Artificial Intelligence (AI), machine learning, and deep learning are taking the healthcare industry by storm. They are not pie in the sky technologies any longer; they are practical tools that can help companies optimize their service provision, improve the standard of care, generate more revenue, and decrease risk. Nearly all major companies in the healthcare space have already begun to use the technology in practice; here I present some of the important highlights of the implementation, and what they mean for other companies in healthcare. AI, machine learning, and deep learning are already increasing profits in the healthcare industry. For example, according to research firm Frost & Sullivan by 2021, AI systems will generate $6.7 billion in global healthcare industry revenue.

Inspur has announced the open-source release of TF2, an FPGA-based efficient AI computing framework. The inference engine of this framework employs the world's first DNN shift computing technology, combined with a number of the latest optimization techniques, to achieve FPGA-based high-performance low-latency deployment of universal deep learning models. This is also the world's first open-sourced FPGA-based AI framework that contains comprehensive solutions ranging from model pruning, compression, quantization, and a general DNN inference computing architecture based on FPGA. The open source project can be found at https://github.com/TF2-Engine/TF2. Many companies and research institutions, such as Kuaishou, Shanghai University, and MGI, are said to have joined the TF2 open source community, which will jointly promote open-source cooperation and the development of AI technology based on customizable FPGAs, reducing the barriers to high-performance AI computing technology, and shortening development cycles for AI users and developers.

Asked what is the biggest misconception about AI, Yoshua Bengio answered without hesitation "AI is not magic." Winner of the 2018 Turing Award (with the other "fathers of the deep learning revolution," Geoffrey Hinton and Yann LeCun), Bengio spoke at the EmTech MIT event about the "amazing progress in AI" while stressing the importance of understanding its current limitations and recognizing that "we are still very far from human-level AI in many ways." Deep learning has moved us a step closer to human-level AI by allowing machines to acquire intuitive knowledge, according to Bengio. Classical AI was missing this "learning component," and deep learning develops intuitive knowledge "by acquiring that knowledge from data, from interacting with the environment, from learning. That's why current AI is working so much better than the old AI."

A new technology using artificial intelligence detects depressive language in social media posts more accurately than current systems and uses less data to do it. The technology, which was presented during the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, is the first of its kind to show that, to more accurately detect depressive language, small, high-quality data sets can be applied to deep learning, a commonly used AI approach that is typically data intensive. Previous psycholinguistic research has shown that the words we use in interaction with others on a daily basis are a good indicator of our mental and emotional state. Past attempts to apply deep learning techniques to detect and monitor depression in social media posts have been shown to be tedious and expensive, explained Nawshad Farruque, a University of Alberta Ph.D. student in computing science who is leading the new study. He explained that a Twitter post saying that somebody is depressed because Netflix is down isn't really expressing depression, so someone would need to "explain" this to the algorithm.

Three years ago, we had maybe six or less AI accelerators, today there's over two dozen, and more are coming. One of the first commercially available AI training accelerators was the GPU, and the undisputed leader of that segment was Nvidia. Nvidia was already preeminent in machine learning (ML) and deep-learning (DL) applications and adding neural net acceleration was a logical and rather straight-forward step for the company. Nvidia also brought a treasure-trove of applications with their GPUs based on the company's proprietary development language CUDA. The company developed CUDA in 2006 and empowered hundreds of Universities to give courses on it. As a result, the thousands of computer science graduates every year came out of school knowing how and wanting to use CUDA.

Three years ago, we had maybe six or less AI accelerators, today there's over two dozen, and more are coming. One of the first commercially available AI training accelerators was the GPU, and the undisputed leader of that segment was Nvidia. Nvidia was already preeminent in machine learning (ML) and deep-learning (DL) applications and adding neural net acceleration was a logical and rather straight-forward step for the company. Nvidia also brought a treasure-trove of applications with their GPUs based on the company's proprietary development language CUDA. The company developed CUDA in 2006 and empowered hundreds of Universities to give courses on it. As a result, the thousands of computer science graduates every year came out of school knowing how and wanting to use CUDA.

OpenAI recently posted on Twitter about teaching computer agents'hide and seek'. We've observed AIs discovering complex tool use while competing in a simple game of hide-and-seek. They develop a series of six distinct strategies and counter strategies, ultimately using tools in the environment to break our simulated physics. In the simulations, seekers are incentivized to maintain line of sight of hiders and hiders are incentivized to avoid line of sight from seekers. The agents environments contain various shelters including cubicles, movable partitions, blocks and ramps. That said, there is no built-in incentive for agents to interact with objects around them.

Optimal engine operation during a transient driving cycle is the key to achieving greater fuel economy, engine efficiency, and reduced emissions. In order to achieve continuously optimal engine operation, engine calibration methods use a combination of static correlations obtained from dynamometer tests for steady-state operating points and road and/or track performance data. As the parameter space of control variables, design variable constraints, and objective functions increases, the cost and duration for optimal calibration become prohibitively large. In order to reduce the number of dynamometer tests required for calibrating modern engines, a large-scale simulation-driven machine learning approach is presented in this work. A parallel, fast, robust, physics-based reduced-order engine simulator is used to obtain performance and emission characteristics of engines over a wide range of control parameters under various transient driving conditions (drive cycles). We scale the simulation up to 3,906 nodes of the Theta supercomputer at the Argonne Leadership Computing Facility to generate data required to train a machine learning model. The trained model is then used to predict various engine parameters of interest. Our results show that a deep-neural-network-based surrogate model achieves high accuracy for various engine parameters such as exhaust temperature, exhaust pressure, nitric oxide, and engine torque. Once trained, the deep-neural-network-based surrogate model is fast for inference: it requires about 16 micro sec for predicting the engine performance and emissions for a single design configuration compared with about 0.5 s per configuration with the engine simulator. Moreover, we demonstrate that transfer learning and retraining can be leveraged to incrementally retrain the surrogate model to cope with new configurations that fall outside the training data space.

Criminals use malware to disrupt cyber-systems. The number of these malware-vulnerable systems is increasing quickly as common systems, such as vehicles, routers, and lightbulbs, become increasingly interconnected cyber-systems. To address the scale of this problem, analysts divide malware into classes and develop, for each class, a specialized defense. In this project we classified malware with machine learning. In particular, we used a supervised multi-class long short term memory (LSTM) model. We trained the algorithm with thousands of malware files annotated with class labels (the training set), and the algorithm learned patterns indicative of each class. We used disassembled malware files (provided by Microsoft) and separated the constituent data into parsed instructions, which look like human-readable machine code text, and raw bytes, which are hexadecimal values. We are interested in which format, text or hex, is more valuable as input for classification. To solve this, we investigated four cases: a text-only model, a hexadecimal-only model, a multi-input model using both text and hexadecimal inputs, and a model based on combining the individual results. We performed this investigation using the machine learning Python package Keras, which allows easily configurable deep learning architectures and training. We hoped to understand the trade-offs between the different formats. Due to the class imbalance in the data, we used multiple methods to compare the formats, using test accuracies, balanced accuracies (taking into account weights of classes), and an accuracy derived from tables of confusion. We found that the multi-input model, which allows learning on both input types simultaneously, resulted in the best performance. Our finding expedites malware classification research by providing researchers a suitable deep learning architecture to train a tailored version to their malware.

Uncertainty Quantification with Statistical Guarantees in End-to-End Autonomous Driving Control Rhiannon Michelmore 1, Matthew Wicker 1, Luca Laurenti 1, Luca Cardelli 1, Y arin Gal 1, Marta Kwiatkowska 1 Abstract -- Deep neural network controllers for autonomous driving have recently benefited from significant performance improvements, and have begun deployment in the real world. Prior to their widespread adoption, safety guarantees are needed on the controller behaviour that properly take account of the uncertainty within the model as well as sensor noise. Bayesian neural networks, which assume a prior over the weights, have been shown capable of producing such uncertainty measures, but properties surrounding their safety have not yet been quantified for use in autonomous driving scenarios. In this paper, we develop a framework based on a state-of-the-art simulator for evaluating end-to-end Bayesian controllers. In addition to computing pointwise uncertainty measures that can be computed in real time and with statistical guarantees, we also provide a method for estimating the probability that, given a scenario, the controller keeps the car safe within a finite horizon. We experimentally evaluate the quality of uncertainty computation by three Bayesian inference methods in different scenarios and show how the uncertainty measures can be combined and calibrated for use in collision avoidance. Our results suggest that uncertainty estimates can greatly aid decision making in autonomous driving. I NTRODUCTION Deep Neural Networks (DNNs) have seen a surge in popularity over the past decade, and their use has become widespread in many fields including safety-critical systems such as medical diagnosis and, in particular, autonomous cars. The latter have driven millions of miles without human intervention [1], [2], but offer few safety guarantees.