The New York times has a piece on labor economist David Autor and his recent paper arguing that imitative AI will restore the middle class. It is an interesting argument, but it seems to reply far too much on technology and far too little on politics.
Autor, correctly in my view, believes that imitative AI is not anywhere close to being able to replace human judgement. Therefore, imitative AI can, at best, be an augmentation for human judgement. Technology generally takes two forms: automation (and thus replacement) or augmentation. Autor argues that because imitative AI is so crap at dealing with reality (he is nicer about it) that it cannot be trusted to be automation practically anywhere. He then suggests that imitative AI will augment human productivity, especially for those less experienced or trained. He has a handful of studies that suggest in certain areas, this kind of augmentation does seem to benefit less experienced, less highly trained employees. He then says that since that work can be spread out to more people, given the imitative AI’s help, the people who do that work with imitative AI should have their wages rise.
The “should” in that sentence is doing a ton of work.
Put aside for the moment whether you believe Autor is correct in his expectations. I do not — I think that imitative AI has its mistake-prone natured baked into it for a variety of reasons — but let us grant the argument. It does not mean, however, that it will lead to a more robust middle class. Once could argue that it will simply drive wages down across the board, as more people can do the work. As expertise becomes less important, workers have less leverage over their employers. It is easy, in fact, to see how this wealth is largely captured by capital in even Autor’s optimistic scenario. That is becasue income distribution is a political problem, not a technological one.
No technology guarantees workers will do better. In fact, the history of technology is largely one of the opposite — new technology driving people into poverty and wealth upwards. It is only after sustained political struggle that the benefits of technological progress begin to accrue even a little bit to workers. Politics, not technology, decide the winners and losers.
Autor, I am certain, knows this. He has been arguing for years that technology causes more pain that it has benefits, at least for working and middle-class people. But by framing his argument around the technological question— the possible benefits of imitative AI — instead of the political fight over how to split the imitative AI pie, he drives the conversation away from politics and towards the kind of lazy techno-optimism that only benefits the owners of these systems.
It is not until the ninth paragraph of the article that the possibility that Autor may be right about how IA will work but wrong about who it will benefit. And the article never really dives into the whys and hows of imitative AI’s putative benefits would be stolen from workers. Perhaps, if the paper had been focused on the politics, the discussion would have been different.
I have no doubt that Autor’s work is above board and meant to, as he puts it, add “value in imagining a positive outcome”. But the result, because of his focus, more attention on the technology and less attention on what he himself seems to understand really matters — the politics.