CSAE Conference Day 1: The sceptics guide to innovation and AI
Apparently, sufficiently large gatherings of development economists can generate freak weather occurrences (n=2). Last year the CSAE conference was held in a freak patch of sunshine sandwiched by the Beast from the East, while today we’ve been alternating between glorious sunshine and vicious hailstorms. I’m no climatologist, but the evidence suggests that a sudden increase in the number of people saying ‘identification’ and ‘welfare impact’ in a concentrated geographical space opens some sort of weather variability vortex.
In that part of the day I wasn’t being pelted by hail, though, I spent a lot of time thinking about innovation and technology. The morning session I attended focused largely on innovation in African firms, while the evening plenary (during which I type, exiled to a spillover room by the popularity of the session) focused on the potential impact of AI and automation – specific innovations – on Africa’s future economic development, and there was some really smart comment throughout. But one thing that we didn’t spend that much time on is why we care about any of this at all.
I’m not being flippant, for a change. Why do we bang on about innovation, AI and the like, apart from an inherent fascination with the new? Ultimately, we care only in so far as it makes our lives better or worse – we care because of the welfare effects it brings. That sounds trite, but it actually helps frame our thinking in important ways when we’re thinking about how innovation matters.
The morning’s discussions focused a lot on innovation in the product or process space on its own terms, but ultimately we shouldn’t care about this at all. We care mainly about increases in welfare generated by more things being produced; things being produced more efficiently or more cheaply (and allowing our scarce resources to be conserved); or because it allows people to produce and exchange stuff more or more effectively and so providing them with a pathway to improving their lives. We care, not necessarily about innovation, but about the application and consequences of technology.
But there’s something about the word technology that seems to focus minds on the new, which is part of our obsession with shiny new toys and new ways of doing things. This seems unhelpful to me. Whether the thing that helps all of this along is new or old doesn’t matter. If what we need is more of an old technology like roads, or running water, or pickaxes, then that’s what we should care about; we only care about innovation if it either creates an easier way of producing or exchanging stuff (or services) or if it makes it harder for people to find any way of making, doing and exchanging as a pathway for improvements in their lives. In most sectors of most economies, adopting old ideas matters more than discovering new ones.
If a misplaced obsession with the new is at the root of an unhelpful fetishization of innovation, it’s a misplaced obsession with the old that’s behind an equally unhelpful fear of innovation that formed the focus of the evening plenary. If the welfare impact of adopting an old technology is the same as that of adopting a
new technology, we should be indifferent between them; so equally we should be indifferent between new ways of working and living and old ones, as long as the welfare implications are the same. If that’s the case, why are we so scared of the possibility that manufacturing might disappear? Or that certain jobs or types of human capital might? Do we think that the set of ways of working that exist now are all that we can do, the pinnacle of human achievement?
Doug Gollin made a similar point with a simple observation. Humans are quite good at doing stuff, and have proven good at doing a huge array of stuff in our recent history. The fear that we will run out of valuable stuff we can do seems to be driven by a linear extrapolation of what things robots can do holding what alternatives we can put humans to constant. But, as Doug pointed out, predicting what technology will do in the future is a notorious fools’ errand – we both underestimate and overestimate technological advances, so the shape of future production is extremely hard to predict. But more subtly, his comments suggested that predicting how people will respond to what technological change does happen is equally foolish.
That’s partly because our behavioural response is hard to predict. Noone knows how we will reallocate our time if we find that some tasks no longer need human intervention – it may be that we do more of other things that do require humans and so create more of some jobs that already exist. That’s what happened when the ATM machine came about – with the time spent physically counting cash and filling out forms saved, we spent more time asking for advice on our finances. Beyond that, it’s hard to predict our own price response. Will labour just become cheaper? If it does, how much cheaper can it become before it makes (some) machines obsolete? Or will it just switch to other areas where the price is higher?
Neither the panel, nor I, were saying that none of this matters – just that thinking about this stuff requires we think flexibly about human welfare, and all the ways open to improve it and to think less rigidly about what we do and how we do it. I think that fosters optimism, but then I’m an optimist.
On non-tech matters, the best thing about the conference is all the opportunity to catch up with amazing researchers doing things I wish I’d thought of; and having my priors challenged. Patrick Premand presented a paper on apprenticeships that went so against my priors I needed a rear-view mirror to see them. I’m sure there are loads of others on twitter – make sure to follow the #OxCSAE2019 hashtag.