Machine Learning Guitar Plugins-could have broader applications?

The possibilities of AI replacing composers and musicians in the future have been discussed. For now, this machine learning, neural network code can eliminate some poor vacuum tubes.

Official GuitarML site. You can train your own sound processor, using Tensorflow/Keras, one has to wonder what other acoustic effects can be created.

NeuralPi A Raspberry Pi port of Guitar ML. (VST3 built with JUCE)

The GuitarML Github page includes link to 11 articles by the developer, Keith Bloeme and a link to Proteus Tone Packs, where you can download 160 (.json ?) tones in a single zip.

This guy compares real amp sound to the modeled version ( at 9:57 after a tutorial)