A fatal crash that occurred when an autonomous SUV operated by Uber struck and killed a pedestrian could have better been avoided if a human was in control of the vehicle, some experts believe. Footage of the incident, which occurred on Sunday in Tempe, Arizona, and resulted in the death of 49-year-old Elaine Herzberg, was released by the local police Wednesday. Experts have suggested that Uber's self-driving technology should have been able to avoid the crash and failed to do so. Experts believe a human driver could have avoided a fatal accident involving Uber's self-driving SUV. The video includes footage from a dashboard camera showing a view outside the car, as well as a view of the operator employed by Uber sitting behind the wheel of the vehicle and take over if the autonomous system does not work as intended.

Machine Learning (ML) is one of the most exciting and dynamic areas of modern research and application. The purpose of this review is to provide an introduction to the core concepts and tools of machine learning in a manner easily understood and intuitive to physicists. The review begins by covering fundamental concepts in ML and modern statistics such as the bias-variance tradeoff, overfitting, regularization, and generalization before moving on to more advanced topics in both supervised and unsupervised learning. Topics covered in the review include ensemble models, deep learning and neural networks, clustering and data visualization, energy-based models (including MaxEnt models and Restricted Boltzmann Machines), and variational methods. Throughout, we emphasize the many natural connections between ML and statistical physics. A notable aspect of the review is the use of Python notebooks to introduce modern ML/statistical packages to readers using physics-inspired datasets (the Ising Model and Monte-Carlo simulations of supersymmetric decays of proton-proton collisions). We conclude with an extended outlook discussing possible uses of machine learning for furthering our understanding of the physical world as well as open problems in ML where physicists maybe able to contribute. (Notebooks are available at https://physics.bu.edu/~pankajm/MLnotebooks.html )

In domains with high knowledge distribution a natural objective is to create principle foundations for collaborative interactive learning environments. We present a first mathematical characterization of a collaborative learning group, a consortium, based on closure systems of attribute sets and the well-known attribute exploration algorithm from formal concept analysis. To this end, we introduce (weak) local experts for subdomains of a given knowledge domain. These entities are able to refute and potentially accept a given (implicational) query for some closure system that is a restriction of the whole domain. On this we build up a consortial expert and show first insights about the ability of such an expert to answer queries. Furthermore, we depict techniques on how to cope with falsely accepted implications and on combining counterexamples. Using notions from combinatorial design theory we further expand those insights as far as providing first results on the decidability problem if a given consortium is able to explore some target domain. Applications in conceptual knowledge acquisition as well as in collaborative interactive ontology learning are at hand.

Humans can learn in a continuous manner. Old rarely utilized knowledge can be overwritten by new incoming information while important, frequently used knowledge is prevented from being erased. In artificial learning systems, lifelong learning so far has focused mainly on accumulating knowledge over tasks and overcoming catastrophic forgetting. In this paper, we argue that, given the limited model capacity and the unlimited new information to be learned, knowledge has to be preserved or erased selectively. Inspired by neuroplasticity, we propose a novel approach for lifelong learning, coined Memory Aware Synapses (MAS). It computes the importance of the parameters of a neural network in an unsupervised and online manner. Given a new sample which is fed to the network, MAS accumulates an importance measure for each parameter of the network, based on how sensitive the predicted output function is to a change in this parameter. When learning a new task, changes to important parameters can then be penalized, effectively preventing important knowledge related to previous tasks from being overwritten. Further, we show an interesting connection between a local version of our method and Hebb's rule,which is a model for the learning process in the brain. We test our method on a sequence of object recognition tasks and on the challenging problem of learning an embedding for predicting $<$subject, predicate, object$>$ triplets. We show state-of-the-art performance and, for the first time, the ability to adapt the importance of the parameters based on unlabeled data towards what the network needs (not) to forget, which may vary depending on test conditions.

Nando de Freitas is a "machine learning professor at Oxford University, a lead research scientist at Google DeepMind, and a Fellow of the Canadian Institute For Advanced Research (CIFAR) in the Neural Computation and Adaptive Perception program." Above, you can watch him teach an Oxford course on Deep Learning, a hot subfield of machine learning and artificial intelligence which creates neural networks--essentially complex algorithms modeled loosely after the human brain--that can recognize patterns and learn to perform tasks. To complement the 16 lectures you can also find lecture slides, practicals, and problems sets on this Oxford web site. If you'd like to learn about Deep Learning in a MOOC format, be sure to check out the new series of courses created by Andrew Ng on Coursera. Oxford's Deep Learning course will be added to our list of Free Online Computer Science Courses, part of our meta collection, 1,300 Free Online Courses from Top Universities.

We're taught from an early age to take notes. I believe the first time I recall starting to take notes was when I entered junior high. In fact, I can even recall being required to take a class on how to take notes – Roman numbers, indentation, etc. The act of taking notes is one of the few activities that has stood the test of time as I continued to do this not only in junior high, but throughout my education and it is something that I continue to do regularly today in my professional career. The reason why notes continue to be so prevalent in all of our lives is not because our teachers were so effective at influencing our young impressionable selves, but it's because taking notes is an extremely useful practice. It allows us to capture the highlights of a class, meeting, or event and serves as a tool to help us recall information or re-enforce learning.

The invention of artificial intelligence has been hotly debated over the years. Some view this tool as the first step toward a world where human professions are no longer necessary. Others see artificial intelligence as a cost-effective means of being more productive during the day. The truth may fall somewhere in between these extremes, particularly when it comes to education. Artificial intelligence has already transformed the face of learning in a major way.

In 2016, DeepMind's AlphaGo famously defeated Lee Sedol, an international Go champion, becoming the first computer program to beat a human world champion. In 2018, LawGeex, an AI contract review platform, pulled the same stunt on human lawyers. The AI system achieved a 94 percent accuracy rate at surfacing risks in non-disclosure agreements (NDAs). Experienced human lawyers average out at 85 percent accuracy for the same task. The study, conducted in collaboration with Duke and Stanford Law Schools, pitted AI against 20 top U.S.-trained lawyers with decades of experience specifically in reviewing NDAs, one of the most common agreements in law.

Interest in machine learning has exploded over the past decade. You see machine learning in computer science programs, industry conferences, and the Wall Street Journal almost daily. For all the talk about machine learning, many conflate what it can do with what they wish it could do. Fundamentally, machine learning is using algorithms to extract information from raw data and represent it in some type of model. We use this model to infer things about other data we have not yet modeled.

Google has started the Intelligent Taiwan initiative to help promote digital transformation of the country's economy and strengthen AI (artificial intelligence) capability of its companies, according to Google's Taiwan managing director Chien Lee-feng. With Intelligent Taiwan, Google hopes to realize its "AI first" concept in Taiwan, Chien indicated, adding Taiwan meets all Google's requirements for developing AI, including those for software, hardware and cloud computing infrastructure. In the initiative, Google will expand its manpower in Taiwan by recruiting over 300 software/hardware engineers and R&D staff, provide digital marketing training for over 50,000 local small- to medium-size businesses and students, and train over 5,000 local students in AI programming in 2018, Chien said. According to Chien, for digital marketing training, Google will provide a free and convenient online platform and facilities in Taichung and Tainan to cater to those who prefer face-to-face training. For AI training for students, Google will first give in-depth instruction to university and senior high school teachers, and seed instructors with government-sponsored Institute for Information Industry (III) using Google-developed Machine Learning Crash Course.