Matt's Thoughts In Between - Issue #23

Facebook earier this year released DensePose, a machine learning framework that allows the creation of 3d models of people from 2d images. IEEE Spectrum has a good piece on some of the darker potential use cases. Drawing on the indispensable Jack Clark, they note that DensePose could be used to enable real-time surveillance of large groups of people (and as Clark notes, the researchers behind it don’t discuss this or other potential malicious uses, which is troubling).

This topic is everywhere. The New York Times had an excellent article this week on China’s use of AI for surveillance. While one of its conclusions is that the technology isn’t actually quite ready yet, it’s also striking how much behaviour is driven simply by the belief that it is (and in a week where a former senior UK police officer launches a startup literally called Panopticon)

It’s worth musing on this in the context of a recent Tyler Cowen column in which he argues that a key development of the last 20 years is that “the governance technologies and strategies of authoritarian regimes have become much more efficient” (See also this fascinating paper on “Informational Autocrats”). It’s worrying, but it’s not clear to me how even democracies can escape a “race to the bottom” in surveillance.

I wrote last week about the political philosophy of Peter Thiel, and a couple of readers sent me thoughts on Ryan Holiday’s recent book Conspiracy, which is largely about Thiel. It looks at the extraordinary story of how Thiel secretly funded Hulk Hogan (yes, the wrestler) to pursue a lawsuit against Gawker, which ultimately led to the publication’s demise. It’s an amazing story and if you don’t have time for the book, the EconTalk podcast with Holiday is, as ever, superb.

Among many other things, the book raises a classic Thiel contrarian question: do we have enough conspiracies? (For others in this genre, see this interview in the NYT in which he muses on whether there’s now too little corruption…) The premise is that many powerful people hated Gawker, but only Thiel’s meticulous, patient and clandestine plot had any impact. Ridding the world of a gossip blog may not be an important achievement, but the question of what other problems might need a “conspiracy” to resolve them is both thought provoking and troubling. (The EconTalk podcast explores this in more detail, including an interesting discussion of the assassination of Julius Caesar)

Maybe the answer is that this is already happening. Tyler Cowen (again) has an interesting take on the most underrated conspiracy theories (briefly discussed on TiB before). Perhaps it’s time to take conspiracies more seriously.

I’m often struck by how powerful certain individual statistics are - governments rise and fall on unemployment numbers, corporations on their earnings reports. But Stephen Gandel has an interesting column in Bloomberg this week on just how malleable these are, at least in a corporate context (This is not to let governments off - there have been a flurry of books in recent years on the extraordinary power and manifold problems of GDP, among other statistics).

The basic idea is that it’s increasingly common for large companies to make a wide range of non-standard adjustments to their earnings when presenting forecasts - adjustments that, unsurprisingly, usually have the effect of improving the story. (The famous recent example in tech is WeWork’s “community adjusted EBITDA” - or as Twitter wits put it, “profit before costs”)

Of course, this matters for shareholders in the relevant companies - but perhaps it matters for all of us. A recent conversation between Noah Smith and Conor Sen makes the compelling case that corporate debt defaults are the most likely cause of the next recession - and that’s far more likely if investors can’t trust the numbers they see.

Thanks as ever for reading. If you like this, you should forward to a friend or share the subscribe link on social media. If you have feedback or would like to share a link, just hit reply.

Until next week,

Matt