Plan for a 40 minute workshop given for people 10+ years of age. They:
- Play Super Mario in an emulator to learn how to play it (and perhaps to identify how they learn how to play it)/
- Use a Genetic Algortihm to train the computer to play. This first one is visual - they can watch the game played (and Mario die) over and over.
- Use a Genetic Algorithm that runs much faster (and no visual feedback) to see how far they can get.
- Kids have an intuitive feeling of how ML works
The computers will need some software installed for this to work. This release works only on Windows computers.
- Copy the zip file (that is part of the release) into the desktop, and unpack.
- Run the "1 - Play Super Mario" file. Make sure you can play (hit the return key to start, left and right arrows to move).
- If they keys do not work, then select the
Options
->Button Mapping
and then configure the return and left and right button. Even thought they look correct! - I would suggest testing the rest to make sure they worked as expected so there are no suprised during the tutorial.
Time (minutes) | Description |
---|---|
0-5 | Everyone get settled, use slide on screen to have them double click on the plain SM game to start |
5-10 | Have everyone play the game. Especially new people. Watch yourself as you learn. Slide on basics of hwo to play and the goal of the game. |
10-15 | Some discussion of how learning works. Repetition, talking to partners, watching others, trying & failing. |
15-20 | Intro machine learning - a simple slide talking about repetitive learning. |
20-25 | Have kids start the learning thing and watch it for a few minutes. |
25-30 | Slides on rain in Seattle (and football). Predict rain. Have them guess the next year's rain fall. Then show fitting. Talk about how it is a really simple math equation. |
30-35 | Show repetitive way to find the proper line - use participation. What if I threw up a line to fit it anywhere. Then ask them which direction I should move it. Etc. Etc. Make the point this is exactly what is happening on their screens. |
35-40 | Explain it takes a long time. Load up a ML thing that has trained all night. Watch how well it is doing. |
40-45 | Give some other examples up on the screen (like the handwriting??). |
45-50 | Questions, and then hang out afterwards and be willing to answer all sorts of questions. |
When | Where |
---|---|
July 28, 2019 | WOMAD Music Festival, 20 kids, 10 computer stations, a large screen up front to project on, and 4 people other than the presenter helping the kids out when something didn't work. |
- https://github.com/AdityaGupta1/waluig-IO
- FAQ for the emulator https://nintaco.com/faq.html
See the issues section of the github repo.