Reynolds’ article reports briefly on a reinforcement learning algorithm with a twist, namely, the introduction of a system that rewards curiosity about its environment. The algorithm was let loose in a games environment and duly explored. As the article notes, DeepMind has also experimented in the field of curiosity-led algorithms. It brings to mind the primary emotion of Seeking described by Jaak Panksepp and Lucy Biven in The Archaeology of Mind.
Link to article: https://www.newscientist.com/article/2132916-curious-ai-learns-by-exploring-game-worlds-and-making-mistakes/