matk

AUC is robust to class imbalance. It can be understood as follows: given an instance of the positive class and an instance of the negative class, it measures the probability of your learned model being able to distinguish between the two.

I agree. These are very marginal computation costs.

This might have the most value as a Node transport though.

lol, yeah; ML seems to be an area where people like making new terms a lot.

So, the influx is due to a plethora of things, which are not all mutually exclusive:

1) The Internet has indeed produced large data sets which allowed statistical AI approaches to flourish opposed to logical AI approaches.

2) Somewhat better algorithms. However, I'd say that the algorithmic development hasn't had as much progress as the comments elude. We've only had a few notable innovations like NMT in the past ~20 years.

3) Computational Power to run the algorithms, so we can perform more experiments on large data sets, more computationally intensive algos, and induce better hypotheses.

4) Libraries like scikit-learn and keras have "democratized AI". Grad students used to implement algorithms themselves in the 2000s; now middle-schoolers are doing ML with the tooling already available.

Those are basically it. I think (2) could even be taken off, because again: learning algorithms have barely changed IMO.

I appreciate the sentiment of your comment, but, what part of a neural net isn't interpretable? Indeed, they do require more careful examination compared to traditional learning techniques. You can examine the receptive field of each node to infer what it detects for.