Machine Learning for Kids (using Skittles candy!)

My daughter was asking how things like Google can be so smart, so we spent 30 minutes doing a basic Machine Learning exercise using Skittles.

Before doing this, I explained to her how computers don't "know" anything, but can make completely random guesses. Some guesses may be terrible, but if you encourage good ones, then it gets better over time.

I then adapted my Machine Learning Using Starbursts article only this time, I used Skittles (it was all I had handy.)

The idea was to get the robot to travel a 10-step path that we determined, using positive reinforcement and then using both positive and negative reinforcement. The total exercise took less than 30 minutes - it went a lot faster once we started with the negative reinforcement.

For more details, check out the original article here:

Or, check out Kathy Ceceri's book "Bots! Robotics Engineering with Hands-On Makerspace Activities," which featured my Starbursts article along with a number of other great activities.