Researchers from Karlsruhe Institute of Tech, MIT and University of Toronto published MovieQA, a dataset that contains 7702 reasoning questions and answers from 294 movies. Their innovative dataset and accuracy metrics provide a well-defined challenge for question/answer machine learning algorithms. The questions range from simpler'Who' did'What' to'Whom' that can be solved by computer vision alone, to'Why' and'How' something happened in the movie, questions that can only be solved by exploiting both the visual information and dialogs. MovieQA is unique in that it contains multiple sources of information – full-length movies, plot synopses, subtitles, scripts and DVS (a service that narrates moves scenes to the visually impaired). With the need to scale to large vocabulary data sets, they relied on a TITAN Black GPU for their overwhelming amount of training data.

Continuing our series of deep learning updates, we pulled together some of the awesome resources that have emerged since our last post on June 20th. Google's DeepMind partners with the National Health Service's Moorfields Eye Hopsital to apply machine learning to spot common eye diseases earlier. The goal is that this leads to a better understanding of eye disease, earlier detection, and treatment. The Harvard NLP and Visual Computing groups announce LSTMVis, a visual analysis tool for recurrent neural networks (RNNs).

Continuing our series of deep learning updates, we pulled together some of the awesome resources that have emerged since our last post on June 20th. In case you missed it, here's the June update, and here's the original set of 20 resources we outlined in April. As always, this list is not comprehensive, so let us know if there's something we should add, or if you're interested in discussing this area further. Google's DeepMind partners with the National Health Service's Moorfields Eye Hopsital to apply machine learning to spot common eye diseases earlier. The five-year research project will draw on one million anonymous eye scans which are held on Moorfields' patient database, with the aim to speed up the complex and time-consuming process of analyzing eye scans (news article).

We are assisting a top international company currently building a Machine Learning and Data Analytics team in Dublin source a number of Software Engineers with proven experience implementing and applying Machine Learning techniques and methodologies in a commercial environment. This is a fantastic opportunity for a Senior Software Engineers with expertise in Machine Learning and Cognitive Computing technologies join a new operation with huge expansion plans for 2016/17 and beyond. This is a fantastic opportunity to work inside a top international company utilising cutting edge tools and techniques.

A bunch of startups is using tech to make everyday tasks much easier. One can book flights and buy mobile phone or insurance plans, simply by using a messaging app and chatting with a robot. Remember JARVIS, 'Iron Man' Tony Stark's intelligent assistant who not only controls everything in Stark's house but was also his closest friend. Some thing similar, far from that powerful yet, is available now to more ordinary folks. They are called chatbots, computer programs that can have automated text conversations with users using artificial intelligence (AI) and natural language processing.

Where did the "cyber" in "cyberspace" come from? Most people, when asked, will probably credit William Gibson, who famously introduced the term in his celebrated 1984 novel, Neuromancer. It came to him while watching some kids play early video games. Searching for a name for the virtual space in which they seemed immersed, he wrote "cyberspace" in his notepad. "As I stared at it in red Sharpie on a yellow legal pad," he later recalled, "my whole delight was that it meant absolutely nothing."

In a recent announcement, researchers at Google's Britain-based DeepMind unit have revealed that they have developed a new artificial intelligence (AI) technology which has the capability to mimic the sound of human voice. The new AI technology developed by DeepMind researchers is called WaveNet. It marks the achievement of a new milestone in the company's AI project. According to DeepMind researchers, the breakthrough WaveNet technology can be described as a deep neural network which is capable of generating raw audio wave forms to produce speech. The technology can apparently outperform the currently-in-use Text-to-Speech systems by reducing the gap in human performance -- which could be demonstrated in an actual AI, that is, human conversation -- by up to 50 percent. Moreover, WaveNet - which can presently use only the English and Chinese languages - also boasts the ability to learn different voices and speech patterns to such an extent that it can not only simulate mouth movements and artificial breaths, but also accents, emotions, and language inflections.

A group of researchers at Google appear to have taken a leaf out of HBO comedy Silicon Valley's book for its latest project. The team has developed a way to use neural networks that mimic the workings of the human brain to compress images more efficiently than traditional methods. The researchers trained an AI system (built using Google's TensorFlow, which the company open sourced last year) to learn how compression works using 6 million photos for reference. It broke these images into 32 x 32 pixel pieces and selected 100 pieces with the least effective compression to learn from; the idea is that training with these difficult bits would make it a cakewalk to handle the rest of the image. The AI then predicts how a image would look after it's compressed and generates that result.

Two of today's major business and intellectual trends offer complementary insights about the challenge of making forecasts in a complex and rapidly changing world. Forty years of behavioral science research into the psychology of probabilistic reasoning have revealed the surprising extent to which people routinely base judgments and forecasts on systematically biased mental heuristics rather than careful assessments of evidence. These findings have fundamental implications for decision making, ranging from the quotidian (scouting baseball players and underwriting insurance contracts) to the strategic (estimating the time, expense, and likely success of a project or business initiative) to the existential (estimating security and terrorism risks). The bottom line: Unaided judgment is an unreliable guide to action. Consider psychologist Philip Tetlock's celebrated multiyear study concluding that even top journalists, historians, and political experts do little better than random chance at forecasting such political events as revolutions and regime changes.1 The second trend is the increasing ubiquity of data-driven decision making and artificial intelligence applications. Once again, an important lesson comes from behavioral science: A body of research dating back to the 1950s has established that even simple predictive models outperform human experts' ability to make predictions and forecasts. This implies that judiciously constructed predictive models can augment human intelligence by helping humans avoid common cognitive traps.

The video analytics industry is typically split into two distinct camps: (1) systems designed around rules and user-specified rules or models and (2) autonomous systems designed around machine learning. Supervised learning systems require heavy training and feedback to achieve the desired output, where unsupervised learning systems train themselves from the input data and require minimal human input. The video analytic solutions we saw in the market a decade ago seem rudimentary compared to today's offerings; partly due to the technology catching up with early promises and partly due to the industry's understanding and level-setting of expectations from the initial splash of analytics hyped as a panacea and the future of security. However, some of the extreme claims such as its ability to replace trained human operators, eliminate the need for well-designed camera placement, completely eliminate false positives, and determine a person's intent ahead of an action have proven to be more hype than reality for many end users.