Existing attention mechanisms, are mostly item-based in that a model is designed to attend to a single item in a collection of items (the memory). Intuitively, an area in the memory that may contain multiple items can be worth attending to as a whole. We propose area attention: a way to attend to an area of the memory, where each area contains a group of items that are either spatially adjacent when the memory has a 2-dimensional structure, such as images, or temporally adjacent for 1-dimensional memory, such as natural language sentences. Importantly, the size of an area, i.e., the number of items in an area, can vary depending on the learned coherence of the adjacent items. By giving the model the option to attend to an area of items, instead of only a single item, we hope attention mechanisms can better capture the nature of the task. Area attention can work along multi-head attention for attending to multiple areas in the memory. We evaluate area attention on two tasks: neural machine translation and image captioning, and improve upon strong (state-of-the-art) baselines in both cases. These improvements are obtainable with a basic form of area attention that is parameter free. In addition to proposing the novel concept of area attention, we contribute an efficient way for computing it by leveraging the technique of summed area tables.

The symbiosis of the two fields, how they overlap, how they inform each other, where they differ, what the past brought us, and what the future brings. Topics include, supervised machine learning, unsupervised learning, reinforcement learning, deep learning, convolutional and recurrent neural networks, decision-making science, AI agents, backpropagation, credit assignment, neuroengineering, neuromorphics, consciousness, general AI, spiking neural networks, data science, and a lot more.

In this work we describe how threat actors may use AI algorithms to bypass AI phishing detection systems. We analyzed more than a million phishing URLs to understand the different strategies that threat actors use to create phishing URLs. Assuming the role of an attacker, we simulate how different threat actors may leverage Deep Neural Networks to enhance their effectiveness rate.

OVER THE past five years researchers in artificial intelligence have become the rock stars of the technology world. A branch of AI known as deep learning, which uses neural networks to churn through large volumes of data looking for patterns, has proven so useful that skilled practitioners can command high six-figure salaries to build software for Amazon, Apple, Facebook and Google. The top names can earn over $1m a year. The standard route into these jobs has been a PhD in computer science from one of America's elite universities. Earning one takes years and requires a disposition suited to academia, which is rare among more normal folk.

In the fall of 2016, I was a Teaching Fellow (Harvard's version of TA) for the graduate class on "Advanced Topics in Data Science (CS209/109)" at Harvard University. I was in-charge of designing the class project given to the students, and this tutorial has been built on top of the project I designed for the class. As a researcher on Computer Vision, I come across new blogs and tutorials on ML (Machine Learning) every day. However, most of them are just focussing on introducing the syntax and the terminology relevant to the field. While people are able to copy paste and run the code in these tutorials and feel that working in ML is really not that hard, it doesn't help them at all in using ML for their own purposes.

Mastering the game of Go with deep neural networks and tree search.

Applying artificial intelligence to self-driving cars to smooth traffic, reduce fuel consumption, and improve air quality predictions may sound like the stuff of science fiction, but researchers at the Department of Energy's Lawrence Berkeley National Laboratory (Berkeley Lab) have launched two research projects to do just that. In collaboration with UC Berkeley, Berkeley Lab scientists are using deep reinforcement learning, a computational tool for training controllers, to make transportation more sustainable. One project uses deep reinforcement learning to train autonomous vehicles to drive in ways to simultaneously improve traffic flow and reduce energy consumption. A second uses deep learning algorithms to analyze satellite images combined with traffic information from cell phones and data already being collected by environmental sensors to improve air quality predictions. "Thirty percent of energy use in the U.S. is to transport people and goods, and this energy consumption contributes to air pollution, including approximately half of all nitrogen oxide emissions, a precursor to particular matter and ozone – and black carbon (soot) emissions," said Tom Kirchstetter, director of Berkeley Lab's Energy Analysis and Environmental Impacts Division, an adjunct professor at UC Berkeley, and a member of the research team.

Recent advancement in artificial intelligence, namely in deep learning, has borrowed concepts from the human brain. The architecture of most deep learning models is based on layers of processing– an artificial neural network that is inspired by the neurons of the biological brain. Yet neuroscientists do not agree on exactly what intelligence is, and how it is formed in the human brain -- it's a phenomena that remains unexplained. Technologist, scientist, and co-founder of Numenta, Jeff Hawkins, presented an innovative framework for understanding how the human neocortex operates, called "The Thousand Brains Theory of Intelligence," at the Human Brain Project Summit in Maaastricht, the Netherlands, in October 2018. The neocortex is the part of the human brain that is involved in higher-order functions such as conscious thought, spatial reasoning, language, generation of motor commands, and sensory perception.

A few years ago I saw this headline news flashing all over the internet. Our dealers are missing up to $18 billion in easy sales. The Chairman and CEO of Caterpillar suggested that the company and its dealers were losing $9 - 18 billion in easy sales revenue as their sales, both internal and dealer networks, weren't monetising the real value of data. They are not tapping into the wealth of real-time customer data now at their fingertips; they are not communicating with each other; and they are not providing customers across the globe with a consistent experience when it comes to everything from e-commerce to parts and services pricing. Long story short, the whole idea was to convert the company's mentality from dumb iron sales to data-driven, machine learning-driven sales.

AI researchers create algorithms and ask questions to be answered by AI..Depositphotos enhanced by CogWorld Is data the new oil? Considering oil's sky-high demand by both industry and government organizations, and the fact that oil only becomes more valuable post-refining and offers a competitive edge to those owning this commodity in surplus, I feel the analogy between data and oil is quite justified. Just as oil requires processing to become usable and is fed to engines to work and create value, data requires cleaning and is fed to algorithms to derive valuable insights and make decisions. While there's only one way to use oil - burning it in an engine to generate energy - data on the other hand can be used in myriad ways for a range of different purposes, from identifying simple statistical trends to training complex robotic AI for independent interaction with humans. The effectiveness and innovativeness of data-driven AI applications depends chiefly on the people working on those applications, such as AI researchers and data scientists, who are among the most sought-after people in the tech sector today.