These new statistics suggest that the spread of AI will not just amount to “more of the same,” and that the onset of AI will introduce new riddles into speculation about the future of work.

Given their difference from previous analyses purporting to discuss AI, Michael Webb’s novel procedures demonstrate that we have a lot to learn about artificial intelligence, and that these are extremely early days in our inquiries. What’s coming may not resemble what we have been experiencing or expect to experience.

Webb’s machine learning statistics suggest AI could bring new patterns of impact across the labor market—ones fundamentally different from those brought by previous technologies.

It’s clear that past automation analyses—including our own, with its amalgamation of robotics, software, and artificial intelligence—have likely obscured AI’s distinctive impact. Based on expert familiarity, previous analyses have almost certainly been dominated by the ways robotics and software have been able to take over numerous routine, highly structured, and repetitive tasks.13

These analyses have tended to suggest that automation’s main effects will be to displace work across the middle of the skill and wage spectrum (such as factory workers and office clerks) while leaving the status quo more or less intact for both high-pay and low-pay interpersonal or nonroutine work (such as chemical engineers and home health aides, respectively).

However, the more refined empirical research presented here suggests that AI’s ability to employ statistics and learning to carry out nonroutine work means that these technologies are set to affect very different parts of the WHAT JOBS ARE AFFECTED BY AI? 23 workforce than previous automation. Most strikingly, it now looks as if whole new classes of well-paid, white-collar workers (who have been less touched by earlier waves of automation) will be the ones most affected by AI.

Eoin Treacy's view

Many better paying jobs rely less on expertise than on workplace protections. It is still mandatory to speak with an insurance agent in the USA when buying insurance. Many European countries dispensed with agents years ago. That single workplace rule which necessitates little more than a box ticking exercise with an agent, and the commissions that agent derives from the policy for every year it is active subsequently represent a massive cost to consumers.

In many cases the world is not waiting on an AI solution to displace workers but on rationalisation of bottlenecks in regulation. That is, in many respects, a much more difficult challenge because it necessitates that vested interests are overcome. That is true whether it is unions holding up atomisation, or teachers refusing to submit to quality control measures, the law society refusing to automate form filling, the medical profession insisting on office visits for prescription filling or banks charging for a host of automated services. These are all areas of life we interact with regularly which are protected by regulations and laws which are often immune to technological advances. They change eventually but progress is likely to be fastest in the private sector.