Microsoft trains AI agents for more realistic gaming

Claudiu Andone
by Claudiu Andone
Windows & Software Expert
0 Comments
Download PDF

  • When you are playing a game against a machine, you want to have tough opponents that can challenge you and that's where AI steps in.
  • A Microsoft research project is now aiming to improve AI agents in the opposite direction, to help you instead.
  • If you want to read more about software programs, go ahead and visit our Software section for more interesting articles.
  • We all like a nice past time with a game and if you're a gamer, we've got the Gaming Hub just for you. 

When it comes to gaming and AI, usually we are talking about better opponents, right? And in this field, Microsoft is even collaborating with their most fierce rivals.

You are always fighting waves of Artificial Intelligence driven agents that oppose you. They are getting increasingly smarter but still not a match for human predictability.

But Microsoft Research’s Project Padia is posing another interesting question: What if we taught AI agents to collaborate with human players instead?

Playing a game will teach AI agents about coop gaming

Project Paidia gameProject Padia is a fruitful collaboration between Microsoft Research Cambridge and Ninja Theory. They are using Ninja Theory’s Bleeding Edge Game to teach AI agents to collaborate with humans.

And the training platform goes back to the Azure Machine Learning technology and Microsoft’s Project Bonsai.

The game, which you can also play yourself on the project’s website, tests the collaboration between you (in white) and the AI agent (in blue).

You are locked in a room and you must work as a team to escape. Throughout the game, the agent will learn from a Reinforcement learning technique.

The final goal is to build deep learning models to help the artificial agents to build-up long-term memory, and to react quickly to new and unpredictable game situations.

What are the possible uses for AI training? 

This not only can lead to a better gaming experience but also produce better simulations. Lastly, AI Reinforcement learning can lead to more important developments in computing.

And let us remind you that not long ago, Microsoft has created an AI supercomputer in Azure, signaling a bright future for this huge field of development.

What do you think about Microsoft’s AI research program? Leave us your thoughts in the Comments section below.