Two ways AI could disrupt the economy via wages

I think it's absolutely core to human security that we view AI as extending from a core of humans and human society.  I will be posting further on that shortly I hope, but as pointed out two days ago at the CNAS Artificial Intelligence and Global Security Summit (Washington DC), one threat is civic disruption stemming from inequality.  And one part of ensuring that AI to extend individual human capacities, either directly as a prosthetic, or indirectly through education.

Unfortunately there are two ways that even using AI could still lead to economic and social disruption even under this model, and they aren't even entirely mutually exclusive.

1. If it is truly easy to train people with AI, or to augment untrained people so that they can perform any job, then this reduces human differentiation which has always been the basis of wage differentials. Wage differentials have been seen as key to maintaining distribution of money through an economy and reducing inequality. I'm not an expert on this theory.
2. It may be that not everyone can be trained or augmented with AI to the same extent.  This is a somewhat taboo topic in education, though it's considered blatantly obvious in athletics.  Some people will talk about randomly allocating political power, but no one ever talks about randomly choosing their nation's soccer team for the World Cup.  In the case that some people are better at working with and using the technology or others, wage differentials will remain, but depending on how they are distributed, this could lead to high wages being concentrated in small number of places, particularly if we don't maintain a good-sized population of workers for every task. But even if wage differentials are well distributed, it could be a very new distribution.

Can it make sense to pay high wages?

I said in the introduction "AI could," but I think it should be obvious that probably we are already witnessing both of the above effects already.  I was very struck by a book, Janesville, that documents the dismantling of a community's economy and the surge of inequality and political polarisation when General Motors closes a plant.  The union-protected wages in that community were a clear, countable cost of keeping the plant there and incentive not to build a new one.  However, the extent to which those wages correctly indexed the value of the community is unclear – it may have been foolish in
terms of quality reduction or loss of logistics support to move the plant, or it may have been worth the cost because it is now so much easier to build those things from scratch.  I don't know which, but given what I heard at the Aspen Institute earlier this year, I think Economics needs to do some hard work figuring that out.

This post was prompted by what I felt was an inadequate answer at a meeting I attended as a speaker last week at the OECD  AI: Intelligent Machines, Smart Policies.   I believe the talks will be on line at some point (they were webcast), this exchange came during the employment and education discussion.  And full disclosure: Janesville is where my mom was raised, and where my grandparents lived when I was growing up. My grandfather worked at the GM plant but only during WWII, when it made munitions.  He was originally a farmer but preferred being a salesman.