Have you ever experienced the dread of overhearing two people, speaking a language you don't understand, begin laughing wildly? You just have to wonder what it is they're talking about, and if it's a joke at your expense. Heck, maybe you even check your teeth to make sure you aren't walking around with half of your lunchtime ham sandwich stuck to your gums. As Wired reports, researchers at OpenAI have made some huge strides in getting bots to communicate with each other, and without actually telling them how to do so. The group published a research paper earlier this week explaining exactly how they were able to accomplish the complex task, and it's all based on reinforcement learning.

The old adage that practice makes perfect applies to machines as well, as many of today's artificially intelligent devices rely on repetition to learn. Deep-learning algorithms are designed to allow AI devices to glean knowledge from datasets and then apply what they've learned to concrete situations. For example, an AI system is fed data about how the sky is usually blue, which allows it to later recognize the sky in a series of images. Complex work can be accomplished using this method, but it certainly leaves something to be desired. For instance, could the same results be obtained by exposing deep-learning AI to fewer examples?

Model predictive control (MPC) is a popular control method that has proved effective for robotics, among other fields. MPC performs re-planning at every time step. Re-planning is done with a limited horizon per computational and real-time constraints and often also for robustness to potential model errors. However, the limited horizon leads to suboptimal performance. In this work, we consider the iterative learning setting, where the same task can be repeated several times, and propose a policy improvement scheme for MPC. The main idea is that between executions we can, offline, run MPC with a longer horizon, resulting in a hindsight plan. To bring the next real-world execution closer to the hindsight plan, our approach learns to re-shape the original cost function with the goal of satisfying the following property: short horizon planning (as realistic during real executions) with respect to the shaped cost should result in mimicking the hindsight plan. This effectively consolidates long-term reasoning into the short-horizon planning. We empirically evaluate our approach in contact-rich manipulation tasks both in simulated and real environments, such as peg insertion by a real PR2 robot.

During the past decade, several areas of speech and language understanding have witnessed substantial breakthroughs from the use of data-driven models. In the area of dialogue systems, the trend is less obvious, and most practical systems are still built through significant engineering and expert knowledge. Nevertheless, several recent results suggest that data-driven approaches are feasible and quite promising. To facilitate research in this area, we have carried out a wide survey of publicly available datasets suitable for data-driven learning of dialogue systems. We discuss important characteristics of these datasets, how they can be used to learn diverse dialogue strategies, and their other potential uses. We also examine methods for transfer learning between datasets and the use of external knowledge. Finally, we discuss appropriate choice of evaluation metrics for the learning objective.

In this paper, we study the problem of designing objective functions for machine learning problems defined on finite \emph{sets}. In contrast to traditional objective functions defined for machine learning problems operating on finite dimensional vectors, the new objective functions we propose are operating on finite sets and are invariant to permutations. Such problems are widespread, ranging from estimation of population statistics \citep{poczos13aistats}, via anomaly detection in piezometer data of embankment dams \citep{Jung15Exploration}, to cosmology \citep{Ntampaka16Dynamical,Ravanbakhsh16ICML1}. Our main theorem characterizes the permutation invariant objective functions and provides a family of functions to which any permutation invariant objective function must belong. This family of functions has a special structure which enables us to design a deep network architecture that can operate on sets and which can be deployed on a variety of scenarios including both unsupervised and supervised learning tasks. We demonstrate the applicability of our method on population statistic estimation, point cloud classification, set expansion, and image tagging.

In this work, we investigate a novel training procedure to learn a generative model as the transition operator of a Markov chain, such that, when applied repeatedly on an unstructured random noise sample, it will denoise it into a sample that matches the target distribution from the training set. The novel training procedure to learn this progressive denoising operation involves sampling from a slightly different chain than the model chain used for generation in the absence of a denoising target. In the training chain we infuse information from the training target example that we would like the chains to reach with a high probability. The thus learned transition operator is able to produce quality and varied samples in a small number of steps. Experiments show competitive results compared to the samples generated with a basic Generative Adversarial Net

We introduce the value iteration network (VIN): a fully differentiable neural network with a `planning module' embedded within. VINs can learn to plan, and are suitable for predicting outcomes that involve planning-based reasoning, such as policies for reinforcement learning. Key to our approach is a novel differentiable approximation of the value-iteration algorithm, which can be represented as a convolutional neural network, and trained end-to-end using standard backpropagation. We evaluate VIN based policies on discrete and continuous path-planning domains, and on a natural-language based search task. We show that by learning an explicit planning computation, VIN policies generalize better to new, unseen domains.

The company, which is on a mission to "solve intelligence," has hired some of the brightest minds in the world, including academics from Oxbridge and research scientists from firms like Facebook and Microsoft. Exactly how DeepMind's researchers work together has been something of a mystery but the FT story sheds new light on the matter. Researchers at DeepMind are divided into four main groups, including a "neuroscience" group and a "frontiers" group, according to the report. The frontiers group is said to be full of physicists and mathematicians who are tasked with testing some of the most futuristic AI theories. "We've hired 250 of the world's best scientists, so obviously they're here to let their creativity run riot, and we try and create an environment that's perfect for that," DeepMind CEO Demis Hassabis told the FT.

BY PATRICK COX: In 2011, scientists made one of the most important discoveries in the history of AI development. They found that graphics processing units (GPUs) are far better at simulating biological learning than central processing units (CPUs). In retrospect, it seems obvious. Human brains are much more like GPUs than CPUs. Both brains and GPUs rely on parallel processing that simulates and predicts real world physics.