Breakthroughs In Artificial Intelligence Research Articles

Advances in intelligent and autonomous navigation systems for small UAS

A significant growth in Unmanned Aircraft System (UAS) operations has been observed over the past decade, largely driven by the emergence of new commercial opportunities and use-cases. This has posed new technological and regulatory challenges in order to address the complex safety, efficiency and sustainability requirements associated with UAS operations in an increasingly congested airspace. The growing need for trusted autonomy in UAS operations imposes demanding performance requirements on Navigation and Guidance Systems (NGS), both in terms of accuracy, integrity, continuity and availability. In most current NGS implementations, system autonomy is tightly constrained within a specified set of operational and environmental conditions through a large number of explicit rules. Recent breakthroughs in Artificial Intelligence (AI)-based methods and the emergence of highly-parallelized processor boards with low form-factor has led to the opportunity to employ Machine Learning (ML) techniques to enhance navigation system performance, particularly for small UAS (sUAS), which account for the majority of current and future unmanned aircraft use-cases. sUAS navigation systems typically employ diverse low Size, Weight, Power and Cost (SWaP-C) sensors such as Global Navigation Satellite System (GNSS) receivers, MEMS-IMUs, magnetometers, cameras and Lidars for localization, obstacle detection and avoidance. This paper presents a comprehensive review of conventional sUAS navigation systems, including aspects such as system architecture, sensing modalities and data-fusion algorithms. Additionally, performance monitoring and augmentation strategies are critically reviewed and assessed against current and future UAS Traffic Management (UTM) requirements. The primary focus is on the identification of key gaps in the literature where the use of AI-based methods can potentially enhance navigation performance. A critical review of AI-based methods and their application to sUAS navigation is conducted, along with an assessment of the performance benefits they provide over conventional navigation systems. Reviewed methods include but are not restricted to Artificial Neural Networks (ANN) such as Convolutional and Recurrent Neural Networks (CNN and RNN), Support Vector Machines (SVM) and ensemble techniques. The key challenges associated with adapting these methods to address sUAS operational objectives are clearly identified. The review also covers the assurance of predictable, deterministic system behaviour which is a key requirement to support system certification. The review and analysis will inform the reader of the applicability of various AI/ML methods to sUAS navigation and autonomous system integrity monitoring, and its likely role in the ongoing UTM evolution.

From Brain Science to Artificial Intelligence

Reviewing the history of the development of artificial intelligence (AI) clearly reveals that brain science has resulted in breakthroughs in AI, such as deep learning. At present, although the developmental trend in AI and its applications has surpassed expectations, an insurmountable gap remains between AI and human intelligence. It is urgent to establish a bridge between brain science and AI research, including a link from brain science to AI, and a connection from knowing the brain to simulating the brain. The first steps toward this goal are to explore the secrets of brain science by studying new brain-imaging technology; to establish a dynamic connection diagram of the brain; and to integrate neuroscience experiments with theory, models, and statistics. Based on these steps, a new generation of AI theory and methods can be studied, and a subversive model and working mode from machine perception and learning to machine thinking and decision-making can be established. This article discusses the opportunities and challenges of adapting brain science to AI.

Machine Learning Applied to Electrified Vehicle Battery State of Charge and State of Health Estimation: State-of-the-Art

The growing interest and recent breakthroughs in artificial intelligence and machine learning (ML) have actively contributed to an increase in research and development of new methods to estimate the states of electrified vehicle batteries. Data-driven approaches, such as ML, are becoming more popular for estimating the state of charge (SOC) and state of health (SOH) due to greater availability of battery data and improved computing power capabilities. This paper provides a survey of battery state estimation methods based on ML approaches such as feedforward neural networks (FNNs), recurrent neural networks (RNNs), support vector machines (SVM), radial basis functions (RBF), and Hamming networks. Comparisons between methods are shown in terms of data quality, inputs and outputs, test conditions, battery types, and stated accuracy to give readers a bigger picture view of the ML landscape for SOC and SOH estimation. Additionally, to provide insight into how to best approach with the comparison of different neural network structures, an FNN and long short-term memory (LSTM) RNN are trained fifty times each for 3000 epochs. The error is somewhat different for each training repetition due to the random initial values of the trainable parameters, demonstrating that it is important to train networks multiple times to achieve the best result. Furthermore, it is recommended that when performing a comparison among estimation techniques such as those presented in this review paper, the compared networks should have a similar number of learnable parameters and be trained and tested with identical data. Otherwise, it is difficult to make a general conclusion regarding the quality of a given estimation technique.

Intelligent fault diagnosis for rotating machinery using deep Q-network based health state classification: A deep reinforcement learning approach

Fault diagnosis methods for rotating machinery have always been a hot research topic, and artificial intelligence-based approaches have attracted increasing attention from both researchers and engineers. Among those related studies and methods, artificial neural networks, especially deep learning-based methods, are widely used to extract fault features or classify fault features obtained by other signal processing techniques. Although such methods could solve the fault diagnosis problems of rotating machinery, there are still two deficiencies. (1) Unable to establish direct linear or non-linear mapping between raw data and the corresponding fault modes, the performance of such fault diagnosis methods highly depends on the quality of the extracted features. (2) The optimization of neural network architecture and parameters, especially for deep neural networks, requires considerable manual modification and expert experience, which limits the applicability and generalization of such methods. As a remarkable breakthrough in artificial intelligence, AlphaGo, a representative achievement of deep reinforcement learning, provides inspiration and direction for the aforementioned shortcomings. Combining the advantages of deep learning and reinforcement learning, deep reinforcement learning is able to build an end-to-end fault diagnosis architecture that can directly map raw fault data to the corresponding fault modes. Thus, based on deep reinforcement learning, a novel intelligent diagnosis method is proposed that is able to overcome the shortcomings of the aforementioned diagnosis methods. Validation tests of the proposed method are carried out using datasets of two types of rotating machinery, rolling bearings and hydraulic pumps, which contain a large number of measured raw vibration signals under different health states and working conditions. The diagnosis results show that the proposed method is able to obtain intelligent fault diagnosis agents that can mine the relationships between the raw vibration signals and fault modes autonomously and effectively. Considering that the learning process of the proposed method depends only on the replayed memories of the agent and the overall rewards, which represent much weaker feedback than that obtained by the supervised learning-based method, the proposed method is promising in establishing a general fault diagnosis architecture for rotating machinery.

The Extended Corporate Mind: When Corporations Use AI to Break the Law

Algorithms may soon replace employees as the leading cause of corporate harm. For centuries, the law has defined corporate misconduct — anything from civil discrimination to criminal insider trading — in terms of employee misconduct. Today, however, breakthroughs in artificial intelligence and big data allow automated systems to make many corporate decisions, e.g., who gets a loan or what stocks to buy. These technologies introduce valuable efficiencies, but they do not remove (or even always reduce) the incidence of corporate harm. Unless the law adapts, corporations will become increasingly immune to civil and criminal liability as they transfer responsibility from employees to algorithms. This Article is the first to tackle the full extent of the growing doctrinal gap left by algorithmic corporate misconduct. To hold corporations accountable, the law must sometimes treat them as if they “know” information stored on their servers and “intend” decisions reached by their automated systems. Cognitive science and the philosophy of mind offer a path forward. The “extended mind thesis” complicates traditional views about the physical boundaries of the mind. The thesis states that the mind encompasses any system that sufficiently assists thought, e.g. by facilitating recall or enhancing decision-making. For natural people, the thesis implies that minds can extend beyond the brain to include external cognitive aids, like rolodexes and calculators. This Article adapts the thesis to corporate law. It motivates and proposes a doctrinal framework for extending the corporate mind to the algorithms that are increasingly integral to corporate thought. The law needs such an innovation if it is to hold future corporations to account for their most serious harms.

DeepSphere: Efficient spherical convolutional neural network with HEALPix sampling for cosmological applications

Convolutional Neural Networks (CNNs) are a cornerstone of the Deep Learning toolbox and have led to many breakthroughs in Artificial Intelligence. These networks have mostly been developed for regular Euclidean domains such as those supporting images, audio, or video. Because of their success, CNN-based methods are becoming increasingly popular in Cosmology. Cosmological data often comes as spherical maps, which make the use of the traditional CNNs more complicated. The commonly used pixelization scheme for spherical maps is the Hierarchical Equal Area isoLatitude Pixelisation (HEALPix). We present a spherical CNN for analysis of full and partial HEALPix maps, which we call DeepSphere. The spherical CNN is constructed by representing the sphere as a graph. Graphs are versatile data structures that can act as a discrete representation of a continuous manifold. Using the graph-based representation, we define many of the standard CNN operations, such as convolution and pooling. With filters restricted to being radial, our convolutions are equivariant to rotation on the sphere, and DeepSphere can be made invariant or equivariant to rotation. This way, DeepSphere is a special case of a graph CNN, tailored to the HEALPix sampling of the sphere. This approach is computationally more efficient than using spherical harmonics to perform convolutions. We demonstrate the method on a classification problem of weak lensing mass maps from two cosmological models and compare the performance of the CNN with that of two baseline classifiers. The results show that the performance of DeepSphere is always superior or equal to both of these baselines. For high noise levels and for data covering only a smaller fraction of the sphere, DeepSphere achieves typically 10% better classification accuracy than those baselines. Finally, we show how learned filters can be visualized to introspect the neural network.

Artificial intelligence and upper gastrointestinal endoscopy: Current status and future perspective.

With recent breakthroughs in artificial intelligence, computer-aided diagnosis (CAD) for upper gastrointestinal endoscopy is gaining increasing attention. Main research focuses in this field include automated identification of dysplasia in Barrett's esophagus and detection of early gastric cancers. By helping endoscopists avoid missing and mischaracterizing neoplastic change in both the esophagus and the stomach, these technologies potentially contribute to solving current limitations of gastroscopy. Currently, optical diagnosis of early-stage dysplasia related to Barrett's esophagus can be precisely achieved only by endoscopists proficient in advanced endoscopic imaging, and the false-negative rate for detecting gastric cancer is approximately 10%. Ideally, these novel technologies should work during real-time gastroscopy to provide on-site decision support for endoscopists regardless of their skill; however, previous studies of these topics remain exvivo and experimental in design. Therefore, the feasibility, effectiveness, and safety of CAD for upper gastrointestinal endoscopy in clinical practice remain unknown, although a considerable number of pilot studies have been conducted by both engineers and medical doctors with excellent results. This review summarizes current publications relating to CAD for upper gastrointestinal endoscopy from the perspective of endoscopists and aims to indicate what is required for future research and implementation in clinical practice.

Improvement of a Deep Learning Algorithm for Total Electron Content Maps: Image Completion

AbstractObservations with a global coverage are very important for space physics research and space weather monitoring. However, due to the technical limitations, it would be very expensive or even impossible to achieve a seamless global coverage even with advanced observational devices. It would be useful to fill missing data gaps to create a global map from the available data, but up until now this has been very challenging. Fortunately, the deep learning method, a recent breakthrough in artificial intelligence, may provide an effective way to solve this problem by making full use of data from reliable observations. In this paper, a promising deep learning algorithm, deep convolutional generative adversarial network (DCGAN), is investigated to fill the missing data of total electron content (TEC) map images. The direct use of DCGAN fails to fill missing data for the completion of TEC maps because there are always missing TEC data in some regions, such as oceans, where the features vary with time and geophysical conditions. Thus, no useful information can be utilized by DCGAN to achieve a meaningful image completion. In order to overcome this shortcoming of the original DCGAN method, a novel regularized DCGAN (R‐DCGAN) is proposed by adding an extra discriminator and some widely used reference TEC maps from the International Global Navigation Satellite Systems Service Ionosphere Working Group. The proposed R‐DCGAN method generates satisfactory ionospheric peak structures at different times and geomagnetic conditions, which demonstrate its effectiveness on filling the missing data in TEC maps. The proposed R‐DCGAN framework can be readily extended to a broad application in other fields of space sciences, particularly for addressing the missing observation data issues.

Artificial intelligence in the lab: ask not what your computer can do for you.

Artificial intelligence in the lab: ask not what your computer can do for you.

Using Artificial Intelligence Technology for Social Determinants and Risk Factors Surveillance

Session topic areaData science methods: machine learning in risk factor surveillance
 Overall objectives or goalBackgroundDecades of research have shown that factors such as living conditions, and not just medical treatments and lifestyles, are strongly associated with the health of individuals and populations. These distal factors (social, economic, cultural and environmental) are collectively called the social determinants of health (SDOH), and affect health inequities (i.e. differences in health outcomes that are avoidable, unfair and unjust). Gathering data on both risk factors (biomedical/clinical) and SDOH is of the utmost importance to quantify their contribution in disease causation at individual and population levels.
 Social determinants of health and biomedical/clinical risk factors surveillance (collectively termed as “risk factor surveillance”) refers to the monitoring of distal and proximal factors that impact the health of individuals and populations and health equity. It offers the opportunity to “forecast” population health, potential disease incidence, and guide intervention programs to prevent disease manifestation. However, current risk factor surveillance data is limited in geographical representation, completion, and content and time. Identifying novel methods of collecting risk factors and SDOH data can allow for opportunities for population health and disease forecasting using high quality, nationally-representative, real-time data.
 Recent breakthroughs in artificial intelligence (AI), such as speech and image recognition, offers new opportunities to develop novel methods to collect risk factor information at individual levels. Meanwhile, we can use intelligent computer systems to process vast amount of data and turn those data into actionable information and knowledge for improving population health.
 Collaborative Session ObjectiveThrough a CIHR-funded project, we are assembling a team of national and international experts including stakeholders, public health officers/physicians, and researchers, who will identify key gaps in risk factor surveillance and data collection technologies. Resulting projects will focus on using AI for risk factor surveillance, for the ultimate purpose of monitoring population health, guiding intervention programs, and preventing disease. Our projects will focus on discovering and refining innovative methods in data collection, management, as well as assessment of data quality (i.e. selection bias). We will engage scientists and knowledge users from the inception of the ideas to ensure the relevancy of the final projects. This project aims to link medical records, clinical information, and SDOH data, to alter the way we conduct surveillance and work with big data.
 Facilitators involved; home institutions
 Dr. Vineet Saini, University of Calgary; Alberta Health Services
 Dr. Mingkai Peng, University of Calgary
 Dr. Hude Quan, University of Calgary; World Health Organization Collaborating Centre for Classification, Measurement and Standardization
 Intended output or outcome
 
 Identify AI technologies for use in risk factor surveillance; innovative methods in data collection, management, as well as assessment of data quality (i.e. selection bias); uses for new data sources in improving health equity.
 Create partnerships between national and international experts in risk factor surveillance

Artificial superintelligence: a futuristic approach

A day does not go by without a news article reporting some amazing breakthrough in artificial intelligence (AI). Many philosophers, futurists, and AI researchers have conjectured that human-level AI will be developed in the next 20 to 200 years. If these predictions are correct, it raises new and sinister issues related to our future in the age of intelligent machines. Artificial Superintelligence: A Futuristic Approach directly addresses these issues and consolidates research aimed at making sure that emerging superintelligence is beneficial to humanity. While specific predictions regarding the consequences of superintelligent AI vary from potential economic hardship to the complete extinction of humankind, many researchers agree that the issue is of utmost importance and needs to be seriously addressed. Artificial Superintelligence: A Futuristic Approach discusses key topics such as: AI-Completeness theory and how it can be used to see if an artificial intelligent agent has attained human level intelligence Methods for safeguarding the invention of a superintelligent system that could theoretically be worth trillions of dollars Self-improving AI systems: definition, types, and limits The science of AI safety engineering, including machine ethics and robot rights Solutions for ensuring safe and secure confinement of superintelligent systems The future of superintelligence and why long-term prospects for humanity to remain as the dominant species on Earth are not great Artificial Superintelligence: A Futuristic Approach is designed to become a foundational text for the new science of AI safety engineering. AI researchers and students, computer security researchers, futurists, and philosophers should find this an invaluable resource.

Deep Learning in Object Recognition, Detection, and Segmentation

As a major breakthrough in artificial intelligence, deep learning has achieved very impressive success in solving grand challenges in many fields including speech recognition, natural language processing, computer vision, image and video processing, and multimedia. This article provides a historical overview of deep learning and focus on its applications in object recognition, detection, and segmentation, which are key challenges of computer vision and have numerous applications to images and videos. The discussed research topics on object recognition include image classification on ImageNet, face recognition, and video classification. The detection part covers general object detection on ImageNet, pedestrian detection, face landmark detection face alignment, and human landmark detection pose estimation. On the segmentation side, thearticle discusses the most recent progress on scene labeling, semantic segmentation, face parsing, human parsing and saliency detection. Object recognition is considered as whole-image classification, while detection and segmentation are pixelwise classification tasks. Their fundamental differences will be discussed in this article. Fully convolutional neural networks and highly efficient forward and backward propagation algorithms specially designed for pixelwise classification task will be introduced. The covered application domains are also much diversified. Human and face images have regular structures, while general object and scene images have much more complex variations in geometric structures and layout. Videos include the temporal dimension. Therefore, they need to be processed with different deep models. All the selected domain applications have received tremendous attentions in the computer vision and multimedia communities. Through concrete examples of these applications, we explain the key points which make deep learning outperform conventional computer vision systems. 1 Different than traditional pattern recognition systems, which heavily rely on manually designed features, deep learning automatically learns hierarchical feature representations from massive training data and disentangles hidden factors of input data through multi-level nonlinear mappings. 2 Different than existing pattern recognition systems which sequentially design or train their key components, deep learning is able to jointly optimize all the components and crate synergy through close interactions among them. 3 While most machine learning models can be approximated with neural networks with shallow structures, for some tasks, the expressive power of deep models increases exponentially as their architectures go deep. Deep models are especially good at learning global contextual feature representation with their deep structures. 4 Benefitting from the large learning capacity of deep models, some classical computer vision challenges can be recast as high-dimensional data transform problems and can be solved from new perspectives. Finally, some open questions and future works regarding to deep learning in object recognition, detection, and segmentation will be discussed.

Deep neural networks: A promising tool for fault characteristic mining and intelligent diagnosis of rotating machinery with massive data

Aiming to promptly process the massive fault data and automatically provide accurate diagnosis results, numerous studies have been conducted on intelligent fault diagnosis of rotating machinery. Among these studies, the methods based on artificial neural networks (ANNs) are commonly used, which employ signal processing techniques for extracting features and further input the features to ANNs for classifying faults. Though these methods did work in intelligent fault diagnosis of rotating machinery, they still have two deficiencies. (1) The features are manually extracted depending on much prior knowledge about signal processing techniques and diagnostic expertise. In addition, these manual features are extracted according to a specific diagnosis issue and probably unsuitable for other issues. (2) The ANNs adopted in these methods have shallow architectures, which limits the capacity of ANNs to learn the complex non-linear relationships in fault diagnosis issues. As a breakthrough in artificial intelligence, deep learning holds the potential to overcome the aforementioned deficiencies. Through deep learning, deep neural networks (DNNs) with deep architectures, instead of shallow ones, could be established to mine the useful information from raw data and approximate complex non-linear functions. Based on DNNs, a novel intelligent method is proposed in this paper to overcome the deficiencies of the aforementioned intelligent diagnosis methods. The effectiveness of the proposed method is validated using datasets from rolling element bearings and planetary gearboxes. These datasets contain massive measured signals involving different health conditions under various operating conditions. The diagnosis results show that the proposed method is able to not only adaptively mine available fault characteristics from the measured signals, but also obtain superior diagnosis accuracy compared with the existing methods.

Applications of Intelligent Evolutionary Algorithms in Optimal Automation System Design

This paper proposes an intelligent evolutionary algorithm that can be applied in the design of optimal automation systems, and employs a multimodal six-bar mechanism optimization design, job shop production scheduling for the fishing equipment industry, and dynamic real-time production scheduling system design cases to show how the technique developed in this paper is highly effective at resolving optimal automation system design problems. Major breakthroughs in artificial intelligence continue to be made in the wake of advanced information technology developments, and the field of intelligent evolutionary algorithms has attracted a particularly large amount of attention from researchers and users in the artificial intelligence community. The successful optimization of automation system design requires interdisciplinary integration, and further requires the use of actual cases, verification, and improvement to ensure implementation in real-world applications.

Artificial intelligence research at IBM

IBM has a history of achieving seminal results in artificial intelligence, such as the first machine translation program, the first competitive game-playing computer, and the first speech recognition system that addressed the general dictation problem. This paper describes the unique characteristics of IBM that enabled researchers to continuously achieve breakthroughs in artificial intelligence.

Education for the 21st Century

What skills will students need to succeed in the 21st century? Mr. Posner tries to look ahead and suggests that our current education policies may be leading us in the wrong direction. A CARTOON from Gary Larson's weirdly insightful feature, Far Side, shows doting parents watching their son play Nintendo while visions of his lucrative future as a Nintendo master dance in their heads. Would this be less funny if instead of Nintendo their son were doing long division and his parents had visions of his lucrative future as a long divider? Before you answer, let me tell you that I know several talented engineers who can trace their careers back to their interest in video games. For them, there was a natural progression from enthusiasm for the games to wondering how such games are designed and implemented to wanting to make games of their own. I've never met anyone who was similarly motivated by long division. I asked one of our engineers, a graduate of MIT, when was the last time he had used long division. He said that, in fact, he had used it only a couple of years earlier, when he and a fellow alum decided to determine whether they still knew how. (They did!) It is a tricky business trying to guess what experiences will motivate an individual to intellectual achievement or what skills or bits of knowledge will wind up being important in a person's life. There are at least two major factors that complicate such predictions. One is the absurd diversity of human beings. I am intimately acquainted with four children who share a common ethnic, cultural, and socioeconomic background. In fact, they were born of the same parents and raised in the same house. And yet with respect to talents, interests, rates of development, and personalities, these children are completely different. By the of our society, each of these children is weird. I'd feel sorry for their parents, except that all the parents I've ever known, including my own, have equally strange children. (Remember Stephen Jay Gould's message: norms are not real.1) It seems likely that the propensity for diversity is a consequence of natural selection. It has been fundamental to the ability of our species to adapt to radically changing environments, particularly those of our own making. The second major factor that complicates the prediction of what skills will be important -- and the one on which most attention seems to focus -- is technology. How many of the skills we spent so much time learning are now done better by machine? Suppose Gary Larson's kid were shown solving differential equations. The cartoon would still be funny. Symbolic math software available for your personal computer is orders of magnitude better than you and I are at solving equations. Automatic spell checkers have eliminated the need to memorize the spelling rules we spent so much time learning. How important is handwriting these days? How long do you think it will be before speech recognition reaches the point where manual entry of text in any form is unnecessary? And search engines have revolutionized and simplified research. None of this technology has required breakthroughs in artificial intelligence. It's all just ordinary information processing in a world where people can carry billions of bytes of memory and gigaflops of processing power in their briefcases. Soon the same computing power will be available with just a cell phone. And I assure you, this process is not going to stop. Do you think composition is beyond automation? I've seen the future in a website that automatically generates Dave Barry columns.2 You seed it with some humorous tidbits, and it does the rest. How long will it be before editorial style is automated? …

Contemporary Systems and Cybernetics

Professor Donald Michie, the Director of the Turing Institute, Glasgow, Scotland, UK, has discussed a breakthrough in artificial intelligence in an article in the UK Observer newspaper. Writing at the end of last year he said that: