Google's DeepMind has recently learned how to complete tricky video game Montezuma's Revenge, famously beat Go world champion Lee Sedol 4-1 and can play a raft of Atari games.
It may sound like an odd choice, but Google said the research could help work out "continuous control problems such as robotic manipulation and locomotion."
The virtual ant can move around its surroundings, kick a ball and is able to successfully aim it towards a 'goal'.
The experiment is part of a wider project from DeepMind of "deep learning of neural networks".
"At DeepMind we have pioneered a combination of approaches to create the first artificial agents to achieve human-level performance across many challenging domains," the firm wrote in a blog post.
"From Atari to Labyrinth, from locomotion through manipulation, to poker and even the game of Go, our deep reinforcement learning agents have demonstrated remarkable progress on a wide variety of challenging tasks."