Thursday, April 18, 2019

Technology as a Substitute for (All Kinds of) Labor

Most economists don't expect automation to cause mass unemployment. Their argument is that, just as it has done for the last few hundred years, technology will replace some jobs but will create even more.

I find that conclusion counterintuitive and unsatisfying for lots of reasons, many of which I hope to share on this blog in the future. But in the mean time, here are two interesting items recently in the news.

First, Google's AI company DeepMind, after conquering the games of Go, Chess, Shogi, and StarCraft II, and in fact basically declaring victory in all two-player perfect information games, recently won a protein folding competition. If you still think that technology will only replace people in low-skilled jobs, think again.

This talk by DeepMind cofounder and CEO Demis Hassabis is fascinating. (Skip to 51:27 for the discussion about AlphaFold.)

Second, economists Daron Acemoglu and Pascual Restrepo just published a short paper titled The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand that seeks to answer the question, "If automation tends to reduce the labor share and has mixed effects on labor demand, why did the labor share remain roughly constant and productivity growth go hand-in-hand with commensurate wage growth over the last two centuries?"

I think it's important to recognize that very recent developments like those being pioneered by DeepMind are radically changing the types of work that can be automated. When machines can learn, they are not constrained by the imaginations or aptitudes of programmers, and can quickly surpass human ability in domains we previously considered reserved for us.

It's going to be a wild ride.

No comments: