Writer David Weinberger and software engineer Yannick Assogba from Google have joined forces to produce a podcast mini-series about playing tic-tac-toe (and tic-tac-two) with machine learning models.
They keep it at a comfortable intermediate level, avoiding vague analogies as well as coding details, and focus on the many human decisions that need to be made in developing such an ML system.
I enjoyed the episodes very much, not only because I learned new things, but especially because the podcast helps to clean up the diffuse image of “magic AI”.
Episodes (~20-30 min each)
- Introducing Tic-Tac-Toe the Hard Way
- Howdy, and the myth of “pouring in data”
- What does a Tic-Tac-Toe board look like to machine learning?
- From Tic-Tac-Toe moves to ML model
- Beating random: What it means to have trained a model
- Give that model a treat! : Reinforcement learning explained
- Head to Head: the Big ML Smackdown!
- Enter Tic-Tac-Two
- Head to Head: The Even Bigger ML Smackdown!
- Lessons learned
On the podcast’s website there are, among other information, links to a web viewer that visualizes how the models trained by Yannick and David play against each other.