A novel system using deep learning technology and custom electroencephalogram (EEG) data techniques has been shown to automatically learn seizure signatures in patients with epilepsy and reduce the amount of raw EEG data needed to be reviewed by a neurologist, according to a study published in EBioMedicine.

Researchers involved in the study obtained scalp EEG data from 365 patients with 171,745 s ictal and 2,185,864 s interictal samples. The data were analyzed during a crowdsourced artificial intelligence (AI) challenge.

Study participants were asked to develop deep learning models for automatic annotation of epileptic seizures in the raw EEG data. Specifically, these participants worked to develop an ictal/interictal classifier that featured high sensitivity as well as low false alarm rates. A challenge platform was subsequently built as a means of preventing the participants from downloading or accessing the relevant data.

Overall, findings indicated that the automatic detection system featured tunable sensitivities ranging from 75.00% to 91.60%. This achievement in high sensitivity rates reduced the amount of raw EEG data needed to be reviewed by a human annotator by maximum achievable reduction factor of 142x and 22x, respectively.

The study researchers stated the algorithm enabled “instantaneous reviewer-managed optimization of the balance between sensitivity” and the amount of raw EEG data that needed to be reviewed.

Ultimately, the deep learning technique allowed participants to learn patient-specific seizure signatures. The study researchers added that the system can then filter “seizure segments out of raw EEG data for verification by an expert neurologist.”

The primary limitation of this study included the relatively small sample size. Additional research is needed to validate the findings from this study.

Besides seizure detection, the study investigators concluded that “the platform enables collaboration between data scientists whilst keeping proprietary or sensitive data secure and protected.”

The post Deep Learning AI System Filters Seizure Verification Data appeared first on Artificial Intelligence.

AI-based assessment and grading in schools is a part of the new normal.

Once upon a time, there were real classrooms with teachers strictly evaluating you during your pen and paper examinations. This might be a story that our future kids will listen to if the ‘new normal’ is planning to stay here. The Covid-19 pandemic brought rapid digital transformation and AI-driven automation into all industries. According to IDC, the worldwide revenues for the artificial intelligence market are forecast to grow 16.4% in 2021 to USD 327.5 billion.

The growing significance of AI is also visible in the education sector. The pandemic-induced shift to online classes impacted many conventional methods of educational institutions. Remote learning has increased the accessibility and efficiency of the education systems. The shift towards virtual classrooms demands strong support from disruptive technologies. There have been many reports about the role of AI technologies in remote classes and virtual education systems and the ease it brings to different administrative tasks. Let us focus on one of the aspects where AI has already made its mark but has also raised some concerns, that is AI in exam evaluation.

AI in Evaluating Academic Performance

Recent reports revealed that Delhi’s new state school board, DBSE will be employing AI-based continuous assessment and game-based assessment in their schools. A Hindustan Times report states that AI in the assessment process will provide real-time learning feedback to teachers and students will be given situations or involved in activities to evaluate their skills and understanding through games. Will this mean that AI algorithms will decide the performance based on the skills and strength of the students? Absolutely, and this is not the first-ever approach. Many educational institutions and universities have already employed AI for assessment and automated grading systems. Schools in China had already started experimenting with automated AI grading systems a few years back.

AI-based assessments and marking techniques can minimize human biases and enhance the speed of evaluation. Instant feedback to both students and teachers is another advantage of AI in evaluations. Pearson, the multinational publishing and educational organization boasts having many AI-based assessment systems in ELT, which provides unbiased and accurate results.

AI paper grading softwares is gaining attention since they can quickly grade papers and assignments without any human intervention. Machine learning and data analytics are the pillars behind the technology of automated assessment. AI and Machine learning algorithms learn from existing data and try to replicate human evaluation patterns with precision.

AI-driven online marking and assessment, automated grading, AI-assisted proctoring are redefining the education system, while reducing bias and frauds. The Telangana State Board of Intermediate Education in India had announced the use of AI for accuracy in their exam results considering the discrepancies in evaluation that led to the suicides of many students. According to the Economic Times report, the board happened to find various errors in the assessment of OMR sheets due to flawed technology.

Are the Grades Reliable?

Despite the claims that AI-based assessment and grading is accurate, fair, and bias-free, there have been many instances where it went totally wrong. An article in the Harvard Business Review reveals how AI grading systems by the International Baccalaureate Organization produced varying results from the predicted ones and the students went on a protest. And, the AI system apparently, just predicted grades on the data fed to the algorithm rather than actually evaluating papers. Another article on The Verge says how a virtual learning platform’s AI-based assessment method focused on specific keywords to determine marks. Most of us would have heard how an automated AI algorithm caused havoc in the UK amid the pandemic by providing biased and lowered A-level results to the students.

We can find more such discrepancies if we dig more. Do these incidents indicate that AI-based exam evaluation systems are not an ideal approach after all? Maybe we are still lagging behind in addressing the flaws of AI and other disruptive technologies. While feeding loads of data to these systems, it must be taken care that the data do not represent any errors. The virtual and remote learning scenario is here to stay and AI will have many positive impacts on the education system. The future of AI is intertwined with us and hence, AI in the evaluation of exams will benefit many if incorporated in the right way.

The post STUDENTS CAN NOW ARGUE WITH AN AI SYSTEM FOR EXTRA MARKS appeared first on Artificial Intelligence.

What more can artificial intelligence do?

IBM has designed an artificial intelligence system that can debate with humans. The company published a paper in the journal called Nature, where one of the team members described the AI system and how well it performed against a human opponent. Chris Reed, a professor in the University of Dundee has published a News & Views article in the same journal throwing light on the history and development of artificial intelligence as a disruptive technology based around the types of logic used in human arguments and the system created by IBM.

As Reed explains in his piece, debating is a skill that humans have been perfecting for thousands of years. It’s usually a type of discussion in which a person or a group persuades others that their opinion on a subject is right. As a step forward, the IBM team has designed an AI system to debate with humans in a live setting. The system is capable of listening to the moderators and opponents and responds in a feminine tone.

Traditionally in debates, people involved tend to cite people who can back up their claims. They might find these citations from prior research or quote well-known phrases used by notable people in the field of argument. Project Debater, as the IBM system is known, scans the internet for similar arguments and uses the information in ways that are convincing.

Another common debate practice is participants attempting to counter the arguments of the opponent. To handle that, Project Debater uses Watson, the IBM system that defeated contestants on the popular game show “Jeopardy”. This system listens to the arguments given by the opponents and searches on the web for rebuttals that were used by others about similar claims.

Project Debater was first tested back in 2019 in a debate session with Harish Natarajan, an expert debater. In that debate, IBM’s system did not defeat Natarajan, but the audience agreed that it performed outstandingly.

Experts at IBM believe that Project Debaters’ failure against Natarajan was an important milestone in the efforts to get robotic systems to master human language. IBM research director Dario Gil said that the experience “is not about winning or losing” but about creating an AI system “that can master the infinitely complex and rich world of human language.”

The Watson computer system used by Project Debater defeated a grand chess master, apart from winning the game show. Ranit Aharanov, manager of the Project Debater team talked in a discussion after the debate. There she said that the goal of the project is to help humans deal with complex decisions.

“It’s not a question of whether AI is going to be better than humans. It can debate both sides, so it can very quickly help you understand both sides of the problem, so you have a wider view of the problem and can make a better decision.”

Project Debater went through another test where it was asked to convince a panel of viewers that telemedicine was a good idea. Almost everyone in the panel agreed that the AI system managed to change their opinions on the topic of conversation which implied that AI systems can participate in human debates in the future, for example, in those that happen on social media.

The post IBM HAS DESIGNED AN AI SYSTEM THAT CAN ALTER YOUR OPINIONS appeared first on Artificial Intelligence.

Artificial intelligence (AI) has been making great strides in recent years sometimes even coming close to being human-like. Now, in a new paper published in Nature magazine, IBM describes a system that can debate with humans and even sometimes win.

“Here we present Project Debater, an autonomous debating system that can engage in a competitive debate with humans,” write the authors. And the system is nothing short of extraordinary.

In tests of Project Debater, the AI was given only 15 minutes to research topics and prepare for debates. Each time, it proceeded to form an opening statement and even layer counterarguments. 

For the most part, the humans won the debate but in one instance it was able to change the stance of nine people. Not bad!

“Project Debater is a crucial step in the development of argument technology and in working with arguments as local phenomena. Its successes offer a tantalizing glimpse of how an AI system could work with the web of arguments that humans interpret with such apparent ease,” Chris Reed writes in a critique of the new project published in Nature magazine.

“Given the wildfires of fake news, the polarization of public opinion and the ubiquity of lazy reasoning, that ease belies an urgent need for humans to be supported in creating, processing, navigating and sharing complex arguments — support that AI might be able to supply.”

In other words, this new AI is not here to replace humans but rather to support them in building better arguments and reasoning with more nuance. If this subject interests you, Scientific American has done a great podcast episode with the research’s lead Noam Slonim which tackles amongst other things whether the AI actually understands the arguments it presents and what that means for the future of debating.

The post IBM Developed an AI System That Engages in Debates with Humans and Convinces Some appeared first on Artificial Intelligence.

As Industry 4.0 takes hold, industrial automation and robotics are replacing many manual tasks in manufacturing. However, when it comes to visual quality inspection, most production lines still employ human workers in the tedious task of examining products and judging defects.

The biggest drawback of manual visual inspection is that humans make mistakes. Tired workers often miss defects that ‘escape’ the quality screens on the production floor and leak into finished goods packages or into integrated systems. When these defects are discovered or surface at a later stage often by end customers, users or consumers, it is too late and very costly to fix. The Cost of Poor Quality (CoPQ) in these cases is significant.  It includes – among other elements – the costs of returned or rejected goods (RMA), scrap, rework and in many cases the negative impact on brand reputation and end customer dissatisfaction.

Israel based Kitov is paving the way towards smart manufacturing, by developing the technology to enable smart computer-driven visual inspection and support manufacturers along their digital transformation path.

KITOV ONE is a Smart 3D, Universal System that can effectively inspect virtually any product. Leveraging advanced 3D computer vision and deep-learning algorithms, KITOV ONE achieves unprecedented levels of detection, eliminating the tedious work and inconsistent results associated with manual inspection. KITOV supports complex 3D structures, numerous materials, and complete inspection specifications.

By imitating human learning processes, KITOV ONE features an intuitive method to teach the system how to optimally inspect almost any product.  Setting up the system does not require programming skills or knowledge of robotics or optics. KITOV ONE software computes and controls the processes of image acquisition and image processing by using pre-set algorithms called detectors. Artificial intelligence capabilities are used to find and classify defects.

“We have developed artificial intelligence (AI) systems for advanced manufacturing that can be intuitively trained within a few hours by a non-expert to automatically plan and perform sophisticated visual inspection tasks on complex 3D products at the highest performance levels.” states Dr. Yossi Rubner, CTO and Founder of Kitov.ai.

By using dashboards and Big Data Analytics Kitov helps manufacturers to identify trends and proactively attend to quality issues early on and by providing powerful insights about manufacturing process and product design can support root cause analysis and elimination of defects.

The post Smart 3D Universal Inspection System Uses Deep Learning appeared first on Artificial Intelligence.

Internet of Things Inc., a software and solutions provider in the artificial intelligence and industrial IoT markets, is pleased to announce that Health Canada has authorized the interim marketing and sales of its fever detection system, ThermalPass. Health Canada has done so under its fast-tracked Medical Device Establishment License (“MDEL”) application process established to help combat COVID-19.

ThermalPass, developed jointly by AI Labs Inc., a wholly owned subsidiary of ITT Inc. and Commersive Solutions Corp, is designed to enhance public safety to reduce the risk of spreading COVID-19 and other fever-bearing contagions. The system will provide fast, touch-free scanning of multiple people at entranceways of high-traffic locations including bus and train stations, schools, malls, office buildings, sports venues, and other public spaces.

“Regulatory approval from Health Canada to commercialize ThermalPass is a significant achievement as the Company prepares for imminent POC testing.” said Michael Lende, President and CEO of Internet of Things Inc. “We would like to thank Health Canada for their rapid review of our submission, allowing us to proceed with the rollout of ThermalPass.”

The ThermalPass system offers unique and distinctive competencies over other fever detecting devices as it uses sensors versus cameras. Infrared cameras are less accurate, more expensive, obtrusive to personal space and infringe on privacy. ThermalPass’ sensors are designed to measure temperature from a distance by detecting an object’s infrared energy. The company’s artificial-intelligence-powered system allows for fast, touch-free scanning of individuals as they pass through the device without impeding walking speed when entering high-traffic public locations. The system instantly and quietly alerts security personnel of any person with a higher-than-normal temperature, enabling staff to conduct a secondary check and maintain traffic flow.

The post Internet of Things Inc. Secures Health Canada Approval to Deploy Breakthrough Fever Detection System, ThermalPass appeared first on Artificial Intelligence.

Researchers at Google Research and the University of California, Berkeley, have recently developed an imitation learning system that could enable a variety of agile locomotion behaviors in robots. Their technique, presented in a paper pre-published on arXiv, allows robots to acquire new skills by imitating animals.

“This project builds on some previous works from computer graphics, which trained simulated characters to move by imitating human motion capture data,” Jason Peng, one of the researchers who carried out the study, told TechXplore. “Most of these techniques were primarily applied in simulation, but in our recent project we took a first step towards applying them to real robots.”

Peng and his colleagues initially trained a four-legged robot to imitate the movements and walking style of a dog within a simulated environment. Their system was trained on motion data recorded from a real dog, using an approach known as reinforcement learning.

“One of the advantages of training in simulation is that it is very fast, so we can simulate months of training in a matter of days,” Peng explained. “Once the robot has been trained in simulation, we can adapt what it has learned to a real robot, using only a few minutes of data collected in the real world.”

The imitation learning method employed by Peng and his colleagues is far more scalable than more traditional techniques for designing robotic controllers. In fact, instead of designing a new controller for every skill that one is trying to reproduce in robots, their approach can simply train robots to achieve specific locomotion styles by showing them a few examples of animals performing the desired movements. The robot can then automatically learn new locomotion skills simply by observing these examples.

Peng and his colleagues evaluated their approach in a series of experiments, training Laikago, a 18-DoF quadruped robot, to reproduce different animal locomotion behaviors, including different ways of running, hopping and turning. Remarkably, their technique allowed the robot to automatically synthesize controllers for a variety of animal locomotion styles, effectively transferring the skills it learned in simulated environments to the real world.

“The most exciting result for us was that the same underlying method can learn a pretty large variety of skills ranging from walking to dynamic hopping and turning and all of the skills learned in simulation can also be transferred to a real robot,” Peng said. “These imitation learning techniques could make it much easier to build large repertoires of skills for robots that can enable them to move and interact more agilely with the real world.”

In the future, the imitation learning system developed by Peng and his colleagues could enable a broader variety of agile movements in animal-inspired robots. Currently, their technique can only be trained using motion data, but the researchers are trying to develop it further, so that it can also learn from videos of animals.

“We are now interested in trying to get robots to imitate different kinds of motion data, such as video clips,” Peng said. “Motion capture data can sometimes be fairly difficult to record, especially from animals, as getting a dog into a mocap studio can be tricky. It would be great if we can just use our phones to record some video clips of what we want the robot to do and then have the robot learn how to reproduce those skills automatically.”

The post A system to reproduce different animal locomotion skills in robots appeared first on Artificial Intelligence.

A simulation system invented at MIT to train driverless cars creates a photorealistic world with infinite steering possibilities, helping the cars learn to navigate a host of worse-case scenarios before cruising down real streets.  

Control systems, or “controllers,” for autonomous vehicles largely rely on real-world datasets of driving trajectories from human drivers. From these data, they learn how to emulate safe steering controls in a variety of situations. But real-world data from hazardous “edge cases,” such as nearly crashing or being forced off the road or into other lanes, are — fortunately — rare.

Some computer programs, called “simulation engines,” aim to imitate these situations by rendering detailed virtual roads to help train the controllers to recover. But the learned control from simulation has never been shown to transfer to reality on a full-scale vehicle.

The MIT researchers tackle the problem with their photorealistic simulator, called Virtual Image Synthesis and Transformation for Autonomy (VISTA). It uses only a small dataset, captured by humans driving on a road, to synthesize a practically infinite number of new viewpoints from trajectories that the vehicle could take in the real world. The controller is rewarded for the distance it travels without crashing, so it must learn by itself how to reach a destination safely. In doing so, the vehicle learns to safely navigate any situation it encounters, including regaining control after swerving between lanes or recovering from near-crashes.  

In tests, a controller trained within the VISTA simulator safely was able to be safely deployed onto a full-scale driverless car and to navigate through previously unseen streets. In positioning the car at off-road orientations that mimicked various near-crash situations, the controller was also able to successfully recover the car back into a safe driving trajectory within a few seconds. A paper describing the system has been published in IEEE Robotics and Automation Letters and will be presented at the upcoming ICRA conference in May.

“It’s tough to collect data in these edge cases that humans don’t experience on the road,” says first author Alexander Amini, a PhD student in the Computer Science and Artificial Intelligence Laboratory (CSAIL). “In our simulation, however, control systems can experience those situations, learn for themselves to recover from them, and remain robust when deployed onto vehicles in the real world.”

The work was done in collaboration with the Toyota Research Institute. Joining Amini on the paper are Igor Gilitschenski, a postdoc in CSAIL; Jacob Phillips, Julia Moseyko, and Rohan Banerjee, all undergraduates in CSAIL and the Department of Electrical Engineering and Computer Science; Sertac Karaman, an associate professor of aeronautics and astronautics; and Daniela Rus, director of CSAIL and the Andrew and Erna Viterbi Professor of Electrical Engineering and Computer Science.

Data-driven simulation

Historically, building simulation engines for training and testing autonomous vehicles has been largely a manual task. Companies and universities often employ teams of artists and engineers to sketch virtual environments, with accurate road markings, lanes, and even detailed leaves on trees. Some engines may also incorporate the physics of a car’s interaction with its environment, based on complex mathematical models.

But since there are so many different things to consider in complex real-world environments, it’s practically impossible to incorporate everything into the simulator. For that reason, there’s usually a mismatch between what controllers learn in simulation and how they operate in the real world.

Instead, the MIT researchers created what they call a “data-driven” simulation engine that synthesizes, from real data, new trajectories consistent with road appearance, as well as the distance and motion of all objects in the scene.

They first collect video data from a human driving down a few roads and feed that into the engine. For each frame, the engine projects every pixel into a type of 3D point cloud. Then, they place a virtual vehicle inside that world. When the vehicle makes a steering command, the engine synthesizes a new trajectory through the point cloud, based on the steering curve and the vehicle’s orientation and velocity.

Then, the engine uses that new trajectory to render a photorealistic scene. To do so, it uses a convolutional neural network — commonly used for image-processing tasks — to estimate a depth map, which contains information relating to the distance of objects from the controller’s viewpoint. It then combines the depth map with a technique that estimates the camera’s orientation within a 3D scene. That all helps pinpoint the vehicle’s location and relative distance from everything within the virtual simulator.

Based on that information, it reorients the original pixels to recreate a 3D representation of the world from the vehicle’s new viewpoint. It also tracks the motion of the pixels to capture the movement of the cars and people, and other moving objects, in the scene. “This is equivalent to providing the vehicle with an infinite number of possible trajectories,” Rus says. “Because when we collect physical data, we get data from the specific trajectory the car will follow. But we can modify that trajectory to cover all possible ways of and environments of driving. That’s really powerful.”

Reinforcement learning from scratch

Traditionally, researchers have been training autonomous vehicles by either following human defined rules of driving or by trying to imitate human drivers. But the researchers make their controller learn entirely from scratch under an “end-to-end” framework, meaning it takes as input only raw sensor data — such as visual observations of the road — and, from that data, predicts steering commands at outputs.

“We basically say, ‘Here’s an environment. You can do whatever you want. Just don’t crash into vehicles, and stay inside the lanes,’” Amini says.

This requires “reinforcement learning” (RL), a trial-and-error machine-learning technique that provides feedback signals whenever the car makes an error. In the researchers’ simulation engine, the controller begins by knowing nothing about how  to drive, what a lane marker is, or even other vehicles look like, so it starts executing random steering angles. It gets a feedback signal only when it crashes. At that point, it gets teleported to a new simulated location and has to execute a better set of steering angles to avoid crashing again. Over 10 to 15 hours of training, it uses these sparse feedback signals to learn to travel greater and greater distances without crashing.

After successfully driving 10,000 kilometers in simulation, the authors apply that learned controller onto their full-scale autonomous vehicle in the real world. The researchers say this is the first time a controller trained using end-to-end reinforcement learning in simulation has successful been deployed onto a full-scale autonomous car. “That was surprising to us. Not only has the controller never been on a real car before, but it’s also never even seen the roads before and has no prior knowledge on how humans drive,” Amini says.

Forcing the controller to run through all types of driving scenarios enabled it to regain control from disorienting positions — such as being half off the road or into another lane — and steer back into the correct lane within several seconds. “And other state-of-the-art controllers all tragically failed at that, because they never saw any data like this in training,” Amini says.

Next, the researchers hope to simulate all types of road conditions from a single driving trajectory, such as night and day, and sunny and rainy weather. They also hope to simulate more complex interactions with other vehicles on the road. “What if other cars start moving and jump in front of the vehicle?” Rus says. “Those are complex, real-world interactions we want to start testing.”

The post System trains driverless cars in simulation before they hit the road appeared first on Artificial Intelligence.