While America struggles to retain net neutrality protections on a piecemeal state-by-state level, India just adopted its own seemingly-strong set of rules that largely ban artificial slowing or blocking of content. The Telecom Regulatory Authority of India (TRAI) proposed the rules back in November, and the country's official Telecommunications Commission (TC) approved them today, resolving a movement to update internet protections that began in 2015. From today onward, ISPs cannot perform actions involving "blocking, degrading, slowing down or granting preferential speeds or treatment to any content," per The Wire's report. Providers are also prohibited from providing zero-rated content. These include remote surgery or autonomous vehicles, which Sundarajan compared to laws allowing ambulances to ignore traffic rules.

Earlier this year Google unveiled Google Duplex -- a human-sounding robot able to have a conversation with a person on a telephone In many cases, the person couldn't distinguish that they were talking to a robot. It is, as Google explained, technology that is meant to carry out real world tasks over the phone, such as scheduling certain types of appointments. It is not meant for everyday conversations, Google also said -- rather Duplex has been trained in a specific domain, such as taking restaurant reservations, and can only hold a conversation in that domain. There is a school of thought that all of the current research and advancements in personal assistants, in artificial technology and in robotics are heading for a convergence of these technologies -- say a walking, talking robot that is able to run errands for you. This development is not here yet but as creations such as Duplex show, they are coming closer to reality.

The fear of the unknown is a basic human feeling, shared by most people. The fear mechanism is one of the most developed devices throughout our evolution. It's there to protect us from evildoers, from predators, from everything that strays from the ordinary, the routine, the repetitive. When asked about artificial intelligence's place in our lives, it usually activates the same sensors in people, eliciting the same response. It's the robots we read about in books, that we see in films.

IT'S robots that mostly come to mind when you ask people about the future of work. Robots taking our jobs, to be specific. And it's a reaction that's two centuries old, in a replay of Lancashire weavers attacking looms and stocking frames at the start of the first Industrial Revolution. A secondary reaction, among a much smaller group, is the creation of new jobs in the coming fourth Industrial Revolution. On the left side are old industries, where some workers are being replaced by robots.

Artificial intelligence (AI) is being used both to monitor and prevent crimes in many countries. AI is used in such areas as bomb detection and deactivation, surveillance, prediction, social media scanning and interviewing suspects. However, for all the hype and hoopla around AI, there is scope for growth of its role in crime management. Currently, a few issues are proving problematic. AI is not uniformly engaged across countries in crime management. There is fierce debate on the ethical boundaries of AI, compelling law enforcement authorities to tread carefully.

We study the problem of attacking a machine learning model in the hard-label black-box setting, where no model information is revealed except that the attacker can make queries to probe the corresponding hard-label decisions. This is a very challenging problem since the direct extension of state-of-the-art white-box attacks (e.g., CW or PGD) to the hard-label black-box setting will require minimizing a non-continuous step function, which is combinatorial and cannot be solved by a gradient-based optimizer. The only current approach is based on random walk on the boundary, which requires lots of queries and lacks convergence guarantees. We propose a novel way to formulate the hard-label black-box attack as a real-valued optimization problem which is usually continuous and can be solved by any zeroth order optimization algorithm. For example, using the Randomized Gradient-Free method, we are able to bound the number of iterations needed for our algorithm to achieve stationary points. We demonstrate that our proposed method outperforms the previous random walk approach to attacking convolutional neural networks on MNIST, CIFAR, and ImageNet datasets. More interestingly, we show that the proposed algorithm can also be used to attack other discrete and non-continuous machine learning models, such as Gradient Boosting Decision Trees (GBDT).

Deep reinforcement learning has recently shown many impressive successes. However, one major obstacle towards applying such methods to real-world problems is their lack of data-efficiency. To this end, we propose the Bottleneck Simulator: a model-based reinforcement learning method which combines a learned, factorized transition model of the environment with rollout simulations to learn an effective policy from few examples. The learned transition model employs an abstract, discrete (bottleneck) state, which increases sample efficiency by reducing the number of model parameters and by exploiting structural properties of the environment. We provide a mathematical analysis of the Bottleneck Simulator in terms of fixed points of the learned policy, which reveals how performance is affected by four distinct sources of error: an error related to the abstract space structure, an error related to the transition model estimation variance, an error related to the transition model estimation bias, and an error related to the transition model class bias. Finally, we evaluate the Bottleneck Simulator on two natural language processing tasks: a text adventure game and a real-world, complex dialogue response selection task. On both tasks, the Bottleneck Simulator yields excellent performance beating competing approaches.

Scikit-multiflow is a multi-output/multi-label and stream data mining framework for the Python programming language. Conceived to serve as a platform to encourage democratization of stream learning research, it provides multiple state of the art methods for stream learning, stream generators and evaluators. scikit-multiflow builds upon popular open source frameworks including scikit-learn, MOA and MEKA. Development follows the FOSS principles and quality is enforced by complying with PEP8 guidelines and using continuous integration and automatic testing. The source code is publicly available at https://github.com/scikit-multiflow/scikit-multiflow.

Scarcity of labeled data is one of the most frequent problems faced in machine learning. This is particularly true in relation extraction in text mining, where large corpora of texts exists in many application domains, while labeling of text data requires an expert to invest much time to read the documents. Overall, state-of-the art models, like the convolutional neural network used in this paper, achieve great results when trained on large enough amounts of labeled data. However, from a practical point of view the question arises whether this is the most efficient approach when one takes the manual effort of the expert into account. In this paper, we report on an alternative approach where we first construct a relation extraction model using distant supervision, and only later make use of a domain expert to refine the results. Distant supervision provides a mean of labeling data given known relations in a knowledge base, but it suffers from noisy labeling. We introduce an active learning based extension, that allows our neural network to incorporate expert feedback and report on first results on a complex data set.

Without any specific way for imbalance data classification, artificial intelligence algorithm cannot recognize data from minority classes easily. In general, modifying the existing algorithm by assuming that the training data is imbalanced, is the only way to handle imbalance data. However, for a normal data handling, this way mostly produces a deficient result. In this research, we propose a class expert generative adversarial network (CE-GAN) as the solution for imbalance data classification. CE-GAN is a modification in deep learning algorithm architecture that does not have an assumption that the training data is imbalance data. Moreover, CE-GAN is designed to identify more detail about the character of each class before classification step. CE-GAN has been proved in this research to give a good performance for imbalance data classification.