimes choosing sentences that sound like nonsense to a human ear.
“That some of the large language models perform as well as they do suggests that they capture something important that the simpler models are missing,” said Dr. Nikolaus Kriegeskorte, PhD, a principal investigator at Columbia’s Zuckerman Institute and a coauthor on the paper. “That even the best models we studied still can be fooled by nonsense sentences shows that their computations are missing something about the way humans process language.”
Consider the following sentence pair that both human participants and the AI’s assessed in the study:
That is the narrative we have been sold.
This is the week you have been dying. More

Capturing blur-free images of fast movements like falling water droplets or molecular interactions requires expensive ultrafast cameras that acquire millions of images per second. In a new paper, researchers report a camera that could offer a much less expensive way to achieve ultrafast imaging for a wide range of applications such as real-time monitoring of drug delivery or high-speed lidar systems for autonomous driving.
“Our camera uses a completely new method to achieve high-speed imaging,” said Jinyang Liang from the Institut national de la recherche scientifique (INRS) in Canada. “It has an imaging speed and spatial resolution similar to commercial high-speed cameras but uses off-the-shelf components that would likely cost less than a tenth of today’s ultrafast cameras, which can start at close to $100,000.”
In Optica, Optica Publishing Group’s journal for high-impact research, Liang together with collaborators from Concordia University in Canada and Meta Platforms Inc. show that their new diffraction-gated real-time ultrahigh-speed mapping (DRUM) camera can capture a dynamic event in a single exposure at 4.8 million frames per second. They demonstrate this capability by imaging the fast dynamics of femtosecond laser pulses interacting with liquid and laser ablation in biological samples.
“In the long term, I believe that DRUM photography will contribute to advances in biomedicine and automation-enabling technologies such as lidar, where faster imaging would allow more accurate sensing of hazards,” said Liang. “However, the paradigm of DRUM photography is quite generic. In theory, it can be used with any CCD and CMOS cameras without degrading their other advantages such as high sensitivity.”
Creating a better ultrafast camera
Despite a great deal of progress in ultrafast imaging, today’s methods are still expensive and complex to implement. Their performance is also limited by trade-offs between the number of frames captured in each movie and light throughput or temporal resolution. To overcome these issues, the researchers developed a new time-gating method known as time-varying optical diffraction.
Cameras use gates to control when light hits the sensor. For example, the shutter in a traditional camera is a type of gate that opens and closes once. In time-gating, the gate is opened and closed in quick succession a certain number of times before the sensor reads out the image. This captures a short high-speed movie of a scene. More

Scientists have confirmed that human brains are naturally wired to perform advanced calculations, much like a high-powered computer, to make sense of the world through a process known as Bayesian inference.
In a study published in the journal Nature Communications, researchers from the University of Sydney, University of Queensland and University of Cambridge developed a specific mathematical model that closely matches how human brains work when it comes to reading vision. The model contained everything needed to carry out Bayesian inference.
Bayesian inference is a statistical method that combines prior knowledge with new evidence to make intelligent guesswork. For example, if you know what a dog looks like and you see a furry animal with four legs, you might use your prior knowledge to guess it’s a dog.
This inherent capability enables people to interpret the environment with extraordinary precision and speed, unlike machines that can be bested by simple CAPTCHA security measures when prompted to identify fire hydrants in a panel of images.
The study’s senior investigator Dr Reuben Rideaux, from the University of Sydney’s School of Psychology, said: “Despite the conceptual appeal and explanatory power of the Bayesian approach, how the brain calculates probabilities is largely mysterious.”
“Our new study sheds light on this mystery. We discovered that the basic structure and connections within our brain’s visual system are set up in a way that allows it to perform Bayesian inference on the sensory data it receives.
“What makes this finding significant is the confirmation that our brains have an inherent design that allows this advanced form of processing, enabling us to interpret our surroundings more effectively.”
The study’s findings not only confirm existing theories about the brain’s use of Bayesian-like inference but open doors to new research and innovation, where the brain’s natural ability for Bayesian inference can be harnessed for practical applications that benefit society. More