The Fujitsu Ltd. RoBoPin communication robot at the Combined Exhibition of Advanced Technologies in Japan on Oct. 4, 2016. Kai-Fu Lee, chairman and CEO of Innovation Works delivered the opening keynote of G-Summit. Lee speaks with authority as a pioneer in speech-recognition. He developed the world's first speaker-independent, continuous speech recognition system and later established Google China. "The future of jobs will change and reinvent every imaginable part of work (from AI)," said Lee.

Here is some great content that you won't find on Google. I hope to add more in the future, and feel free to email me at [email protected] if you want to add some of your links. It is easy to remember this page: the URL is BannedOnGoogle.com. It's not that the articles below are black-listed by Google, but most likely, Google algorithms are not working properly: either they can't find the page or can only find the mobile version (issue with Google's indexation algorithm) or instead, when searching for the article's title, Google returns irrelevant articles, or a copy of the article that is illegaly stolen and hosted elsewhere (issue with Google's web page scoring / ranking / attribution algorithms.) To learn more about these problems (how to design a good search engine or improve Google) click here, and here.

Amazon Lex, Amazon Polly, and Amazon Rekognition are based on the same proven, highly scalable Amazon technology built by the thousands of deep learning and machine learning experts across the company. Amazon AI services all provide high-quality, high-accuracy AI capabilities that are scalable and cost-effective. Amazon AI services are fully managed services so there are no deep learning algorithms to build, no machine learning models to train, and no up-front commitments or infrastructure investments required. This frees developers to focus on defining and building an entirely new generation of apps that can see, hear, speak, understand, and interact with the world around them. To learn more about Amazon Lex, Amazon Polly, or Amazon Rekognition, visit: https://aws.amazon.com/amazon-ai

Amazon's cloud services platform Amazon Web Services recently announced three AI services it said will make it easy for developers to build apps that can understand natural language, turn text into speech, have conversations using voice or text, analyze images and recognize faces, objects and scenes. This, in turn, underscores the increasing importance of AI to consumers, brands and marketers, but also raises some questions about how it will – and should – be developed. Building apps with AI capabilities has been challenging to date because doing so requires access to vast amounts of data and specialized expertise in machine learning and neural networks, Amazon said in a press release. "The combination of better algorithms and broad access to massive amounts of data and cost-effective computing power provided by the cloud is making AI a reality for application developers," added Raju Gulabani, vice president of databases, analytics and AI at AWS, in a statement. "Thousands of machine learning and deep learning experts across Amazon have been developing AI technologies for years to predict what customers might like to read, to drive efficiencies in our fulfillment centers through robotics and computer vision technologies and to give customers our AI-powered virtual assistant, Alexa. Now, we are making the technology underlying these innovations available to any developer…we are excited to see how customers use Amazon Lex, Amazon Polly and Amazon Rekognition to build a new generation of apps that have human-like intelligence and can see, hear, speak and interact with people and their environments."

In this work we investigate intra-day patterns of activity on a population of 7,261 users of mobile health wearable devices and apps. We show that: (1) using intra-day step and sleep data recorded from passive trackers significantly improves classification performance on self-reported chronic conditions related to mental health and nervous system disorders, (2) Convolutional Neural Networks achieve top classification performance vs. baseline models when trained directly on multivariate time series of activity data, and (3) jointly predicting all condition classes via multi-task learning can be leveraged to extract features that generalize across data sets and achieve the highest classification performance.

In an independence model, the triplets that represent conditional independences between singletons are called elementary. It is known that the elementary triplets represent the independence model unambiguously under some conditions. In this paper, we show how this representation helps performing some operations with independence models, such as finding the dominant triplets or a minimal independence map of an independence model, or computing the union or intersection of a pair of independence models, or performing causal reasoning. For the latter, we rephrase in terms of conditional independences some of Pearl's results for computing causal effects.

Until recently, research on artificial neural networks was largely restricted to systems with only two types of variable: Neural activities that represent the current or recent input and weights that learn to capture regularities among inputs, outputs and payoffs. There is no good reason for this restriction. Synapses have dynamics at many different time-scales and this suggests that artificial neural networks might benefit from variables that change slower than activities but much faster than the standard weights. These "fast weights" can be used to store temporary memories of the recent past and they provide a neurally plausible way of implementing the type of attention to the past that has recently proved very helpful in sequence-to-sequence models. By using fast weights we can avoid the need to store copies of neural activity patterns.

Current deep learning architectures are growing larger in order to learn from complex datasets. These architectures require giant matrix multiplication operations to train millions of parameters. Conversely, there is another growing trend to bring deep learning to low-power, embedded devices. The matrix operations, associated with both training and testing of deep networks, are very expensive from a computational and energy standpoint. We present a novel hashing based technique to drastically reduce the amount of computation needed to train and test deep networks. Our approach combines recent ideas from adaptive dropouts and randomized hashing for maximum inner product search to select the nodes with the highest activation efficiently. Our new algorithm for deep learning reduces the overall computational cost of forward and back-propagation by operating on significantly fewer (sparse) nodes. As a consequence, our algorithm uses only 5% of the total multiplications, while keeping on average within 1% of the accuracy of the original model. A unique property of the proposed hashing based back-propagation is that the updates are always sparse. Due to the sparse gradient updates, our algorithm is ideally suited for asynchronous and parallel training leading to near linear speedup with increasing number of cores. We demonstrate the scalability and sustainability (energy efficiency) of our proposed algorithm via rigorous experimental evaluations on several real datasets.

Dropout has been witnessed with great success in training deep neural networks by independently zeroing out the outputs of neurons at random. It has also received a surge of interest for shallow learning, e.g., logistic regression. However, the independent sampling for dropout could be suboptimal for the sake of convergence. In this paper, we propose to use multinomial sampling for dropout, i.e., sampling features or neurons according to a multinomial distribution with different probabilities for different features/neurons. To exhibit the optimal dropout probabilities, we analyze the shallow learning with multinomial dropout and establish the risk bound for stochastic optimization. By minimizing a sampling dependent factor in the risk bound, we obtain a distribution-dependent dropout with sampling probabilities dependent on the second order statistics of the data distribution. To tackle the issue of evolving distribution of neurons in deep learning, we propose an efficient adaptive dropout (named \textbf{evolutional dropout}) that computes the sampling probabilities on-the-fly from a mini-batch of examples. Empirical studies on several benchmark datasets demonstrate that the proposed dropouts achieve not only much faster convergence and but also a smaller testing error than the standard dropout. For example, on the CIFAR-100 data, the evolutional dropout achieves relative improvements over 10\% on the prediction performance and over 50\% on the convergence speed compared to the standard dropout.

The AUV three-dimension path planning in complex turbulent underwater environment is investigated in this research, in which static current map data and uncertain static-moving time variant obstacles are taken into account. Robustness of AUVs path planning to this strong variability is known as a complex NP-hard problem and is considered a critical issue to ensure vehicles safe deployment. Efficient evolutionary techniques have substantial potential of handling NP hard complexity of path planning problem as more powerful and fast algorithms among other approaches for mentioned problem. For the purpose of this research Differential Evolution (DE) technique is conducted to solve the AUV path planning problem in a realistic underwater environment. The path planners designed in this paper are capable of extracting feasible areas of a real map to determine the allowed spaces for deployment, where coastal area, islands, static/dynamic obstacles and ocean current is taken into account and provides the efficient path with a small computation time. The results obtained from analyze of experimental demonstrate the inherent robustness and drastic efficiency of the proposed scheme in enhancement of the vehicles path planning capability in coping undesired current, using useful current flow, and avoid colliding collision boundaries in a real-time manner. The proposed approach is also flexible and strictly respects to vehicle's kinematic constraints resisting current instabilities.