TiB 99: facial recognition and the race to the bottom; what (doesn't) cause polarisation; the long run effect of the Cultural Revolution; and more...

Forwarded this email? Subscribe here. Enjoy this email? Forward it to a friend.

The New York Times, under the headline, “The Secretive Company That Might End Privacy as We Know It”, profiled Clearview AI, a startup that can apparently identify a person from any photo. Clearview has scraped billions of images from the public web and trained a model to find a match from a photo. There are useful applications in security - Clearview’s initial market - but understandably lots of people are worried about the implications. 

As usual, Ben Thompson has the best analysis. I’d highlight two points. First, the fact that any individual can do at micro-scale what Clearview does at global scale is a difference of kind, not just magnitude (“Quantity has a quality all of its own”, one might say?) We’re a long way from knowing how to deal with this societally. See Benedict Evans “million interns” analogy on machine learning capabilities for more on this - or Albert Wenger’s ”right to be represented by a bot“ for an optimistic take.

Second, what Clearview does has been technically possible for Facebook,et al, for a long time, but norms held them back: ”Clearview AI, though, ignored norms, and found a market”. I expect we’ll see this increasingly: in a world with fewer gatekeepers, anything that can be done will be done. That might worry us when it comes to privacy destruction, but it should worry us a lot when it comes to, for example, cheaper bioterrorism. Countering this dynamic seems one of the most important problems for smart people to work on…

Political polarisation is a hot topic on both sides of the Atlantic. It’s used as a lens for looking at everything from social media regulation to the future of democracy, so it’s important to understand. I came across three new-ish papers that cast helpful light on aspects of this - on the basic facts of polarisation, its impact in the economy and how it’s changing how we think.

First, this paper looks at a comparative history of polarisation in the West since 1980. It shows that while polarisation has increased markedly in the US, that’s not true everywhere (it’s actually fallen in Britain). This is important for understanding causes: it casts doubt on narratives that blame global phenomena like the rise of social media or inequality (though see an interesting counterpoint on inequality here). The authors suggest that partisan cable news and worsening racial tensions are better explanations for the US position.

I’m not sure I fully buy the second paper, which argues that shared partisan preferences in the workforce are a predictor of both more and more successful mergers and acquisitions. The effect size is modest and there are lots of confounding factors. But it’s something that could become increasingly pronounced as polarisation provides a positive feedback loop for more corporate activism (previous coverage). Finally, this short paper argues that polarisation increasingly shapes (US) people’s perceptions of reality(!) That seems like very bad news, whatever your politics.

A favourite occasional TiB topic is historical forces that end up having very long run effects. Tanner Greer pointed to an interesting example last week: the impact of political violence during the Cultural Revolution on attitudes and behaviours in China fifty years later. This new paper by Yuhua Wang shows that people in areas that suffered greater violence during the Cultural Revolution - even those who weren’t born at the time - are both less trusting of and more obedient to the state today.

It’s interesting to speculate on just how enduring effects like these are. One of my favourite papers of last year - discussed here - was Joe Henrich et al’s piece that argued that Catholic Church family policy shaped Western psychology over periods of centuries. Similarly, Tanner Greer, in his commentary on Wang’s paper, points to this study that suggests that political repression in 17th and 18th century China appears to have permanently damaged social capital in the affected areas. 

The impact of the Cultural Revolution may be more shorter lived - Wang shows that the effect diminishes over generations - but he speculates that there’s a dilemma for authoritarian governments: you can buy obedience at the cost of stoking long run resentment, which can come back to haunt you. I worry this is too optimistic a take. Perhaps “low trust, high obedience” is actually the norm in history; revolutions is much rarer than repression, alas. 

Thanks as always for reading - and welcome new readers! If you enjoy TiB, please forward it to a friend or two. And it’s always great to hear from you: feel free to get in touch on Twitter or by hitting reply.

Until next week,

Matt