Skip to main content

How to get better at video games, according to babies - Brian Christian

552,955 Views

8,114 Questions Answered

TEDEd Animation

Let’s Begin…

In 2013, a group of researchers wanted to create an AI system that could beat every Atari game. They developed a system called Deep Q Networks (DQN) and less than two years later, it was superhuman. But there was one notable exception. When playing Montezuma’s Revenge, DQN couldn’t score a single point. What was it that made this game so vexingly difficult for AI? Brian Christian investigates.

Additional Resources for you to Explore

To learn more about this topic, read Brian Christian’s book The Alignment Problem, which inspired this TED-Ed lesson. For a basic understanding of how artificial intelligence learns, watch this video. Watch this video and this video series called Deep RL Bootcamp for a deep dive into reinforcement learning. Another great resource for deep learning and reinforcement learning is this course taught by DeepMind's David Silver.

For a closer look at the AI research discussed in this lesson, see this paper for more about DQN and Atari, this paper for the use of intrinsic motivation to play Montezuma's Revenge, and this paper this paper for a discussion of how a tv screen can pose a distraction to curiosity-driven AI agents.

Next Section »

About TED-Ed Animations

TED-Ed Animations feature the words and ideas of educators brought to life by professional animators. Are you an educator or animator interested in creating a TED-Ed Animation? Nominate yourself here »

Meet The Creators

  • Educator Brian Christian
  • Director Gavin Edwards
  • Narrator Jack Cutmore-Scott
  • Music Salil Bhayani, cAMP Studio
  • Sound Designer cAMP Studio, Amanda P.H. Bennett
  • Director of Production Gerta Xhelo
  • Editorial Director Alex Rosenthal
  • Producer Bethany Cutmore-Scott
  • Editorial Producer Dan Kwartler

More from How Things Work