Microsoft's Project Malmo AI platform goes open source

The system, now available to all, uses Minecraft to test artificial intelligence protocols.
Written by Charlie Osborne, Contributing Writer
Scott Eklund/Red Box Pictures

Microsoft has released artificial intelligence system Project Malmo to the open-source community.

On Thursday, the Redmond giant revealed the shift of Project Malmo from the hands of a small group of computer scientists in a private preview to GitHub, a code repository for open-source projects.

Project Malmo is an artificial intelligence system used to test out AI algorithms using the Minecraft world.

Formerly referred to as Project AIX, the platform has been developed in order to give startups a cheap, effective way to test out artificial intelligence programming without the need to build robots to test commands and comprehension with physical subjects.

See also: The world of Minecraft is now used for artificial intelligence innovation

The platform's primary goal is to boost research in general AI, including teaching agents how to learn, hold conversations, make decisions and complete complex tasks. According to Katja Hofmann, a researcher in Microsoft's Cambridge, UK research lab, Project Malmo can also be used in "reinforcement learning," in which AI agents learn to complete a task by trial and error -- and are rewarded when they make the right choice.

"We're trying to put out the tools that will allow people to make progress on those really, really hard research questions," Hofmann said.

Microsoft has also added new functionality to the platform on launch which will allow developers to create bots which can learn to talk to each other -- as well as people. Researchers will also be able to overclock Minecraft to increase the speed of experiments beyond the game's usual pace.

Project Malmo consists of a Java mod and code to help AI agents act within the world of Minecraft, and is compatible with Windows, Linux or Mac OS.

You can check out the project here.

10 steps to learn how to hack

Editorial standards