Google DeepMind’s terrifying AI learned how to walk on its own

Google’s AI branch built a new artificial intelligence capable of learning how to walk, run, jump, and climb, without receiving specific instructions. DeepMind, the company whose AI beat the world’s greatest Go player, wrote in a blog post how it plans to create artificial intelligence with enough physical smarts to master sophisticated motors skills. The company created several strange AI body types and tested them on diverse digital terrains, or obstacle courses, to see how well they could adapt in different environments. DeepMind posted a video of the results, which are as hilarious as they are terrifying. Each of the figures was programmed with sensors that told it how its body was positioned. A set of simple reward signals was used to encourage the agents to get through the course. The rest is the AI learning how it can continue moving forward through the obstacles without being terminated. For now, AI is still in innocent-baby phase, but DeepMind says its ultimate goal is to “produce flexible and natural behaviors that can be reused and adapted to solve tasks.” The advances it makes can then be programmed into future robots to make them seem more human. You’ll be relieved to know that…


Link to Full Article: Google DeepMind’s terrifying AI learned how to walk on its own

Pin It on Pinterest

Share This

Join Our Newsletter

Sign up to our mailing list to receive the latest news and updates about homeAI.info and the Informed.AI Network of AI related websites which includes Events.AI, Neurons.AI, Awards.AI, and Vocation.AI

You have Successfully Subscribed!