Bookmark and Share this Story:
  • Digg
  • StumbleUpon
  • Facebook
  • Fark
  • Google
  • Reddit
  • del.icio.us
  • Mixx
  • Slashdot
  • E-mail this story to a friend!

Great… Now Google’s AI Knows How To Parkour!

Posted by Hoob on July 11th, 2017
(Be the First to Rate This Post)
Loading ... Loading ...

What’s worse than rogue artificial intelligence? How about a rogue artificial intelligence that could win Ninja Warrior? You know this and I know this, but perhaps someone needs to tell the folks over at Google’s Deepmind project.

Our benevolent overlords just released a video of their DeepMind artificial intelligence system learning how to navigate a virtual parkour course through a process of reinforcement learning.

At its most basic level, the system was as follows: the faster the AI moved across the terrain, the greater the rewards. Additional incentives and penalties were added for more complex programs.

The AI used a trial and error system to figure out how to move forward as fast as possible without “terminating.”

Source: Geekologie

RSS feed | Trackback URI

Comments »

No comments yet.

You're not logged in. Please log in to comment. Not a member? Register and join YouBentMyWookie now. It's fun, free, and pure unadulterated awesomeness!


Trackback responses to this post