Robot Learning

Can you humans picture a humanoid robot learning to bowl a perfect 300 game in less than an hour and
from then on only needing to walk down any new alley which was presented and characterize any defects
including inaccurate placement of the pins before proceeding to bowl perfect games.  At this point most
humans assume that not bowling a perfect score is due to variations in the human and inability to toss the
perfect ball each time.  Whereas this might be true, once humans have seen a robot bowl, they start to see
things differently.  

When a few human bowlers watched a robot bowl, it changed their view of the game.  The robot has the
ability to bowl the ball more precisely and uniformly, of course.  The robot perfected its game on alley 1 and
then surprised everyone by moving to alley 2.  After not getting a perfect score on alley 2, the robot
inspected the alley for being level and having dents in the wood as well as noting how precisely the pins
were being spotted.  He actually walked down the alley each frame and noted carefully the exact
positioning of each pin before tossing the ball.  On alley 3, he was able to bowl two perfect strings without
checking the position of the pins beyond the first three frames.  It was an interesting sight to watch.  He
ended up delivering perfect games on each of the 12 alleys before the morning was over.

The lesson is that once the bowler is almost perfect in execution, the variations in alley slope, dents, and
imprecise positioning of the pins become more apparent.  The human bowlers that saw the importance of
these variations will no doubt start to blame their not winning on these imperfections and demand more
precise equipment.  


Surprises for People
Robot Learning 03