A number of people have commented on a new paper from Robert Gordon, professional pessimist. Many people have identified the specific issues with the logic in this essay. But I think it’s important to discuss a central problem with the very idea of Gordonesque gloom.
First, distinguish between positive and negative pessimism. If NASA were to tell us, with great confidence, that an astroid will strike Earth tomorrow there is no case for disdain. This is a scientific judgement and, at least philosophically, would be akin to standing in front of a speeding train and claiming that it won’t hit you. That is not the sort of pessimism that concerns us. But Gordon is making a much more powerful claim, a pessimism about what won’t happen – that our entrepreneurs cannot create another industrial revolution, that we’ve pretty much done the most we can with robots, and artificial intelligence is limited to the grocery store.
This requires a certain knowledge about the trend of technological progress and the economic value thereof. Gordon knows much more about the former than most and may well be the reigning expert on the latter. Unfortunately that doesn’t help us out. Because, believe it or not, most any of us can make this claim without any technical skill and little more than economic knowledge.
If you knew that some form of AI was going to revolutionize the world, and that building it is tractable, there’s a good chance it already exists or will be built relatively soon. Because that’s all you need to know to make a profit from basically nothing (patent the idea and rent the rights out when someone who can build it builds it). But it’s very rare that we make something from nothing so most of us, like Gordon, don’t see anything great about the future.
But that only means that me, you and Gordon don’t know what that invention will be, not that some arbitrary such invention won’t be. So Gordon is claiming that the space of all future invention is limited and the costs of finding the marginal source of technological growth are limited and, equivalently, that he knows the space of all future growth.
That’s a rather strong claim. He rather strongly asserts that the last three stagnant decades are a better indication than the last century. Choosing the postwar boom may be an outlier but, accordingly, so too would choosing a bad period like the past fifty years arbitrarily. By claiming that he understands growth will slow, Gordon implicitly declares that he understands the mechanics of future innovation.