Slightly scary AI

On the whole I haven’t been among the most pessimistic people about the likely impact of a potential Artificial Intelligence revolution on the economy and life – although not blithely optimistic either about the scale of the adjustment that will be needed in labour markets and education. But

by Jerry Kaplan has made me more apprehensive. This is not because of anything he writes about the economy, which is standard fare, but rather because of something he says about AI in the first half of the book (quite a short and very readable volume).

[amazon_image id=”0300213557″ link=”true” target=”_blank” size=”medium” ]Humans Need Not Apply: A Guide to Wealth and Work in the Age of Artificial Intelligence[/amazon_image]

It starts with an account of high frequency trading and the Flash Crash of 2010. This will be familiar to anybody who has read Michael Lewis’s very entertaining

. It seems pretty clear to me that financial regulators need to rein in HFT – not that they are showing any sign of interest in doing so – and there is even a proposed solution recommended in Al Roth’s terrific book about market structure,
, as well as in this volume. That is to regulate to allow computer trading to occur only once every second. This would re-create liquidity in the markets (which are illiquid at the milli- or nano-second timescale) and stop the speed arms race.

However, what Kaplan points out is that other AI applications will have undesirable consequences because they will look like the swarms of computers trading in financial markets, and doing it super-well with no application of judgment. One example is the use of AI to personalize the offers made to shoppers online, which will become so efficient that the synthetic intelligence will be able to price discriminate perfectly, extract all the consumer surplus in each market, and undo the hope that online retailing would lead to less rather than more price dispersion. Nobody will be forced to buy, of course. “But while you may exercise freedom as an individual, collectively we will not. Synthetic intellects are fully capable of managing the behaviour of groups to a fine statistical precision while permitting individuals to roam in whatever direction their predictable little hearts desire.”

The book asks many other thought-provoking questions about social norms and ethics. Will my personal robot be allowed to stand in a queue for me? Or repark my car to avoid tickets all day? How can we ensure robots understand what are the moral boundaries on their actions when they’ve been programmed to fulfil a certain kind of task?

It also has one interesting policy suggestion, the “job mortgage”, a means of allowing people to train and retrain without being tied to one specific employer: the government loans people money and is repaid from their subsequent earnings. This would replace the existing student loan schemes and apprenticeships, which Kaplan sees as too restrictive for the period of upheaval ahead.

There seem to be lots of books on this subject out now, but I thought this one worth a read, especially for an economist like me, as the author’s background is a technical one and he explains the technology trends very clearly.


3 thoughts on “Slightly scary AI

  1. Pingback: Slightly scary AI | Homines Economici

  2. Diane, does Roth’s book explain why the kindle edition is priced at paperback price + vat (when, I assume, the marginal cost of each new kindle edition is tiny) and who is getting the additional proflit?

Comments are closed.