Ourapproach, Exploration through Learned Language Abstraction (ELLA) providesintermediate rewards to an agent for completing relevant low-level behaviors as it tries to solve a complex, sparse rewardtask.

To aid the study of generalization in a more realistic setting that is still tractable for current RL algorithms, we present Avalon2: a benchmark and high-performance simulator.

Reinforcement learning (RL) is a powerful tool for optimal control that has found great success in Atari games, the game of Go, robotic control, and building optimization. RL is also very brittle; agents often overfit to their training environment and fail to generalize to new settings. Unsupervised environment design (UED) has been proposed as a solution to this problem, in which the agent trains in environments that have been specially selected to help it learn. Previous UED algorithms focus on trying to train an RL agent that generalizes across a large distribution of environments. This is not necessarily desirable when we wish to prioritize performance in one environment over others. In this work, we will be examining the setting of robust RL building control, where we wish to train an RL agent that prioritizes performing well in normal weather while still being robust to extreme weather conditions. We demonstrate a novel UED algorithm, ActivePLR, that uses uncertainty-aware neural network architectures to generate new training environments at the limit of the RL agent's ability while being able to prioritize performance in a desired base environment. We show that ActivePLR is able to outperform state-of-the-art UED algorithms in minimizing energy usage while maximizing occupant comfort in the setting of building control.

The study of zero-shot generalisation (ZSG) in deep Reinforcement Learning (RL) aims to produce RL algorithms whose policies generalise well to novel unseen situations at deployment time, avoiding overfitting to their training environments. Tackling this is vital if we are to deploy reinforcement learning algorithms in real world scenarios, where the environment will be diverse, dynamic and unpredictable. This survey is an overview of this nascent field. We rely on a unifying formalism and terminology for discussing different ZSG problems, building upon previous works. We go on to categorise existing benchmarks for ZSG, as well as current methods for tackling these problems. Finally, we provide a critical discussion of the current state of the field, including recommendations for future work. Among other conclusions, we argue that taking a purely procedural content generation approach to benchmark design is not conducive to progress in ZSG, we suggest fast online adaptation and tackling RL-specific problems as some areas for future work on methods for ZSG, and we recommend building benchmarks in underexplored problem settings such as offline RL ZSG and reward-function variation.

Machines are getting smarter and smarter every year, but artificial intelligence is yet to live up to the hype that's been generated by some of the world's largest technology companies. Artificial Intelligence can excel at specific narrow tasks such as playing chess but it struggles to do more than one thing well. A seven-year-old has far broader intelligence than any of today's AI systems, for example. "AI algorithms are good at approaching individual tasks, or tasks that include a small degree of variability," Edward Grefenstette, a research scientist at Meta AI, formerly Facebook AI Research. "However, the real world encompasses the significant potential for change, a dynamic which we are bad at capturing within our training algorithms, yielding brittle intelligence," he added.

Machine Intelligence is increasing every year, but still Artificial Intelligence (AI) is about to live up to the reputation or hype that is created abut it by the technology companies in the world. AI still has a long way to go before behaving like a human intelligence but we will certainly see the evolution of AI in 2022. AI can excel at specific tasks but it struggles to do more than the specified tasks. It is better at doing programmed tasks than taking intelligent decisions or situational decisions. Edward Grefenstette, a Research Scientist at Meta AI said, "AI algorithms are good at approaching individual tasks, or tasks that include a small degree of variability. However, the real world encompasses significant potential for change, a dynamic which we are bad at capturing within our training algorithms, yielding brittle intelligence."

Machines are getting smarter and smarter every year, but artificial intelligence is yet to live up to the hype that's been generated by some of the world's largest technology companies. AI can excel at specific narrow tasks such as playing chess but it struggles to do more than one thing well. A seven-year-old has far broader intelligence than any of today's AI systems, for example. "AI algorithms are good at approaching individual tasks, or tasks that include a small degree of variability," Edward Grefenstette, a research scientist at Meta AI, formerly Facebook AI Research, told CNBC. "However, the real world encompasses significant potential for change, a dynamic which we are bad at capturing within our training algorithms, yielding brittle intelligence," he added.

Machine learning models have already mastered Chess, Go, Atari games and more, but in order for it to ascend to the next level, researchers at Facebook intend for AI to take on a different kind of game: the notoriously difficult and infinitely complex NetHack. "We wanted to construct what we think is the most accessible'grand challenge' with this game. It won't solve AI, but it will unlock pathways towards better AI," said Facebook AI Research's Edward Grefenstette. "Games are a good domain to find our assumptions about what makes machines intelligent and break them." You may not be familiar with NetHack, but it's one of the most influential games of all time.

Arnold Schwarzenegger means it when he says: "I'll be back," but not everyone is thrilled there's a new Terminator film out this week. In labs at the University of Cambridge, Facebook and Amazon, researchers fear Terminator: Dark Fate could mislead the public on the actual dangers of artificial intelligence (AI). AI pioneer Yoshua Bengio told BBC News he didn't like the Terminator films for several reasons. "They paint a picture which is really not coherent with the current understanding of how AI systems are built today and in the foreseeable future," says Prof Bengio, who is sometimes called one of the "godfathers of AI" for his work on deep learning in the 1990s and 2000s. "We are very far from super-intelligent AI systems and there may even be fundamental obstacles to get much beyond human intelligence."

Arnold Schwarzenegger means it when he says: "I'll be back," but not everyone is thrilled there's a new Terminator film out this week. In labs at the University of Cambridge, Facebook and Amazon, researchers fear Terminator: Dark Fate could mislead the public on the actual dangers of artificial intelligence (AI). AI pioneer Yoshua Bengio told BBC News he didn't like the Terminator films for several reasons. "They paint a picture which is really not coherent with the current understanding of how AI systems are built today and in the foreseeable future," says Prof Bengio, who is sometimes called one of the "godfathers of AI" for his work on deep learning in the 1990s and 2000s. "We are very far from super-intelligent AI systems and there may even be fundamental obstacles to get much beyond human intelligence."