Over the summer I wrote a short and relatively non-technical piece looking at what machine learning might have to offer to archives and special collections. This has just come out and can be found here
The International Council on Archives, which invited the piece, has also translated it into French, for those who prefer that language.
Recently, my friend Dan Cohen of Northeastern University has also looked at very similar topics in his wonderful newsletter Humane Ingenuity. See
I may do a longer version of this piece at some point in future, greatly expanding the discussion of facial recognition, which is quickly turning out to be an extremely interesting and complicated issue.