For the first three years of OpenAI, I dreamed of becoming a machine learning expert but made little progress towards that goal. Over the past nine months, I've finally made the transition to being a machine learning practitioner. It was hard but not impossible, and I think most people who are good programmers and know (or are willing to learn) the math can do it too. There are many online courses to self-study the technical side, and what turned out to be my biggest blocker was a mental barrier -- getting ok with being a beginner again. A founding principle of OpenAI is that we value research and engineering equally -- our goal is to build working systems that solve previously impossible tasks, so we need both.

UNIST scientists have recently developed an interactive visual analytics system that enables traffic congestion exploration, surveillance, and forecasting based on vehicle detector data. Their system can predict traffic conditions for the next 5 to 15 minutes at an error rate of fewer than four kilometers an hour. This intelligent visual analytics system empowers traffic congestion exploration, observation, and determining dependent on vehicle detector information. Through domain expert collaboration, we have extricated task requirements, consolidated the Long-Short Term Memory (LSTM) model for congestion forecasting, and designed a weighting technique for distinguishing the reasons for congestion and congestion propagation directions. The system then visualized the traffic situation for easier comprehension: Congestion levels and average driving speed, for instance, are described using colors and shapes.

In acquiring Vertex.AI for its Movidius unit, Intel is envisioning a tomorrow where deep learning will feature in many aspects of business. Chip giant Intel has acquired Vertex.AI, a Seattle start-up that is developing deep learning for every platform. The start-up will join the Movidius group, which is focused on self-learning and artificial intelligence (AI) technology on a myriad of devices. Intel acquired Movidius in 2016 for an undisclosed sum, rumoured to be in the region of $300m. This was part of a $1bn spending spree on AI tech companies, including Mighty AI, DataRobot, Lumiata, AEye and others.

Summary: Combining multiple artificial intelligence agents sheds light on the aging process and can help further understanding of what contributes to healthy aging. There are two kinds of age: chronological age, which is the number of years one has lived, and biological age, which is influenced by our genes, lifestyle, behavior, the environment, and other factors. Biological age is the superior measure of true age and is the most biologically relevant feature, as it closely correlates with mortality and health status. The search for reliable predictors of biological age has been ongoing for several decades, and until recently, largely without success. Since 2016 the use of deep learning techniques to find predictors of chronological and biological age has been gaining popularity in the aging research community.

Computer vision is still an emerging technology for businesses. But in September 2018, Gartner analyst Bern Elliot included computer vision in a list of enterprise AI technologies that organizations should at least be experimenting with, particularly in combination with machine learning and deep learning algorithms. Amsterdam UMC, which operates two teaching hospitals in the Netherlands, is doing just that. It has started using a computer vision AI application running in a SAS Institute Inc. analytics system to measure the size of liver tumors in colorectal cancer patients. The automated medical-image analysis is faster and more accurate than manual measurements by radiologists, according to Geert Kazemier, a surgery professor and director of surgical oncology at Amsterdam UMC.

Frontier Technology Conference (FTC) is San Francisco's first and seminal frontier technology conference. It is a two-day conference of learning, inspiration, and networking. The conference will feature Augmented Reality, Artificial Intelligence, and Blockchain (DLT's). Specific included technologies are XR (Augmented Reality, Virtual Reality, and Mixed Reality), AI (Computer Vision, Machine Learning, and Deep Learning), and Blockchain and other Distributed Ledger Technologies. It caters to top executives, technologists, developers, entrepreneurs, and other frontier technology professionals, as well as enthusiasts and the curious-minded.

The ability to accurately predict the surrounding environment is a foundational principle of intelligence in biological and artificial agents. In recent years, a variety of approaches have been proposed for learning to predict the physical dynamics of objects interacting in a visual scene. Here we conduct a systematic empirical evaluation of several state-of-the-art unsupervised deep learning models that are considered capable of learning the spatio-temporal structure of a popular dataset composed by synthetic videos of bouncing objects. We show that most of the models indeed obtain high accuracy on the standard benchmark of predicting the next frame of a sequence, and one of them even achieves state-of-the-art performance. However, all models fall short when probed with the more challenging task of generating multiple successive frames. Our results show that the ability to perform short-term predictions does not imply that the model has captured the underlying structure and dynamics of the visual environment, thereby calling for a careful rethinking of the metrics commonly adopted for evaluating temporal models. We also investigate whether the learning outcome could be affected by the use of curriculum-based teaching.

Algorithms are powerful and necessary tools behind a large part of the information we use every day. However, they may introduce new sources of bias, discrimination and other unfair practices that affect people who are unaware of it. Greater algorithm transparency is indispensable to provide more credible and reliable services. Moreover, requiring developers to design transparent algorithm-driven applications allows them to keep the model accessible and human understandable, increasing the trust of end users. In this paper we present EBAnO, a new engine able to produce prediction-local explanations for a black-box model exploiting interpretable feature perturbations. EBAnO exploits the hypercolumns representation together with the cluster analysis to identify a set of interpretable features of images. Furthermore two indices have been proposed to measure the influence of input features on the final prediction made by a CNN model. EBAnO has been preliminarily tested on a set of heterogeneous images. The results highlight the effectiveness of EBAnO in explaining the CNN classification through the evaluation of interpretable features influence.

Deep learning (DL) has shown great potential in medical image enhancement problems, such as super-resolution or image synthesis. However, to date, little consideration has been given to uncertainty quantification over the output image. Here we introduce methods to characterise different components of uncertainty in such problems and demonstrate the ideas using diffusion MRI super-resolution. Specifically, we propose to account for $intrinsic$ uncertainty through a heteroscedastic noise model and for $parameter$ uncertainty through approximate Bayesian inference, and integrate the two to quantify $predictive$ uncertainty over the output image. Moreover, we introduce a method to propagate the predictive uncertainty on a multi-channelled image to derived scalar parameters, and separately quantify the effects of intrinsic and parameter uncertainty therein. The methods are evaluated for super-resolution of two different signal representations of diffusion MR images---DTIs and Mean Apparent Propagator MRI---and their derived quantities such as MD and FA, on multiple datasets of both healthy and pathological human brains. Results highlight three key benefits of uncertainty modelling for improving the safety of DL-based image enhancement systems. Firstly, incorporating uncertainty improves the predictive performance even when test data departs from training data. Secondly, the predictive uncertainty highly correlates with errors, and is therefore capable of detecting predictive "failures". Results demonstrate that such an uncertainty measure enables subject-specific and voxel-wise risk assessment of the output images. Thirdly, we show that the method for decomposing predictive uncertainty into its independent sources provides high-level "explanations" for the performance by quantifying how much uncertainty arises from the inherent difficulty of the task or the limited training examples.

Internet-of-Things (IoT) devices that are limited in power and processing capabilities are susceptible to physical layer (PHY) spoofing attacks owing to their inability to implement a full-blown protocol stack for security. The overwhelming adoption of multicarrier communications for the PHY layer makes IoT devices further vulnerable to PHY spoofing attacks. These attacks which aim at injecting bogus data into the receiver, involve inferring transmission parameters and finding PHY characteristics of the transmitted signals so as to spoof the received signal. Non-contiguous orthogonal frequency division multiplexing (NC-OFDM) systems have been argued to have low probability of exploitation (LPE) characteristics against classic attacks based on cyclostationary analysis. However, with the advent of machine learning (ML) algorithms, adversaries can devise data-driven attacks to compromise such systems. It is in this vein that PHY spoofing performance of adversaries equipped with supervised and unsupervised ML tools are investigated in this paper. The supervised ML approach is based on estimation/classification utilizing deep neural networks (DNN) while the unsupervised one employs variational autoencoders (VAEs). In particular, VAEs are shown to be capable of learning representations from NC-OFDM signals related to their PHY characteristics such as frequency pattern and modulation scheme, which are useful for PHY spoofing. In addition, a new metric based on the disentanglement principle is proposed to measure the quality of such learned representations. Simulation results demonstrate that the performance of the spoofing adversaries highly depends on the subcarriers' allocation patterns used at the transmitter. Particularly, it is shown that utilizing a random subcarrier occupancy pattern precludes the adversary from spoofing and secures NC-OFDM systems against ML-based attacks.