What do you like best?
The fact that here lies an engine that mimics the brain in terms of how we dream, or more precisely, how we hallucinate, is in fact, cool. It gives us an insight on how a Convolutional Neural Network works, by giving a vague visualization of its layer to layer activations. And the source code, it's in fact short ( and sweet ! )
What do you dislike?
Well, since this engine relies on a pre-trained Convolutional Neural Network, that was trained on ImageNet data, and boosts any activations at the layer we are "dreaming", it usually ends up activating dogs more often than anything else. It would have been much more cooler if the network was exposed to more objects and training data.
Recommendations to others considering the product
Train the "Inception" network on more data, with more classes, and compare the results of the "trippy" activations.
What business problems are you solving with the product? What benefits have you realized?
I have been trying to realize the working of the human brain since quite a while, and never quite bumped into the idea of hallucinating !! If the engine is tweaked further with more real world training data and "classes", I guess it would be possible to generate more "data" with help of these dreams.