Neural networks are changing the Internet. Inspired by the networks of neurons inside the human brain, these deep mathematical models can learn discrete tasks by analyzing enormous amounts of data. They've learned to recognize faces in photos, identify spoken commands, and translate text from one language to another. They're helping to choose what you see when you query the Google search engine or visit your Facebook News Feed. All this is sharpening the behavior of online services.

While everyone keeps saying that robots are not a job security threat, it is also true that robots are steadily getting better at tasks. Enter this little bad boy, the robot that won Amazon's Picking Challenge. A team of engineers from Netherland's TU Del have won this year's challenge, both in the picking and stowing finals. They dubbed their creation "Delft." The cool thing is their robot is no ordinary warehouse bot; it relies on a suction cup, a "two-fingered" gripper, and the combination of deep learning artificial intelligence and depth-sensing cameras to get the job done.

Everywhere I turn, I seem to encounter discussions on Machine Learning and Artificial Intelligence (AI). The Nasscom conference on Big Data and Analytics in June was heavily AI focused. The cover of last week's issue of The Economist reads "March of the Machines" with a special report on AI. Analytics websites are full of it. So why is the analytics community so upbeat about this technology?

In this benchmark, I've used a Windows 10 Pro 64 Bit computer with Intel Core i7 6700HQ 2.60 GHz with 32 Gb RAM and NVIDIA GeForce GTX 960M. As a programming environment, I've used Python 2.7 (Anaconda distribution) and Jupyter. The code I've written is this (without matplotlib functions and float32 numbers, in order to use the GPU): So, Numpy is on average 300% slower than Theano (with GPU support). The spikes should be due to CPU overload, multitasking or memory swapping. However, it's absolutely clear that Theano (I'm going to test also Tensorflow) should be the best choice if you want to implement deep learning algorithms (in particular if you have a good GPU).

I had a recent meeting with a person who was introduced to Machine Learning for the first time. It was interesting to know how someone totally new to the field would interpret what Machine Learning would be. He could instantly connect the term learning with what most Data Scientists would call Reinforcement Learning. A machine could observe phenomena and refine itself by itself is what he thought.It is weird that the one branch of Artificial Intelligence a layman could best connect to is the one least studied. I hope that Google Deepmind's recent work in Deep Reinforcement learning would change this.

A designer's guide to AI. Leveraging user centered design principles, the author rightly states that AI will enable designers to create bespoke experiences right out of the box for each user. Importantly, these experiences need to a) create emotionally-aware relationships with the user, b) respond to needs that haven't yet been explicitly expressed, c) prevent negative emotional responses when a user is upset with an AI-caused result and d) be sensitive to sociology. A list of further reading resources is included. There's been a resurgence of neuroscience-inspired AI architectures in the past few years, with Numenta being one of the leaders. Their VP of Research, Subutai Ahmad, argues that environmental sensory inference and behavior generation are intricately tied together, and critical for learning really general purpose representations.

Here's What We Can Expect from Deep Learning in 2016 and Beyond Robots Need "Common Sense" A.I. to Work in Our Uncertain World

The Sunway TiahuLight machine is the fastest supercomputer in the world running on the 10 million-core with a peak of 125 petaflops. The TaihuLight supercomputer is being harnessed for some interesting work on deep neural networks. What is fascinating here is that currently, the inference side of such workloads can scale to many processors, but the training side is often scale-limited hardware and software-wise. Fu described an ongoing project on the Sunway TaihuLight machine to develop an open source deep neural network library and make the appropriate architectural optimization for both high performance and efficiency on both the training and inference parts of deep learning workloads. "Based on this architecture, we can provide support for both single and double precision as well as fixed point," he explains.

I trained an LSTM Recurrent Neural Network (a deep learning algorithm) on the first four Harry Potter books. I then asked it to produce a chapter based on what it learned. He looked like Madame Maxime. When she strode up the wrong staircase to visit himself. "I'm afraid I've definitely been suspended from power, no chance -- indeed?" said Snape.

Naive Bayes Nearest Neighbour (NBNN) is a simple and effective framework which addresses many of the pitfalls of K-Nearest Neighbour (KNN) classification. It has yielded competitive results on several computer vision benchmarks. Its central tenet is that during NN search, a query is not compared to every example in a database, ignoring class information. Instead, NN searches are performed within each class, generating a score per class. A key problem with NN techniques, including NBNN, is that they fail when the data representation does not capture perceptual (e.g.~class-based) similarity. NBNN circumvents this by using independent engineered descriptors (e.g.~SIFT). To extend its applicability outside of image-based domains, we propose to learn a metric which captures perceptual similarity. Similar to how Neighbourhood Components Analysis optimizes a differentiable form of KNN classification, we propose "Class Conditional" metric learning (CCML), which optimizes a soft form of the NBNN selection rule. Typical metric learning algorithms learn either a global or local metric. However, our proposed method can be adjusted to a particular level of locality by tuning a single parameter. An empirical evaluation on classification and retrieval tasks demonstrates that our proposed method clearly outperforms existing learned distance metrics across a variety of image and non-image datasets.