Today, we're very excited to announce that we're expanding our capabilities in machine learning by acquiring Magic Pony Technology, a London-based technology company that has developed novel machine learning techniques for visual processing. Magic Pony's technology – based on research by the team to create algorithms that can understand the features of imagery – will be used to enhance our strength in live and video and opens up a whole lot of exciting creative possibilities for Twitter. The team includes 11 PhDs with expertise across computer vision, machine learning, high-performance computing, and computational neuroscience, who are alumni of some of the top labs in the world. We are continuing to build strength into our deep learning teams with world-class talent to help Twitter be the best place to see what's happening and why it matters, first.

Twitter today is taking another step to build up its machine learning muscle, and also potentially to improve how it delivers photos and videos across its apps: the company is acquiring Magic Pony Technology (that is really the name), a company based out of London that has developed techniques of using neural networks (systems that essentially are designed to think like human brains) and machine learning to provide expanded data for images -- used, for example, to enhance a picture or video taken on a mobile phone; or to help develop graphics for virtual reality or augmented reality applications. Terms of the deal are not being disclosed but we have two separate sources who tell us that Twitter is paying 150 million in all for the deal. This takes into account retention bonuses for the staff, which numbers about 11, including co-founders Zehan Wang and CEO Rob Bishop. "Machine learning is increasingly at the core of everything we build at Twitter," said Jack Dorsey, Twitter CEO and co-founder, in a statement. "Magic Pony's machine learning technology will help us build strength into our deep learning teams with world-class talent, so Twitter can continue to be the best place to see what's happening and why it matters, first. We value deep learning research to help make our world better, and we will keep doing our part to share our work and learnings with the community."

The artificial intelligence from Google's DeepMind Technologies division is impressively versatile, there's no doubt. Late last year, it became the first neural network in history to defeat a professional player at Go, the Chinese board game whose human players had stumped computers for years, by besting world-ranked player Lee Sedol. It has demonstrated a prowess for video games, too -- it taught itself to emerge victorious in 49 different games for the Atari 2600 console and navigate digital 3D-maze called Labyrinth. And now, Google's human-like AI has learned how to play a sport of a different nature: soccer. DeepMind's latest experiment involves teaching an ant-like digital bug to maneuver a soccer ball into a goal.

We investigate the use of Deep Neural Networks for the classification of image datasets where texture features are important for generating class-conditional discriminative representations. To this end, we first derive the size of the feature space for some standard textural features extracted from the input dataset and then use the theory of Vapnik-Chervonenkis dimension to show that hand-crafted feature extraction creates low-dimensional representations which help in reducing the overall excess error rate. As a corollary to this analysis, we derive for the first time upper bounds on the VC dimension of Convolutional Neural Network as well as Dropout and Dropconnect networks and the relation between excess error rate of Dropout and Dropconnect networks. The concept of intrinsic dimension is used to validate the intuition that texture-based datasets are inherently higher dimensional as compared to handwritten digits or other object recognition datasets and hence more difficult to be shattered by neural networks. We then derive the mean distance from the centroid to the nearest and farthest sampling points in an n-dimensional manifold and show that the Relative Contrast of the sample data vanishes as dimensionality of the underlying vector space tends to infinity.

Great hardware needs great software. To help data scientists and developers make the most of the vast opportunities in deep learning, we're announcing today at the International Supercomputing show, ISC16, a trio of new capabilities for our deep learning software platform. The three -- NVIDIA DIGITS 4, CUDA Deep Neural Network Library (cuDNN) 5.1 and the new GPU Inference Engine (GIE) -- are powerful tools that make it even easier to create solutions on our platform. NVIDIA DIGITS 4 introduces a new object detection workflow, enabling data scientists to train deep neural networks to find faces, pedestrians, traffic signs, vehicles and other objects in a sea of images. This workflow enables advanced deep learning solutions -- such as tracking objects from satellite imagery, security and surveillance, advanced driver assistance systems and medical diagnostic screening.

Last month, Google (NASDAQ:GOOG) (NASDAQ:GOOGL) announced, among other things, the Tensor Processing Unit (TPU). While a number of Seeking Alpha commenters have hailed this innovation (here and here), none of them have really touched on the technicals reasons behind this, other than general comments on'better machine learning'. This article is a brief summary of my thoughts on Google's move into custom machine learning hardware. Let's begin by clarifying what exactly the TPU is. It's an application specific integrated circuit (ASIC).

We've long known that there are many tasks that computers can perform faster – and better – than humans. Of course, we still have to teach computers HOW to do these tasks, and when using conventional programming techniques we have to be very specific about what computers should do and when. With machine learning, we're essentially teaching computers how to learn what to do, and some of them are becoming better than we are at complex tasks. For example, machine learning is a key enabler of self-driving cars and experts predict that they will eventually be safer than human-driven vehicles. That's just one example of how machine learning is letting us use computers in new ways to do new things.

The TensorFlow project is bigger than you might realize. The fact that it's a library for deep learning, and its connection to Google, has helped TensorFlow attract a lot of attention. Cool stuff, but--especially for someone hoping to explore machine learning for the first time--TensorFlow can be a lot to take in. Let's break it down so we can see and understand every moving part. We'll explore the data flow graph that defines the computations your data will undergo, how to train models with gradient descent using TensorFlow, and how TensorBoard can visualize your TensorFlow work. The examples here won't solve industrial machine learning problems, but they'll help you understand the components underlying everything built with TensorFlow, including whatever you build next!

Twitter has bought a machine learning startup that can automatically sharpen low-resolution and blurred video in real time. The social network announced this morning that it had acquired London-based Magic Pony Technology for an undisclosed sum, with Twitter CEO Jack Dorsey tweeting that the move will help the the company reach its goal of "making Twitter the first and best place to see what's happening in the world." The benefits of Magic Pony's tech are clear for Twitter. Earlier this year, the company unveiled some of its machine learning research, showing how its algorithms can essentially upgrade the resolution of low-res videos using ordinary graphics cards. The Magic Pony team includes 11 PhDs, with their expertise ranging across computer vision, computational neuroscience, and deep learning.

Current deep learning technology is not advanced enough for computers to understand language, a major figure in the field said today. The ability to learn the way people learn -- through observation and experience -- is what Facebook will use to teach chatbots and computers to carry on a conversation like a human, said Yann LeCun, the head of Facebook's artificial intelligence (A.I.) research lab. LeCun spoke about A.I. and steps being taken to make virtual assistant M less reliant on human training at the 2016 Wired Business Conference, as Wired reported. Humans have played a role in the decision-making process for Facebook's M since the bot debuted last year, before the launch of the company's bot platform. Facebook has been researching ways to make machines understand language more independently.