Teach. Test. Break. Build.
Can you break this AI?
First, you’ll have to teach it.
About 3 minutes · No coding · No account · Camera optional
What you’ll do
Pick a future you care about. Teach a small machine-learning model with your own examples. Test it. Then try to break it — and when it fails, work out why, give it a better example, and test it again. In about three minutes you will have done the real work of building AI: choosing the evidence, questioning a confident answer, and deciding when to trust it.
Two ways to play
Live Vision Lab — at the Open House table, with a camera pointed down at two piles of cards. Teach it, then turn a card, cover it, change the background, and watch what happens.
Accessible Data Lab — on any phone or computer, no camera needed. Train on readings such as temperature and vibration, then take the model somewhere your examples never went. Same lesson, same three minutes.
Privacy
Camera privacy: The camera is used only for this activity. Images are processed in this browser and are not saved or uploaded.
Future Forge asks for no name, email or account, and keeps no record of your visit. Everything runs inside your own browser; nothing is sent anywhere.
Accessibility
Every step works from the keyboard and with a screen reader. Nothing requires a camera, a mouse, sound or a quick reaction. Charts have text equivalents, and motion is reduced when your device asks for it. Future Forge is designed toward WCAG 2.2 AA. If something does not work for you, write to ai.institute@clarkson.edu.
Build the Future with AI at Clarkson.
TRAIL: Future Forge was designed by Stephen Casper for the Clarkson AI Institute and built with Claude (Anthropic) as engineering collaborator. The model you train runs entirely in your browser.