DeepMind is Google’s new Artificial Intellgence Development team. The team has recently been training its AI to adapt some apparently strange walking techniques in virtual space. The new technique which is under development process is known as the reinforcement learning. The reinforcement learning technique revolves round the objective of incentivizing the Artificial Intelligence. This means the AI will aim an objective and apply a set of sensors and abilities in what is described as a ‘rich environment’.
DeepMind teaches its AI to adapt with linear as well as non-linear terrains. The strange walking techniques which apparently seems funny are the complex behaviours that enables the AI to survive non-linear bumpy terrains and head towards its destination. DeepMind AI thus learned how to use its various virtual limbs in an environment that had many options.
Three body variants have been incorporated in the AI for learning the movement techniques. The first Planar is a two-legged, armless, and headless creature that learned to crouch, jump, run, and duck through a military-style obstacle and hurdle course. The Planar walker in course of its journey was able to learn to use its “knee” to gain extra purchase over one of the bigger hurdles.
Next comes the Quadroped; a small, four-legged ball that quickly gained knowledge how to hop and balance from plank to plank, over large drops, as well as navigate around or over large blocks. The stable and flawless movement is a treat to watch.