If large parts of society automate at the same time, it causes problems. In the end, once those problems have shaken out, society is normally better off for it, but the transition can take generations. The classic example of that problem is the Industrial Revolution. There is no doubting that it was better to be a factory worker in 1900 Britain than it was to be an agricultural labourer in 1750; but in the midpoint, the era of 18 hour days, Corn Laws, and the Peterloo Massacre, that clarity breaks down.
The two big problems that such a shift can bring are a concentration of wealth and skills mismatches which leave millions unemployable. The former comes as ever more of the returns to production accumulate to the owners of machinery (and in the modern reprisal, intellectual property), rather than the labourers; and the latter comes from the impossibility of rapidly retraining an entire population if their skills have been rendered obsolete.
But neither of those problems are resolvable through standard macroeconomic thought. The former isn't even seen as a problem at all by most economists, and the latter is seen as just a bigger example of the normal churn in the job market, ignoring the fact that a country where 2 per cent of the country is stuck with obsolete skills is very different from one where 20 per cent is.
If the robots do displace middle class jobs, then presumably the capitalist robot owners will have a lot of extra change lying around. The immediate impact is yet another surge in inequality. But presumably they’ll be looking around to spend their surplus on something, and that something might be the goods and services of an industry that will hire the newly jobless to produce them. This is traditionally how technological displacement goes. Reasons for pessimism notwithstanding, it can’t be entirely discounted that things will turn out this way again.
Anyways, just because we’ll have to wait a while to know anything for sure is no reason to ignore the anecdotal evidence, or for that matter to refrain from speculating about the potential consequences of a big economic transformation. Best to be prepared and so forth.
Something Cardiff misses, though, is that this revolution in automation isn't just affecting physical labour. Automation in the form of algorithmic creation has hit journalists  and lawyers , just as actual robots have hit doctors and researchers . That may seem like a technical distinction, but there's an important difference: the concentration of capital which is fairly inevitable with physical machinery isn't inevitable at all with software.
Consider two worlds, one in which every solicitor is fired to be replaced with Microsoft Word 2015 and its new "auto-write legal letter", and the other in which every solicitor is fired to be replaced with the open-source (and so free) Open Office 2015, with the same feature. In the former, almost all of the gains will go Microsoft, with a little bit more going to businesses which can afford the license taking custom from businesses which can't; in the latter, where the importance of having capital to pay for the software license is diminished, the concentration may not be quite so big. Either way it's not great for solicitors, but if the savings were passed on to customers rather than recouped by Microsoft, that's probably the better outcome.
All of which is to say that if the rise of the robots continues, reassessing our intellectual property regime may be important not just because it could boost innovation , but because it could be the only way to deal with the new world.