Vision Recognition AI for Zookeepers is the first in a series of workshops designed to give kids an overview of Artificial Intelligence (AI) and Machine Learning (ML) technologies and societal implications.
This activity starts several offline visual puzzle activities to help kids “think like an AI”. They learn how an AI perceives images and how that perception differs from ours. They also learn how AIs need a lot of data to learn to recognize different classes of images. Next, the kids put their knowledge to work to train an AI vision camera system to work. We tell the kids that they are zookeepers, and we brainstorm ways to use the cameras to improve their jobs, the lives of the animals, and the experience for visitors at the zoo. The kids use laser-cut zoo pieces and plastic zoo animals and visitors to set up “Normal” and “Problem” scenarios for the cameras to recognize. They train the AI vision system to recognize these scenarios and generalize to new situations by training it on as many examples as possible. For example, for an animal escape problem scenario, they train it on as many different escapes as possible so that it can recognize any of them.
What should we do with the extra time that the zookeepers now have because of the automation? Hire fewer zookeepers and lower the prices for the zoo? Take care of more animals? Talk to visitors and educate them about animals?
Students are given an introduction to basic principles of machine learning and how computers might recognize images and objects, similar to how someone would put a puzzle together.
Kids put their knowledge to work to train an AI vision camera system to recognize their faces and expressions.
This is the laser cut box that houses the AI vision camera system.
Students have a hands-on experience of training the computer and collecting data through taking photos of their animals.