Very misleading title about a very interesting development! A link to the actual paper is here: https://arxiv.org/abs/1802.01561 https://arxiv.org/abs/1802.0156

Very misleading title about a very interesting development! A link to the actual paper is here: https://arxiv.org/abs/1802.01561 https://arxiv.org/abs/1802.01561. In less clickbaity terms, DeepMind has developed a scalable system for distributed training of certain types of reinforcement learning agents. They use that to train an agent to solve a pretty big collection of different reinforcement learning problems.