https://en.wikipedia.org/wiki/Tsetlin_machine
They are discrete, individually interpretable, and can be configured into complicated architectures.
These guys are trying to make chips for ML using Tsetlin machines...