Sparse coding with an overcomplete basis set: a strategy employed by V1? Vision Res.

Last year, Google's DeepMind A.I. development house released a "tool" called DeepDream that let neural networks loose on innocent imagery, with truly terrifying results. Though it was hailed as a window into the secret experiences of A.I, in reality, DeepDream was a demonstration of just how primitive the mind of modern a A.I. really is. But could even an artificial intelligence be alien enough to pervert the peaceful, soothing imagery of the great Bob Ross? A new video titled "Deeply Artificial Trees," which applies DeepDream to every frame of a Bob Ross video, proves that the answer to this question is, well, yes. Further, it even applies the technique to the audio -- making the experience truly nightmarish.

No two stem cells are identical, even if they are genetic clones. This stunning diversity is revealed today in an enormous publicly available online catalogue of 3D stem cell images. The visuals were produced using deep learning analyses and cell lines altered with the gene-editing tool CRISPR. And soon the portal will allow researchers to predict variations in cell layouts that may foreshadow cancer and other diseases. The Allen Cell Explorer, produced by the Allen Institute for Cell Science in Seattle, Washington, includes a growing library of more than 6,000 pictures of induced pluripotent stem cells (iPS) -- key components of which glow thanks to fluorescent markers that highlight specific genes.

From a layman's perspective, I approached Nvidia's Deep Learning Institute event with a profound sense of unease. Would I be swept along in a sea of jargon and acronyms, furtively and futilely attempting to navigate the esoteric world of data science? In a Royal Institution lecture theatre populated by Microsoft delegates and Credit Suisse's IT team, would I feel totally and completely alienated by the formidable – not to mention inaccessible – world of deep learning? No, is the short answer to that. The world of deep learning is a complex one, I'll admit, but it's an electrifying one too: a phenomenon that wields innumerable possibilities for the streamlining and advancement of everyday life at home and in business.

Artificial intelligence has, all of a sudden, become the next big thing. It is not so much sweeping across our world as seeping into it, with a combination of enormous computing power and the latest'deep learning' techniques promising to reshape our lives. But can we use this extensively to deliver something that people actually "do" want? Whether sophisticated AI turns out to be friend or foe, we must come to grips with the possibility that as we move further into 2017, the greatest intelligence on the planet may be silicon-based. This year will see people-machine relationships become more pronounced, nuanced, fluid and yet deeply personalized.

When Microsoft acquired deep learning startup Maluuba in January, Maluuba's highly respected advisor, the deep learning pioneer Yoshua Bengio, agreed to continue advising Microsoft on its artificial intelligence efforts. Bengio, head of the Montreal Institute for Learning Algorithms, recently visited Microsoft's Redmond, Washington, campus, and took some time for a chat. Let's start with the basics: What is deep learning? Yoshua Bengio: Deep learning is an approach to machine learning, and machine learning is a way to try to make machines intelligent by allowing computers to learn from examples about the world around us or about some specific aspect of it. Deep learning is particular among all the machine learning methods in that it is inspired by some of the things we know about the brain.

Machine learning, especially its subfield of Deep Learning, had many amazing advances in the recent years, and important research papers may lead to breakthroughs in technology that get used by billions of people. The research in this field is developing very quickly and to help our readers monitor the progress we present the list of most important recent scientific papers published since 2014. The criteria we used to select the 20 top papers are by using citation counts from three academic sources: scholar.google.com; Since the number of citations varied among sources and are estimated, we listed the results from academic.microsoft.com For each paper we also give the year it was published, a Highly Influential Citation count (HIC) and Citation Velocity (CV) measures provided by semanticscholar.org.

However, AI may have a part to play in identifying the subtle signs that betray the kinds of human behavior that can lead to these kinds of crashes. If we look at the crash of the late 2000's it is clear that it was caused, in part, by people historically mis-selling mortgages to people who couldn't afford them. AI and deep learning make it simple to avoid these kinds of mistakes, whether somebody knew they were making a mistake or not. Similarly, when looking at stock market crashes, AI is designed to incorporate data from a wide variety of areas and formats, which means that it should be able to determine whether a stock is over or undersold and the inherent risk of its purchase fairly easily. It has the potential to somewhat stabilize the often unpredictable and unrealistic rise and fall of the stock market, even if it's through prompting behavior rather than stopping it altogether.

We are excited to announce our new RL Tuner algorithm, a method for enchancing the performance of an LSTM trained on data using Reinforcement Learning (RL). We create an RL reward function that teaches the model to follow certain rules, while still allowing it to retain information learned from data. We use RL Tuner to teach concepts of music theory to an LSTM trained to generate melodies. When I joined Magenta as an intern this summer, the team was hard at work on developing better ways to train Recurrent Neural Networks (RNNs) to generate sequences of notes. As you may remember from previous posts, these models typically consist of a Long Short-Term Memory (LSTM) network trained on monophonic melodies. This means that melodies are fed into the network one note at a time, and it is trained to predict the next note in the sequence.

Unless you have been living under the rock, you must have heard of the revolution that deep learning and convolutional neural networks have brought in computer vision. Computers have achieved near-human level accuracy for most of the tasks. This problem gets worse for an application like object detection where multiple windows at different locations and scale need to be processed. Models that achieve state of the art accuracy are too large to fit into mobile devices or small devices like Raspberry Pi. Even if you somehow manage to live with the large size of models, the amount of run-time memory(RAM) required to run these models is way too high and limits their usage.