Thoughts in Between

by Matt Clifford

TiB 213: The world’s surprising best university; AI safety today; China’s soft power failure; and more...

Welcome new readers! Thoughts in Between is a newsletter (mainly) about how technology is changing politics, culture and society and how we might do it better.

It goes out every Tuesday to thousands of entrepreneurs, investors, policy makers, politicians and others. It’s free.

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

Reasons to worry about AI safety

Ryan Fedasiuk has a good piece on near-term AI safety in Foreign Policy. One of the most immediate ways AI could be dangerous is its integration into military technology. Given the fragility of machine learning models - and the high tensions between the US and China - accidents or unexpected behaviour in this context could be disastrous. Fedasiuk proposes three steps the US and China should agree to limit the risk: rigorous testing for any military AI deployment; a formal channel for inter-military crisis communications; and a ban on using AI in nuclear command and control systems.

These all seems sensible, but I worry represents a low bar for policymakers on both sides. There are real risks beyond the military sphere. AI is a historical rarity in being a technology where private sector actors appear to have stronger capabilities than national security institutions. Moreover, many credible observers believe the progress of the leading AI labs is accelerating (as we discussed a few weeks ago, Metaculus's best guess on the timeline for AGI moved forwards several years in the last few months!) and multiple new actors are entering the space.

It's worth reading this piece from 2019 on the risks posed to AI safety by the US's strategic technological "decoupling" of China that we've discussed here many times before. As Matt Sheehan writes, if we make it harder for Western and Chinese researchers to interact and collaborate, what we gain in geopolitical advantage we may lose in additional AI risk. It's good to see pieces like Fedasiuk's in the mainstream media that treat AI safety as a real and current concern, but I still fear we have too few politicians and policymakers, anywhere, who can and want to take it seriously.

What is the best university in the world?

Paolo Martellini, Todd Schoellman and Jason Sockin have a fascinating new paper on ranking global universities based on estimates of the human capital of their graduates. One intuitive way to do this would be to look at average earnings after graduation - but of course you immediately run into the problem that the country you work in has a big impact on salaries. The authors use a large and unusual dataset from Glassdoor to correct for this. By looking at how college graduates' earnings change when they move country, they're able to estimate for each college what the average earnings would be if everyone was graduating into the same labour market.

The results are fascinating. Their top 50 includes many familiar names - Harvard, Columbia, etc - but the top 10 is dominated by the Indian Institutes of Technology; number one is… IIT Ropar (you guessed, obviously). To be clear, this isn't a measure of college "value add", just a measure of how highly the market values a college's graduates (I don't have space to get into it here, but the paper does a great job of various robustness checks to reduce the chance they're measuring something else). I'm a sucker for ranking pretty much anything, but as well as the intrinsic interest, the authors discover a lot of results with real-world impact. 

First, the paper suggests that poorer countries not only experience "brain drain" as a high proportion of their graduate talent, but as a disproportionate percentage of their best graduates. Second, "college graduate quality" predicts a country's number of Nobel prize winners, entrepreneurs and executives. Third, countries vary wildly in the quality of the talent they attract; the winners are the UK and the US (but for how long?); the losers are Japan and Korea. Much more in the full paper, which I highly recommend.

Why isn’t China a soft power heavyweight?

Tanner Greer asks an interesting question in a new essay: why has China's "soft" cultural power grown so little relative to its economy? China's economic and military growth has been impressive, but there are very few successful Chinese cultural exports. In Dan Wang's 2021 letter, which Greer quotes, he argues that:

the country has produced two cultural works over the last four decades since reform and opening that have proved attractive to the rest of the world: the Three-Body Problem and TikTok

Wang and Greer have different explanations. Wang sees it as the simple consequence of "the deadening hand of the state": creativity cannot coexist with censorship. Greer thinks this overstates the case; there's undoubtedly been a crackdown on dissent over the last decade... but China's cultural power was no stronger before the Xi era. For Greer the bigger issue is the "an information ecosystem sealed off from the rest of humankind". If you look at other non-Western countries, such as South Korea, that have achieved global success in cultural exports in recent years, they've done it by taking advantage of a shared set of channels, above all YouTube.

I'd add two things. First, a lot of China's recent cultural power has come from exporting censorship (the "values tariff" we discussed in TiB 86), rather than creativity - and that might suit the CCP rather well (it's also presumably one of the free speech risks to Twitter in the Elon Musk era). Second, the core Western ideas are just really popular! This is why you get China insisting that it's a democracy rather than attacking the concept head on. I strongly suspect that "soft power with Chinese characteristics" will continue to look more like Hong Kong (2019 edition, of course) than Hollywood.

Quick links

  1. Dall-E-ance. Medieval Dall-E. And the limitations of DALL-E.
  2. My kingdom for a chip. Projected semiconductor market share by country, 2030
  3. Cheery reading. Five plausible Russia-Ukraine nuclear scenarios.
  4. What’s up, doc? Turns out, surgeon quality matters a lot.
  5. Ok, Boomer! Home equity accumulation by generation. Wow.

BONUS: Remember TiB 181 and the wild story about Arm losing control of Arm China? Looks like we’re in the grand finale…

Thank you, etc

Thanks for reading Thoughts in Between. If you like it, please do share it. Forwarding to a friend or posting on Twitter is most effective.

And feel free to hit reply if you have comments, questions or suggestions.

Until next week,

Matt Clifford

PS: Lots of newsletters get stuck in Gmail’s Promotions tab. If you find it in there, please help train the algorithm by dragging it to Primary. It makes a big difference.