Bayesian reward learning from demonstrations enables rigorous safety and uncertainty analysis when performing imitation learning. However, Bayesian reward learning methods are typically computationally intractable for complex control problems. We propose a highly efficient Bayesian reward learning algorithm that scales to high-dimensional imitation learning problems by first pre-training a low-dimensional feature encoding via self-supervised tasks and then leveraging preferences over demonstrations to perform fast Bayesian inference. We evaluate our proposed approach on the task of learning to play Atari games from demonstrations, without access to the game score. For Atari games our approach enables us to generate 100,000 samples from the posterior over reward functions in only 5 minutes using a personal laptop. Furthermore, our proposed approach achieves comparable or better imitation learning performance than state-of-the-art methods that only find a point estimate of the reward function. Finally, we show that our approach enables efficient high-confidence policy performance bounds. We show that these high-confidence performance bounds can be used to rank the performance and risk of a variety of evaluation policies, despite not having samples of the reward function. We also show evidence that high-confidence performance bounds can be used to detect reward hacking in complex imitation learning problems.

An important problem that arises in reinforcement learning and Monte Carlo methods is estimating quantities defined by the stationary distribution of a Markov chain. In many real-world applications, access to the underlying transition operator is limited to a fixed set of data that has already been collected, without additional interaction with the environment being available. We show that consistent estimation remains possible in this challenging scenario, and that effective estimation can still be achieved in important applications. Our approach is based on estimating a ratio that corrects for the discrepancy between the stationary and empirical distributions, derived from fundamental properties of the stationary distribution, and exploiting constraint reformulations based on variational divergence minimization. The resulting algorithm, GenDICE, is straightforward and effective. We prove its consistency under general conditions, provide an error analysis, and demonstrate strong empirical performance on benchmark problems, including off-line PageRank and off-policy policy evaluation.

Recent advances in machine learning highlight successes on a small set of tasks where a large number of labeled examples have been collected and exploited. These include image classification with 1.2 million labeled examples Deng et al. (2009) and French-English machine translation with 40 million paired sentences Bojar et al. (2014). For common tasks, however, collecting clean labels is costly, as they require human expertise (as in medical imaging) or physical interactions (as in robotics), for example. Thus collected real-world datasets follow a long-tailed distribution, in which a dominant set of tasks only have a small number of training examples Wang et al. (2017). Inspired by human ingenuity in quickly solving novel problems by leveraging prior experience, meta-learning approaches aim to jointly learn from past experience to quickly adapt to new tasks with little available data Schmidhuber (1987); Thrun & Pratt (2012). This has had a significant impact in few-shot supervised learning, where each task is associated with only a few training examples. By leveraging structural similarities among those tasks, one can achieve accuracy far greater than what can be achieved for each task in isolation Finn et al. (2017); Ravi & Larochelle (2016); Koch et al. (2015); Oreshkin et al. (2018); Triantafillou et al. (2019); Rusu et al. (2018). The success of such approaches hinges on the following fundamental question: When can we jointly train small data tasks to achieve the accuracy of large data tasks? We investigate this tradeoff under a canonical scenario where the tasks are linear regressions in d-dimensions and the regression parameters are drawn i.i.d.

Statistical inference using pairwise comparison data has been an effective approach to analyzing complex and sparse networks. In this paper we propose a general framework for modeling the mutual interaction in a probabilistic network, which enjoys ample flexibility in terms of parametrization. Within this set-up, we establish that the maximum likelihood estimator (MLE) for the latent scores of the subjects is uniformly consistent under a near-minimal condition on network sparsity. This condition is sharp in terms of the leading order asymptotics describing the sparsity. The proof utilizes a novel chaining technique based on the error-induced metric as well as careful counting of comparison graph structures. Our results guarantee that the MLE is a valid estimator for inference in large-scale comparison networks where data is asymptotically deficient. Numerical simulations are provided to complement the theoretical analysis.

The key distinguishing property of a Bayesian approach is marginalization, rather than using a single setting of weights. Bayesian marginalization can particularly improve the accuracy and calibration of modern deep neural networks, which are typically underspecified by the data, and can represent many compelling but different solutions. We show that deep ensembles provide an effective mechanism for approximate Bayesian marginalization, and propose a related approach that further improves the predictive distribution by marginalizing within basins of attraction, without significant overhead. We also investigate the prior over functions implied by a vague distribution over neural network weights, explaining the generalization properties of such models from a probabilistic perspective. From this perspective, we explain results that have been presented as mysterious and distinct to neural network generalization, such as the ability to fit images with random labels, and show that these results can be reproduced with Gaussian processes. Finally, we provide a Bayesian perspective on tempering for calibrating predictive distributions.

Policy gradient methods in reinforcement learning update policy parameters by taking steps in the direction of an estimated gradient of policy value. In this paper, we consider the statistically efficient estimation of policy gradients from off-policy data, where the estimation is particularly non-trivial. We derive the asymptotic lower bound on the feasible mean-squared error in both Markov and non-Markov decision processes and show that existing estimators fail to achieve it in general settings. We propose a meta-algorithm that achieves the lower bound without any parametric assumptions and exhibits a unique 3-way double robustness property. We discuss how to estimate nuisances that the algorithm relies on. Finally, we establish guarantees on the rate at which we approach a stationary point when we take steps in the direction of our new estimated policy gradient.

I wrote this paper because technology can really improve people's lives. With it, we can live longer in a healthy body, save time through increased efficiency and automation, and make better decisions. To get to the next level, we need to start looking at intelligence from a much broader perspective, and promote international interdisciplinary collaborations. Section 1 of this paper delves into sociology and social psychology to explain that the mechanisms underlying intelligence are inherently social. Section 2 proposes a method to classify intelligence, and describes the differences between weak and strong intelligence. Section 3 examines the Chinese Room argument from a different perspective. It demonstrates that a Turing-complete machine cannot have strong intelligence, and considers the modifications necessary for a computer to be intelligent and have understanding. Section 4 argues that the existential risk caused by the technological explosion of a single agent should not be of serious concern. Section 5 looks at the AI control problem and argues that it is impossible to build a super-intelligent machine that will do what it creators want. By using insights from biology, it also proposes a solution to the control problem. Section 6 discusses some of the implications of strong intelligence. Section 7 lists the main challenges with deep learning, and asserts that radical changes will be required to reach strong intelligence. Section 8 examines a neuroscience framework that could help explain how a cortical column works. Section 9 lays out the broad strokes of a road map towards strong intelligence. Finally, section 10 analyzes the impacts and the challenges of greater intelligence.

Cyber-attacks can occur at machine speeds that are far too fast for human-in-the-loop (or sometimes on-the-loop) decision making to be a viable option. Although human inputs are still important, a defensive Artificial Intelligence (AI) system must have considerable autonomy in these circumstances. When the AI system is model-based, its behavior responses can be aligned with risk-aware cost/benefit tradeoffs that are defined by user-supplied preferences that capture the key aspects of how human operators understand the system, the adversary and the mission. This paper describes an approach to automated cyber response that is designed along these lines. We combine a simulation of the system to be defended with an anytime online planner to solve cyber defense problems characterized as partially observable Markov decision problems (POMDPs).

Classifying our data and predicting the outcomes from our historical data are huge tasks at the moment. For performing these tasks, we have a robust family of Supervised Learning Algorithms called Naive Bayes' Classifiers. Naive Bayes' Classifiers are wholly based on the Bayes' Theorem which gives us the probability of an event, given that another event has already occurred. This is symbolically expressed as P(A B), i.e. Probability of event A will occur given that event B has already occurred.

Most current reinforcement learning algorithms are not capable of effectively handling such a large number of possible actions per turn. Poor sample efficiency, consequently, results in agents that are unable to pass bottleneck states, where they are unable to proceed because they do not see the right action sequence to pass the bottleneck enough times to be sufficiently reinforced. Building on prior work using knowledge graphs in reinforcement learning, we introduce two new game state exploration strategies. We compare our exploration strategies against strong baselines on the classic text-adventure game, Zork1, where prior agent have been unable to get past a bottleneck where the agent is eaten by a Grue.