Why it’s interesting
AI for Oceans turns training data into a visible little game. You label pictures as fish or not fish, then watch a machine-learning model apply those examples to a much larger ocean. The satisfying twist is that the model follows your data, including your mistakes, so careless or deliberately odd labels produce an equally odd cleanup result.
How to try it
Open the official level and click through the short introduction. Use the two classification buttons for each sample, finish a training round, and inspect what the model keeps or removes. For a sharper comparison, repeat the activity with a few intentionally inconsistent labels. Later levels let you choose a new word and train the model on randomly generated fish traits.
Requirements and sources
The browser activity is free and does not require an account for the core loop. A full classroom lesson is planned for about 45 minutes, while the basic interaction is understandable in a few minutes. Code.org's official lesson documents the mechanics and learning goals, and its 2019 annual report places the launch in December 2019. The cover is the current official curriculum image.