Thoughts in Between

by Matt Clifford

Matt's Thoughts In Between - Issue #73

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

AI conquers poker

For the first time, an AI can dominate humans in multi-player Texas hold-‘em, the most popular format of poker.  Given the need to bluff and deal with hidden information, poker has usually been seen as a major AI challenge. AIs have been able to beat humans in two player poker for a while, but multi-player is a much more complex challenge, as there is no simple game theoretic “solution”. 

The AI, built by researchers at Facebook and Carnegie Mellon, played against itself many times, in a similar way to AlphaZero, to create a “blueprint” map of strategies. It then uses the blueprint as a baseline and real-time searches for an improved strategy as the game situation develops. There’s a good overview of the methodology here

As I’ve discussed before, when AI surpasses human capabilities in a field, it’s fascinating to hear the expert human response to encountering a completely different kind of mind. There are some good examples in Facebook's announcement ("It's a monster bluffer"). It’s also striking that this AI is many times less computationally intensive than AlphaGo - $144 to train the blueprint (vs $26m to train AlphaStar, though not an apples-to-apples comparison) and uses two CPUs in game vs 1,920 for AlphaGo! It’s another important AI milestone and a glimpse of lots more to come. 

What kind of kleptocracy do you want?

Vladimir Putin gave an interview to the FT a couple of weeks ago, to which Branko Milanovich has written an interesting response. Milanovich is no apologist for Putin but, drawing on the ideas of Mankur Olson, notes that if your only choice is what kind of kleptocracy to live under, you might prefer Putin’s stable version to the more chaotic one under Yeltsin (though there are some signs that “oligarchic pluralism” is back)

That sounds odd, maybe even offensive, to Western ears, but Paul Krugman argues that this has actually been the normal set of options in history - and those with the opportunity to live in a stable kleptocracy have been the lucky ones. Liberal democratic capitalism is a modern aberration - and the idea that it represents the End of History has seldom looked so insecure.

As we’ve discussed before, to survive, democracy needs to deliver consistent income growth to ordinary people. It’s done that historically, but as Alex Tabbarok notes, it’s not certain that will continue: if countries choose democracy because democracies are rich, the growing presence of rich - or even just non-kleptocratic - non-democracies represents a fundamental challenge - and one we’re not rising to on either side of the Atlantic. 

The dangerous allure of "simplism"

Ian Leslie has an important piece in the New Statesman on "simplism" - the idea that many of our political and social woes come from the allure of simple solutions to complicated problems (If you don't read Leslie's excellent newsletter, The Ruffian, do sign up). Leslie argues that in general we overestimate our knowledge (e.g. try to explain how a zip works), but in most fields, when we're confronted with our ignorance, the result is humility. When this happens in politics, though, our response is often anger and denial.

The obvious example is Brexit, where simplism has been raised to a fine art, but it's one of many (and not exclusive to one side of the political spectrum). It strikes me that simplism is a specific instance of the more general failure mode that Tyler Cowen calls "mood affiliation", which I think is one of the most important and underrated explanations of a lot of human behaviour.

It's easy to mock simplism. Most of the challenges we face (from Brexit to climate change) involve attempts to control complicated systems - and we know from cybernetics that any control mechanism must be at least as complicated as the system it is controlling (This is known as Ashby's Law and is worth exploring). But picking the right complicated solution is challenging - and the more complicated and illegible it is, the bigger the risk of disaster. We must reject simplism - but doing so, alas, doesn't provide any easy answers.

Quick Links

  1. There she blows. The returns of whaling expeditions vs venture capital
  2. Grey matters. China will get older than the US next year
  3. Flyover state. Air passenger volume has doubled in 15 years
  4. Black death. VW's cheating on emissions standards was much, much worse than you think
  5. We were promised flying cars. Peter Thiel on conservative nationalism (Counterpoint: the crucial role of foreign-born entrepreneurs)

Your feedback

Thank you for reading. I'd love it if you would forward Thoughts in Between to a friend who might enjoy it. And just reply if you have any comments or talk to me on Twitter.

Until next week,