Thanks for reading and appreciate the kind words Richard. I agree with all your points, especially on signal decay and the (lack of) novelty of ML in investing. Financial data has been mined for decades now, the ones who have applied machine learning to investing are often the same researchers who created the techniques in the first place, hired by hedge funds. It's naive to think investors have been oblivious to ML up until now.
On RenTech, I did a write up a couple of months ago which I think you may also find interesting.
Its largely structured as a response to a lot of commentary I've seen lately on RenTech being so successful because they use ML.
Short version: they're not unique in applying ML or using big data, their biggest edge is the veil of secrecy that they've managed to maintain over how they invest.