"If new tech was breaking the frontiers of automation and making it possible to automate jobs which couldn't be done before, these features should've changed."
This is definitely not self-evident, moreover, I'd totally disagree with it.
The authors are essentially missing the point by trying to categorize jobs or features of jobs as 'automatable' when in the grand scheme, it's just not how it works.
Everyone's jobs are evolving due to automation.
1) Automation has been steadily happening due to progress in IT for decades.
2) AI, in particular, is a nascent field, moreover, it's really on a 'tool' in the tool chest of IT systems overall. The time horizon for impact for direct impact of 'AI' will be a long way off, and when it is, it won't be directly apparent in most cases. There will be very few jobs where we'll be able to say 'AI replaced me'. Maybe Taxi drivers.
3) And my point - is that automation is tremendously difficult to measure in reality, and that it changes the nature of some jobs more than others, and in ways that cannot be measured by using employment numbers.
The 'AI replaced the taxi drivers' example, derived from our 'Industrial Revolution' thinking about automation and productivity, is an outlier.
Seeking to find ways in which 'AI replaced cab drivers' is probably worthwhile on some level, but it's going to miss 95% of the revolution.
Think of an Amazon product picker - their jobs have changed tremendously due to automation. There may be all sorts of flavours of AI baked into an Amazon warehouse pickers job now and in the future: a lot of AI in the vision/3D control for the robots which incrementally replace humans. AI in the route optimisation, AI in labour scheduling (no longer done by humans), AI in facilities management (ordering parts, service automatically).
All of this may lead to considerably more efficient workplace and yet not a single job may be lost nor might it lead us to think of some magic tech replacing some workers in a specific way, rather 'the system got smarter'.
The economic equilibrium may in fact lead to considerably more people being hired to do 'different kinds of more automated picking', not fewer.
The AI revolution is the same as the information revolution and it's happening across the system in myriad ways that will be very hard to measure as we would like.