Pages

June 27, 2012

This robot will beat you at rock-paper-scissors 100 percent of the time



When someone claims they're good at rock-paper-scissors, they're usually just trying to psyche you out so they can predict your next move. But this robot, created by the Ishikawa Oku Laboratory in Japan, doesn't need to psyche you out, because it knows it will beat you every single time.

How can it do it? One thing it certainly doesn't do is any kind of high-level analysis of the game. It doesn't put your last sequence of moves through a complex semantic analysis and try to predict the move. It doesn't use anti-random tactics like "five scissors in a row" to throw you off. All it needs is a high-speed camera and quick reflexes.

Yes, the robot cheats. By watching the image from a camera that can determine the position of your hand every millisecond, it is aware of your move the very moment you make it. And as soon as your hand starts to form that rock, the robot is giving you some paper to wrap it up. At the very end of the video, you can see the tiny delay between the human making a move and the robot reacting — but it happens so fast that you wouldn't notice except when shown in slow-motion.

Janken


Doesn't seem fair, does it? But since humans can't see or react on a millisecond timescale, such a tactic would never even be considered. It's precisely because reaction time doesn't come into it that rock-paper-scissors becomes a game of psychology or dumb luck. Similarly, our vision isn't good enough to read a person's cards in the reflection of their eyes, and we're not skillful enough with our hands to control the outcome of dice.

Robots, on the other hand, could be fast enough, far-sighted enough, or precise enough that they could do any of those things. In this case, the robot can think and react in the time it takes you to make your play. Sure it's cheating — but it's still quite a trick.

The lab built the setup to show how quickly a robot arm and vision system could accommodate human input; that hand could just as easily be remotely operating in a hostile area or amplifying those movements to pick up construction materials.

Of course, what everyone wants to know is: What happens if they build a second one and have them play each other?

4 comments:

  1. "What happens if they build a second one and have them play each other?"

    Nothing.
    Each is dependent on the other moving first.

    ReplyDelete
    Replies
    1. I can build one that can fake the first move.....so what!!!

      Delete
  2. Where's John Henry (the steel drivin' man) now that we need him?
    Marilyn

    ReplyDelete
  3. By definition the robot is cheating, and the designer is just using technology to cover it's track.

    ReplyDelete