The ultimate goal of artificial intelligence research has always been to construct autonomous agents that can function in complex and higly dynamic environments. Such intelligent agents should be able to improve their performance over time learning from their experiences while exploring their environment. Reinforcement Learning (RL) is a discipline of AI with a focus on developing algorithms for such agents.
Over the years, RL has often come under fire for not being able to produce algorithms that can successfully solve other than the smaller problems. Unfortunately, in most cases, critics of RL happen to be people who are poorly informed and often misinformed about RL. As a result, the University of Michigan Reinforcement Learning Group has decided to maintain two very interesting wikis about myths, misstatements and successes of Reinforcement Learning. The first wiki answers many questions about the goals, algorithms and scalability of RL. The second wiki contains a list of applications that have successfully used RL. These include applications in robot control, human-computer interaction, computer games, economics, marketing and operations research. Both wikis are highly non-technical and contain a wealth of information.