The historian Suetonius wrote that, when approached by a man offering a cheap technology for transporting heavy columns to Rome the emperor Vespasian refused to use the device for fear it would displace workers. "How will it be possible for me to feed the populace?" he asked. The fear that new technology will destroy jobs is, therefore, an ancient one.

This is the theme of Carl Benedikt Frey's The Technology Trap. Although Dr Frey is best known for his work with Michael Osborne predicting (pdf) that new technologies threaten millions of jobs, his book is not an exercise in futurology. Most of it is a fascinating history of technical change.

There are, he says - echoing Acemoglu and Restrepo - two main types of technical change. Some enable workers to do more, making them more productive and creating new work. Others, though, simply replace labour. Frey argues that much of pre-industrial technical change was labour-replacing, and was often suppressed by states fearful of the unrest that would cause: Elizabeth I refused a patent for William Lee's knitting machine in 1589 on these grounds. And, he argues, the early decades of the industrial revolution was also an era of labour-replacing techniques - hence the "Engels pause", years of stagnant wages. A big reason for the industrial revolution, Frey shows, is that the British state shifted from supressing technical change to using coercion to enable such change: many Luddites were hanged. Frey nicely rebuts the silly just-so stories of some right-libertarians who tend to see capitalism as natural.

It was only later in the industrial revolution that technical change actually created jobs, and goodish ones even for unskilled work. In the long sweep of history, however, this sort of change - and the improved living conditions for the masses it brought - was unusual. This fits with Walter Scheidel's work from a very different perspective, which has argued that periods of increased inequality are historically very rare.

From this perspective, the labour-replacing changes that Frey fears will come from AI and robotics are in a sense a reversion to the historic norm. As futurology one might quibble with that; people are bad at predicting technical change. But turn the question around: how likely is it that new technologies will provide good jobs for unskilled men of the sort we saw in the first half of the 20^th century?

So, where might we argue with Frey's analysis? I agree with John Thornhill that his policy suggestions (around helping workers adapt (via education, changes in zoning laws and suchlike) are weak, but this is common in many books analysing our new times.

Instead, I suspect Diane Coyle is right to argue that Frey treats technical change as exogenous when in fact it isn't. For example, the distinction between labour-replacing and labour-enabling technical change, whilst insightful, distracts us from another type - the sort that enables capitalist exploitation such as the power-biased technical change discussed by Skott and Guy. We should ask: if we have greater worker ownership, what sort of technical change would we see? Mightn't it be more labour-enabling?

In not seeing such change an endogenous, Frey under-estimates our current paradox - that (to paraphrase Solow) robots are everywhere except in the data. The key facts of recent years are that unemployment is at a 45-year low; that business investment has flatlined for years; and that productivity is stagnating. The data suggest that people are stealing robots' jobs, not vice versa.

Frey is right to say that such stagnation is common in the early years of new technologies. But I suspect this isn't the only explanation. It's also the case that there are big barriers to robotization such as: firms' fears that new technologies will be out-competed by future cheaper tech; that firms with intangible assets struggle to raise finance for lack of collateral; that low wages and a quiescent workforce mean there's little need to invest in labour-saving techniques; or that the tech crash and financial crisis have near-permanently depressed animal spirits.

Frey's work, however, poses two deep issues which we should all think about. One is: are sensible economists right to want supply-side policies that boost productivity? If we get these, mightn't we open the Pandora's box that Frey fears, of significant labour-displacing investment? Is good macro policy aimed at maintaining full employment really sufficient safeguard here?

The other is that technology changes culture. The jobs at threat from AI are, mostly, those that require us to follow algorithms: robots do that better than us. Humans' comparative advantage lies instead in creativity and soft skills. Should we do more to foster these, by changing educational priorities? How would culture, society and humans change if much of what we now think of as skilled professional work is instead done by machines?