14 years ago, when I was master student, using machine learning(ML) algorithms for research were very useful but learning rate could be not satisfied mostly.
For my graduated thesis, I remember training phase took 2 days approximately.
But today, everything in ML/AI is easier than before.
Open source world growth quadratically and we could find any library __that we need__ in open source repositories. Without knowing the theory of neural computing, we could do anything in ML by using open source projects such as WEKA, H2O. In addition, deep learning libraries in Python are awesome today (GPU usage instead CPU helped to reputation of Phyton libraries).
I have too many Rube Goldberg projects which AI based, one of them my AI kit watches the kitchen of the Michelin Star Chefs and learns to cook :).
Why not? For Go play, It was worked.
https://deepmind.com/research/alphago/
A.I./ML in programming languages for memory management could be nice. But it comes to the source code analyzing/DF Topic actually everything seems rule based. So statistical methods are very helpful such as classification. Using classification algorithms and announce we are using ML etc comes to me funny and not smart.