Open-source reinforcement learning (RL) environments have played a crucial role in driving progress in the development of AI algorithms. In modern RL research, there is a need for simulated environments that are performant, scalable, and modular to enable their utilization in a wider range of potential real-world applications. Therefore, we present Jumanji, a suite of diverse RL environments specifically designed to be fast, flexible, and scalable. Jumanji provides a suite of environments focusing on combinatorial problems frequently encountered in industry, as well as challenging general decision-making tasks. By leveraging the efficiency of JAX and hardware accelerators like GPUs and TPUs, Jumanji enables rapid iteration of research ideas and large-scale experimentation, ultimately empowering more capable agents. Unlike existing RL environment suites, Jumanji is highly customizable, allowing users to tailor the initial state distribution and problem complexity to their needs. Furthermore, we provide actor-critic baselines for each environment, accompanied by preliminary findings on scaling and generalization scenarios. Jumanji aims to set a new standard for speed, adaptability, and scalability of RL environments.

We consider the problem of subset selection where one is given multiple rankings of items and the goal is to select the highest ``quality'' subset. Score functions from the multiwinner voting literature have been used to aggregate rankings into quality scores for subsets. We study this setting of subset selection problems when, in addition, rankings may contain systemic or unconscious biases toward a group of items. For a general model of input rankings and biases, we show that requiring the selected subset to satisfy group fairness constraints can improve the quality of the selection with respect to unbiased rankings. Importantly, we show that for fairness constraints to be effective, different multiwinner score functions may require a drastically different number of rankings: While for some functions, fairness constraints need an exponential number of rankings to recover a close-to-optimal solution, for others, this dependency is only polynomial. This result relies on a novel notion of ``smoothness'' of submodular functions in this setting that quantifies how well a function can ``correctly'' assess the quality of items in the presence of bias. The results in this paper can be used to guide the choice of multiwinner score functions for the subset selection setting considered here; we additionally provide a tool to empirically enable this.

In efforts to keep up with the rapid progress and use of large language models, gender bias research is becoming more prevalent in NLP. Non-English bias research, however, is still in its infancy with most work focusing on English. In our work, we study how grammatical gender bias relating to politeness levels manifests in Japanese and Korean language models. Linguistic studies in these languages have identified a connection between gender bias and politeness levels, however it is not yet known if language models reproduce these biases. We analyze relative prediction probabilities of the male and female grammatical genders using templates and find that informal polite speech is most indicative of the female grammatical gender, while rude and formal speech is most indicative of the male grammatical gender. Further, we find politeness levels to be an attack vector for allocational gender bias in cyberbullying detection models. Cyberbullies can evade detection through simple techniques abusing politeness levels. We introduce an attack dataset to (i) identify representational gender bias across politeness levels, (ii) demonstrate how gender biases can be abused to bypass cyberbullying detection models and (iii) show that allocational biases can be mitigated via training on our proposed dataset. Through our findings we highlight the importance of bias research moving beyond its current English-centrism.

Temporal-difference (TD) learning is widely regarded as one of the most popular algorithms in reinforcement learning (RL). Despite its widespread use, it has only been recently that researchers have begun to actively study its finite time behavior, including the finite time bound on mean squared error and sample complexity. On the empirical side, experience replay has been a key ingredient in the success of deep RL algorithms, but its theoretical effects on RL have yet to be fully understood. In this paper, we present a simple decomposition of the Markovian noise terms and provide finite-time error bounds for TD-learning with experience replay. Specifically, under the Markovian observation model, we demonstrate that for both the averaged iterate and final iterate cases, the error term induced by a constant step-size can be effectively controlled by the size of the replay buffer and the mini-batch sampled from the experience replay buffer.

We propose a novel hierarchical Bayesian model for learning with a large (possibly infinite) number of tasks/episodes, which suits well the few-shot meta learning problem. We consider episode-wise random variables to model episode-specific target generative processes, where these local random variables are governed by a higher-level global random variate. The global variable helps memorize the important information from historic episodes while controlling how much the model needs to be adapted to new episodes in a principled Bayesian manner. Within our model framework, the prediction on a novel episode/task can be seen as a Bayesian inference problem. However, a main obstacle in learning with a large/infinite number of local random variables in online nature, is that one is not allowed to store the posterior distribution of the current local random variable for frequent future updates, typical in conventional variational inference. We need to be able to treat each local variable as a one-time iterate in the optimization. We propose a Normal-Inverse-Wishart model, for which we show that this one-time iterate optimization becomes feasible due to the approximate closed-form solutions for the local posterior distributions. The resulting algorithm is more attractive than the MAML in that it is not required to maintain computational graphs for the whole gradient optimization steps per episode. Our approach is also different from existing Bayesian meta learning methods in that unlike dealing with a single random variable for the whole episodes, our approach has a hierarchical structure that allows one-time episodic optimization, desirable for principled Bayesian learning with many/infinite tasks. The code is available at \url{https://github.com/minyoungkim21/niwmeta}.

Intelligence is a difficult concept to define, despite many attempts at doing so. Rather than trying to settle on a single definition, this article introduces a broad perspective on what intelligence is, by laying out a cascade of definitions that induces both a nested hierarchy of three levels of intelligence and a wider-ranging space that is built around them and approximations to them. Within this intelligence space, regions are identified that correspond to both natural - most particularly, human - intelligence and artificial intelligence (AI), along with the crossover notion of humanlike intelligence. These definitions are then exploited in early explorations of four more advanced, and likely more controversial, topics: the singularity, generative AI, ethics, and intellectual property. Consider the space of all possible forms of intelligence, what can be called the intelligence space. We know very little about the nature of this space other than how it is currently populated with various, often only partially understood, instances of natural and artificial intelligence, although some earlier thoughts on this space - typically under different names - can be found in [2-5]. Rosenbloom [6] takes a different tack at understanding this space, in attempting to define a set of dichotomies whose cross product spans technologies underlying artificial and human intelligence, with the possibility of it spanning a much larger swath of the intelligence space. Here, another tack is taken, of starting with a generic, trilevel definition of intelligence that anchors the space and hypothesizing that a range of approximations to this definition could flesh out much of the full space (Section 1). This space is then exploited by exploring the relationship between natural and artificial intelligence via the mapping of human intelligence, humanlike intelligence, artificial intelligence, and cognitive science onto regions of this space (Section 2) and then beginning an exploration into the implications of this all for four more advanced, and likely more controversial topics: the singularity, as it relates to intelligence; ethics, with a particular focus on its relationship to artificial intelligence (AI); the current white-hot topic of generative AI, with a particular focus on large language models (LLMs); and intellectual property, with a particular focus on whether or not its creation might be ascribed to AI systems (Section 3).

Systems based on algorithms and machine learning are increasingly used to guide or outright make decisions which strongly impact human lives; thus it is imperative to take fairness into account when designing such systems. Notions of fairness in computer science can be classified into those that try to capture fairness towards a group (Hardt et al., 2016; Hébert-Johnson et al., 2018; Kearns et al., 2018; Kleinberg et al., 2017) vs. those that try to be fair to each individual Dwork et al. (2012); Kim et al. (2018, 2020). In our work, we focus on the latter notion. The most widely studied notion of individual fairness is due to the seminal work of Dwork et al. (2012): it assumes that a metric space on observable features of individuals captures similarity, and requires that outcomes of a resource allocation mechanism satisfy a certain Lipschitz continuity condition with respect to the given metric. Intuitively, this ensures that individuals who are similar according to the metric will be treated similarly by the mechanism. We consider a setting in which individuals have preferences over the outcomes of the resource allocation mechanism, focusing on the important setting of two-sided markets. Applications of this setting abound: matching students to schools, job fair participants to interviews, doctors to hospitals, patients to treatments, drivers to passengers in ride hailing, or advertisers to ad slots/users in online advertising (Abdulkadiroğlu and Sönmez, 2003; Bronfman et al., 2015; Mehta et al., 2013; Roth, 1986; Roth et al., 2007), to name a

Determining the optimal sample complexity of PAC learning in the realizable setting was a central open problem in learning theory for decades. Finally, the seminal work by Hanneke (2016) gave an algorithm with a provably optimal sample complexity. His algorithm is based on a careful and structured sub-sampling of the training data and then returning a majority vote among hypotheses trained on each of the sub-samples. While being a very exciting theoretical result, it has not had much impact in practice, in part due to inefficiency, since it constructs a polynomial number of sub-samples of the training data, each of linear size. In this work, we prove the surprising result that the practical and classic heuristic bagging (a.k.a. bootstrap aggregation), due to Breiman (1996), is in fact also an optimal PAC learner. Bagging pre-dates Hanneke's algorithm by twenty years and is taught in most undergraduate machine learning courses. Moreover, we show that it only requires a logarithmic number of sub-samples to reach optimality.

Action advising is a knowledge transfer technique for reinforcement learning based on the teacher-student paradigm. An expert teacher provides advice to a student during training in order to improve the student's sample efficiency and policy performance. Such advice is commonly given in the form of state-action pairs. However, it makes it difficult for the student to reason with and apply to novel states. We introduce Explainable Action Advising, in which the teacher provides action advice as well as associated explanations indicating why the action was chosen. This allows the student to self-reflect on what it has learned, enabling advice generalization and leading to improved sample efficiency and learning performance - even in environments where the teacher is sub-optimal. We empirically show that our framework is effective in both single-agent and multi-agent scenarios, yielding improved policy returns and convergence rates when compared to state-of-the-art methods

We propose a hybrid neural network (NN) and PDE approach for learning generalizable PDE dynamics from motion observations. Many NN approaches learn an end-to-end model that implicitly models both the governing PDE and constitutive models (or material models). Without explicit PDE knowledge, these approaches cannot guarantee physical correctness and have limited generalizability. We argue that the governing PDEs are often well-known and should be explicitly enforced rather than learned. Instead, constitutive models are particularly suitable for learning due to their data-fitting nature. To this end, we introduce a new framework termed "Neural Constitutive Laws" (NCLaw), which utilizes a network architecture that strictly guarantees standard constitutive priors, including rotation equivariance and undeformed state equilibrium. We embed this network inside a differentiable simulation and train the model by minimizing a loss function based on the difference between the simulation and the motion observation. We validate NCLaw on various large-deformation dynamical systems, ranging from solids to fluids. After training on a single motion trajectory, our method generalizes to new geometries, initial/boundary conditions, temporal ranges, and even multi-physics systems. On these extremely out-of-distribution generalization tasks, NCLaw is orders-of-magnitude more accurate than previous NN approaches. Real-world experiments demonstrate our method's ability to learn constitutive laws from videos.