2018 non-fiction, a layperson's introduction to the way in which
algorithms are allowed to affect life. (Another Book of the Week
condensation.)
This is a pop-science book on a subject about which I know quite
a lot. Therefore I'm mostly reviewing it not for people like me, but
rather for non-technical people who could use a fairly high-level view
of what's currently going on, what it might lead to, and what one
might want to do about it now and in the future.
Fry is distinctly more optimistic than I am about the willingness of
governments to attempt to regulate large companies that can promise to
get them re-elected; she talks of "the Facebook scandal" as though it
were something that would change anything about the way Facebook does
business.
An early attempt to classify algorithms into four broad types
flounders badly, but otherwise this is good stuff: a gentle
introduction to false-positive and false-negative errors, explanations
of why neural nets fail, and generally the way in which humans are
forced to accept algorithmic judgments (e.g. on no-fly lists or
mortgage applications) with no avenue of appeal even when the
"algorithm" was just someone's badly-written spreadsheet.
If you read my blog you probably aren't in the target audience for the
book, but it wouldn't be a terrible thing to give to a worried
relative. Or even a relative who isn't sufficiently worried.
Comments on this post are now closed. If you have particular grounds for adding a late comment, comment on a more recent post quoting the URL of this one.