Thanks, it is a topic I've dabbled in as a curiosity. (I've replicated the Iris dataset experiment, digit recognition, and some experiments with 'particle swarms'), and in light of the other day announcement of a 'sentient AI' from Google (Even if it's limited to text based interactions).

There came a point where I realized that AI is stupid, it just reaches stupid a lot faster than any human could.... For example, with one AI they came to realize that as far as the AI was concerned, would you believe that the difference between a wolf and a husky is if there's snow in the background? (snow = wolf)

So, as soon as I saw "deep learning" that means they created a dataset where they tagged stuff as "russian disinformation" whether or not it was in reality. So, it, with mathematical objectivity, told them all the stuff that fit what they classified as "russian disinformation" (and the other categories).