Pic: Getty ImagesArtificial intelligence, blockchain, cryptocurrencies - three terms you need to scatter through your conversation if you want to come across as a tech guru. On Tech Tent this week we examine these trends and ask a futurologist to predict which of them will make rapid progress over the next decade. This week saw another major achievement by Google's Deep Mind, when it showed that a neural network could learn to play Go in just three days, without even looking at how humans play this complex game. AlphaGo Zero took on the previous version of the program, developed with human expertise, and beat it by 100 games to nil. The company now hopes to use this technique in other areas such as drug development.

There's been way too much fear-mongering news articles around the latest version of DeepMind's AlphaGo. Let's set the record straight, AlphaGo is an incredible technology and it's not terrifying at all. I'll go over the technical details of how AlphaGo really works; a mixture of deep learning and reinforcement learning. That's what keeps me going.

HPE announced new purpose-built platforms and services capabilities to help companies simplify the adoption of Artificial Intelligence, with an initial focus on a key subset of AI known as deep learning. Inspired by the human brain, deep learning is typically implemented for challenging tasks such as image and facial recognition, image classification and voice recognition. To take advantage of deep learning, enterprises need a high performance compute infrastructure to build and train learning models that can manage large volumes of data to recognize patterns in audio, images, videos, text and sensor data. Many organizations lack several integral requirements to implement deep learning, including expertise and resources; sophisticated and tailored hardware and software infrastructure; and the integration capabilities required to assimilate different pieces of hardware and software to scale AI systems. Based on the HPE Apollo 6500 system in collaboration with Bright Computing to enable rapid deep learning application development, this solution includes pre-configured deep learning software frameworks, libraries, automated software updates and cluster management optimized for deep learning and supports NVIDIA Tesla V100 GPUs.

From the moment we are born, we begin using our senses to build a coherent model of the world. As we grow, we constantly refine our model and access it effortlessly as we go about our lives. If we see a ball rolling onto the street, we might reason that a child could have kicked it there. When asked to pour a glass of wine, we wouldn't search for a bottle opener if the wine is already in the decanter. If we are told, "Sally hammered the nail into the floor," and asked whether the nail was vertical or horizontal, we can imagine the scenario with the appropriate level of detail to answer confidently: vertical [1].

Which authors of this paper are endorsers? Disable MathJax (What is MathJax?)

Developers, data scientists, and researchers are solving today's complex challenges with breakthroughs in artificial intelligence, deep learning, and high performance computing (HPC). NVIDIA is working with Amazon Web Services to offer the newest and most powerful GPU-accelerated cloud service based on the latest NVIDIA Volta architecture: Amazon EC2 P3 instance. Using up to eight NVIDIA Tesla V100 GPUs, you will be able to train your neural networks with massive data sets using any of the major deep learning frameworks faster than ever before. Then use the capabilities of GPU parallel computing, running billions of computations, to infer and identify known patterns or objects. With over 500 GPU-accelerated HPC applications accelerated, including the top ten HPC applications and every deep learning framework, you can quickly tap into the power of the Tesla V100 GPUs on AWS to boost performance, scale-out, accelerate time to results, and save money.

In order for machine learning to be deployed and trusted in many applications, it is crucial to be able to reliably explain why the machine learning algorithm makes certain predictions. For example, if an algorithm classifies a given pathology image to be a malignant tumor, then the doctor may need to know which parts of the image led the algorithm to this classification. How to interpret black-box predictors is thus an important and active area of research. A fundamental question is: how much can we trust the interpretation itself? In this paper, we show that interpretation of deep learning predictions is extremely fragile in the following sense: two perceptively indistinguishable inputs with the same predicted label can be assigned very different interpretations. We systematically characterize the fragility of several widely-used feature-importance interpretation methods (saliency maps, relevance propagation, and DeepLIFT) on ImageNet and CIFAR-10. Our experiments show that even small random perturbation can change the feature importance and new systematic perturbations can lead to dramatically different interpretations without changing the label. We extend these results to show that interpretations based on exemplars (e.g. influence functions) are similarly fragile. Our analysis of the geometry of the Hessian matrix gives insight on why fragility could be a fundamental challenge to the current interpretation approaches.

Connections between nodes of fully connected neural networks are usually represented by weight matrices. In this article, functional transfer matrices are introduced as alternatives to the weight matrices: Instead of using real weights, a functional transfer matrix uses real functions with trainable parameters to represent connections between nodes. Multiple functional transfer matrices are then stacked together with bias vectors and activations to form deep functional transfer neural networks. These neural networks can be trained within the framework of back-propagation, based on a revision of the delta rules and the error transmission rule for functional connections. In experiments, it is demonstrated that the revised rules can be used to train a range of functional connections: 20 different functions are applied to neural networks with up to 10 hidden layers, and most of them gain high test accuracies on the MNIST database. It is also demonstrated that a functional transfer matrix with a memory function can roughly memorise a non-cyclical sequence of 400 digits.

Availability of an explainable deep learning model that can be applied to practical real world scenarios and in turn, can consistently, rapidly and accurately identify specific and minute traits in applicable fields of biological sciences, is scarce. Here we consider one such real world example viz., accurate identification, classification and quantification of biotic and abiotic stresses in crop research and production. Up until now, this has been predominantly done manually by visual inspection and require specialized training. However, such techniques are hindered by subjectivity resulting from inter- and intra-rater cognitive variability. Here, we demonstrate the ability of a machine learning framework to identify and classify a diverse set of foliar stresses in the soybean plant with remarkable accuracy. We also present an explanation mechanism using gradient-weighted class activation mapping that isolates the visual symptoms used by the model to make predictions. This unsupervised identification of unique visual symptoms for each stress provides a quantitative measure of stress severity, allowing for identification, classification and quantification in one framework. The learnt model appears to be agnostic to species and make good predictions for other (non-soybean) species, demonstrating an ability of transfer learning.