Project - Kyoto, baby

AI assisted learning for neural networks.

Core idea: 

https://playground.tensorflow.org/ clone with specific targets.

• Group A: Users get ai assistance for the first 5 examples (clipy style), then they have 5 test cases without AI.
• Group B: Users get 5 test examples they need to optimise for, with generic rules of how different parameters affect performance, then 5 test cases
  

We record (everything but we focus on):

• Performance on the 5 test questions (complexity vs performance) and speed
• Engagement between clicks
• If Michalis approves -> performance on the NN assignment (post intervention performance)