The Arcade Learning Environment (ALE) is an evaluation platform that poses the challenge of building AI agents with general competency across dozens of Atari 2600 games. It supports a variety of different problem settings and it has been receiving increasing attention from the scientific community, leading to some high-profile success stories such as the much publicized Deep Q-Networks (DQN). In this article we take a big picture look at how the ALE is being used by the research community. We show how diverse the evaluation methodologies in the ALE have become with time, and highlight some key concerns when evaluating agents in the ALE. We use this discussion to present some methodological best practices and provide new benchmark results using these best practices. To further the progress in the field, we introduce a new version of the ALE that supports multiple game modes and provides a form of stochasticity we call sticky actions. We conclude this big picture look by revisiting challenges posed when the ALE was introduced, summarizing the state-of-the-art in various problems and highlighting problems that remain open.

Convolutional neural networks (CNNs) have been extensively applied for image recognition problems giving stateof-the-art results on recognition, detection, segmentation and retrieval. In this work we propose and evaluate several deep neural network architectures to combine image information across a video over longer time periods than previously attempted. We propose two methods capable of handling full length videos. The first method explores various convolutional temporal feature pooling architectures, examining the various design choices which need to be made when adapting a CNN for this task. The second proposed method explicitly models the video as an ordered sequence of frames. For this purpose we employ a recurrent neural network that uses Long Short-Term Memory (LSTM) cells which are connected to the output of the underlying CNN. Our best networks exhibit significant performance improvements over previously published results on the Sports 1 million dataset (73.1% vs. 60.9%) and the UCF-101 datasets with (88.6% vs. 88.0%) and without additional optical flow information (82.6% vs. 73.0%).

Causation by omission is instantiated when an effect occurs from an absence, as in The absence of nicotine causes withdrawal or Not watering the plant caused it to wilt. The phenomenon has been viewed as an insurmountable problem for process theories of causation, which specify causation in terms of conserved quantities, like force, but not for theories that specify causation in terms of statistical or counterfactual dependencies. A new account of causation challenges these assumptions. According to the force theory, absences are causal when the removal of a force leads to an effect. Evidence in support of this account was found in 3 experiments in which people classified animations of complex causal chains involving force removal, as well as chains involving virtual forces, that is, forces that were anticipated but never realized. In a 4th experiment, the force theory's ability to predict synonymy relationships between different types of causal expressions provided further evidence for this theory over dependency theories. The findings show not only how causation by omission can be grounded in the physical world but also why only certain absences, among the potentially infinite number of absences, are causal.

We introduce TextWorld, a sandbox learning environment for the training and evaluation of RL agents on text-based games. TextWorld is a Python library that handles interactive playthrough of text games, as well as backend functions like state tracking and reward assignment. It comes with a curated list of games whose features and challenges we have analyzed. More significantly, it enables users to handcraft or automatically generate new games. Its generative mechanisms give precise control over the difficulty, scope, and language of constructed games, and can be used to relax challenges inherent to commercial text games like partial observability and sparse rewards. By generating sets of varied but similar games, TextWorld can also be used to study generalization and transfer learning. We cast text-based games in the Reinforcement Learning formalism, use our framework to develop a set of benchmark games, and evaluate several baseline agents on this set and the curated list.

This article addresses the challenge of learning to play many different video games with little domainspecific knowledge. Specifically, it introduces a neuro-evolution approach to general Atari 2600 game playing. Four neuro-evolution algorithms were paired with three different state representations and evaluated on a set of 61 Atari games. The neuro-evolution agents represent different points along the spectrum of algorithmic sophistication -including weight evolution on topologically fixed neural networks (Conventional Neuro-evolution), Covariance Matrix Adaptation Evolution Strategy (CMA-ES), evolution of network topology and weights (NEAT), and indirect network encoding (HyperNEAT). State representations include an object representation of the game screen, the raw pixels of the game screen, and seeded noise (a comparative baseline). Results indicate that direct-encoding methods work best on compact state representations while indirect-encoding methods (i.e. HyperNEAT) allow scaling to higher-dimensional representations (i.e. the raw game screen). Previous approaches based on temporaldifference learning had trouble dealing with the large state spaces and sparse reward gradients often found in Atari games. Neuro-evolution ameliorates these problems and evolved policies achieve state-of-the-art results, even surpassing human high scores on three games. These results suggest that neuro-evolution is a promising approach to general video game playing.