HOW HAS ARTIFICIAL intelligence, associated with hubris and disappointment since its earliest days, suddenly become the hottest field in technology? The term was coined in a research proposal written in 1956 which suggested that significant progress could be made in getting machines to "solve the kinds of problems now reserved for humans…if a carefully selected group of scientists work on it together for a summer". That proved to be wildly overoptimistic, to say the least, and despite occasional bursts of progress, AI became known for promising much more than it could deliver. Researchers mostly ended up avoiding the term, preferring to talk instead about "expert systems" or "neural networks". The rehabilitation of "AI", and the current excitement about the field, can be traced back to 2012 and an online contest called the ImageNet Challenge.

The state-of-the-art in object recognition has undergone dramatic changes in the last 20 years. In this talk, I will review the progression of the field and discuss why various approaches both succeeded and failed. The talk will cover visual recognition from the early 90's, including handwritten digit and face detection, to the current state-of-the-art in deep learning applied to object categorization. Algorithms will be explained at an intuitive level. The talk is aimed at the non-expert in computer vision with some knowledge of machine learning.

Last month NVIDIA, our favorite GPU vendor, dived into the converged appliance space. In fact we might call their new NVIDIA DGX-1 a hyperconverged supercomputer in a 4U box. Designed to support the application of GPU's to Deep Learning (i.e. The price is surprisingly affordable, and can replace the 250 server cluster you might otherwise need for effective Deep Learning. Despite the obvious opportunities, enterprises face a lot of obstacles in putting machine learning (and esp.

Recently, neural networks have become the tool of choice for a variety of tough computer-science problems: Facebook uses them to identify faces in photos, Google uses them to identify everything in photos. Apple uses them to figure out what you're saying to Siri, and IBM uses them for operationalizing business unit synergies. Can neural networks help you find the emoji when you really need it? This post will outline some of the engineering behind Dango, allowing us to automatically learn from hundreds of millions of real-world uses of emoji, and distill this down to a tool small and fast enough to predict emoji for you in real time on your phone. Dango is a floating assistant that runs on your phone and predicts emoji, stickers and GIFs based on what you and your friends are writing in any app.

This year at Cannes Lions, judges are recognizing work at the intersection of AI and advertising. The Next Rembrandt, a campaign by JWT Amsterdam for the Dutch bank ING, used "deep learning" algorithms to analyze more than 168,000 painting fragments by the 17th century master to create a digital image based on his existing work. Then, a 3D printer fabricated a painting based on the image, replicating the depth and texture of original Rembrandt works to a surprising degree. The campaign took home a Grand Prix at Cannes in the Cyber Lions category, which recognizes "ideas indigenous to, or enhanced by, the digital environment," and an additional Grand Prix in the Creative Data Lions category. This year's entries were more sophisticated than those in the past, said Cyber Lions jury president Chloe Gottlieb, SVP and executive creative director at R/GA. "The data is not an output from the creativity," said Gottlieb.

A group of researchers from Beth Israel Deaconess Medical Center (BIDMC) and Harvard Medical School (HMS) have developed a way to train artificial intelligence to read and interpret pathology images. Scientists tested the artificial intelligence (AI) during a competition at the annual International Symposium of Biomedical Imaging, where it was tasked to look for breast cancer in images of lymph nodes. It turns out it can detect breast cancer accurately 92 percent of the time and won in two separate categories during the contest. Andrew Beck from BIDMC says they used the deep learning method, which is commonly used to train AI to recognize speech, images and objects. They fed the machine with hundreds of slides marked to indicate which parts have cancerous cells and which have normal ones.

Just a few weeks back, scientists at Google's artificial intelligence division DeepMind announced they were developing a "kill switch" to ensure that intelligent machines do not go all Terminator on us. Now, it seems, Google's AI-related concerns …… Read More Google released a new paper on a highly controversial topic: safety rules for Artificial Intelligence. Artificial intelligence is either the bright shining future of technology or an insidious threat that could endanger all of mankind, depending on your point of view. The words Artificial Intelligence can bring to mind far-fetched, sci-fi ideas and a society where robots have replaced humans. Well, this idea may not be too far off given Google's recent innovations.

Since 2009, Microsoft has engaged with academic pioneers of deep learning and has created industry-scale successes in speech recognition as well as in speech translation, object recognition, automatic image captioning, natural language, multimodal processing, semantic modeling, web search, contextual entity search, ad selection, and big data analytics. Much of these successes are attributed to the availability of big datasets for training deep models, the powerful general-purpose GPU computing, and the innovations in deep learning architectures and algorithms. In this talk, a selected overview will be given to highlight our center's work in some of these exciting applications, as well as the lessons we have learned along the way as to what tasks are best solved by deep learning methods.

Google has released a paper on AI safety in collaboration with researchers from Stanford University, University of California, Berkeley and OpenAI.

This paper focuses on the problem of explaining predictions of psychological attributes such as attractiveness, happiness, confidence and intelligence from face photographs using deep neural networks. Since psychological attribute datasets typically suffer from small sample sizes, we apply transfer learning with two base models to avoid overfitting. These models were trained on an age and gender prediction task, respectively. Using a novel explanation method we extract heatmaps that highlight the parts of the image most responsible for the prediction. We further observe that the explanation method provides important insights into the nature of features of the base model, which allow one to assess the aptitude of the base model for a given transfer learning task. Finally, we observe that the multiclass model is more feature rich than its binary counterpart. The experimental evaluation is performed on the 2222 images from the 10k US faces dataset containing psychological attribute labels as well as on a subset of KDEF images.