This set of activities introduces the concepts underlying machine learning and provide hands-on experience creating your own machine learner (artificial intelligence). New AI models are built from existing image classification AI models using Teachable Machine. Students are guided to explore the space of training data and the associated results.

Students will learn the impressive abilities and the critical weaknesses of machine learning techniques.

This workshop was built with support from the Google exploreCSR program and TensorFlow.

Indigenous students exploring machine learning in 2022

alt

Students found that:

  • attempts to train the AI to recognise a face would only get an AI that recognised where that person sat
  • pens are too small for any AI to accurately find in this space
  • the inception model for imagenet has a lot of strange ideas
  • imagenet models are obsessed with colour

alt