Welcome new readers! Forwarded this email? Subscribe here. Enjoy this email? Forward it to a friend.

We’ve talked several times - see, e.g., TiB 71 and TiB 119 - about the growing scale of machine learning models and the associated increases in the cost of training them. What’s the impact? A new paper argues that it’s made AI more elitist: starting from the deep learning revolution, papers accepted at top machine learning conferences are increasingly dominated by researchers from the largest companies and top-ranked universities (interestingly, this doesn’t appear to be true in non-ML computer science disciplines).

Nevertheless, the conclusion doesn’t quite ring true. As Gwern points out here, it’s hard to look at the last decade of machine learning and not see an extraordinary opening up of the “state of the art” to individual researchers. True, they cannot pay tens of millions of dollars to train models with billions of parameters, but the combination of open research norms (Arxiv is a minor miracle), open source tooling and infrastructure, and so much public discussion and education by the field’s luminaries are impressive resources for democratisation.

There’s another dimension here worth pondering: the amounts spend by organisations like OpenAI or DeepMind are enormous relative to the resources of individuals, but they’re rounding errors compared to budgets of nation states. OpenAI raised $1bn to much fanfare - but this is a couple of percentage points of the UK’s defence budget. If you share the view, as many smart people do, that AI is a geopolitically important technology, it’s really very inexpensive. Any medium sized power can afford the cost of a world-class research lab. Talent, not cash, is the bottleneck today. 

Western democracies are increasingly divided along educational lines. To a first approximation, the populist electoral victories of 2016 on both sides of the Atlantic were the revolt of those without a university degree. College-educated voters are now firmly entrenched as the core of the liberal/left electoral coalition in the US and UK. We talked about this way back in TiB 11 - and the Thomas Piketty paper linked there is still probably the best guide to the phenomenon.

It’s easy to point to ways in which the interests of the groups either side of the university dividing line have diverged. But a new paper by academics Jean Twenge and Bell Cooper suggests something more fundamental is going on: the gap in self-evaluated happiness by education (and income) level has grown significantly over the last 50 years. You see this starkly in this chart from the paper:

That this is driven primarily by falling happiness among the worst off is particularly troubling - particularly when you combine it with this horrific data on the growing rates of “deaths of despair” among those without a degree. As Peter Turchin says, it’s hard to maintain societal stability and coherence when psychological inequality becomes stark. Understanding the causes of and cures for this divide feels of pressing importance. 

We’ve talked a lot this year about growing technological conflict between China and the West (see, e.g., TiB 124, TiB 126, TiB 134). Matt Sheehan has a good piece on the Macro Polo blog (or tweet summary here) on the way Silicon Valley’s relationship with China has evolved over the last decade. 

Sheehan sees three broad periods: from 2010 to 2013, China was viewed as a copycat and a potential market (Apple, LinkedIn and Uber all tried to go big on China); from 2014 to 2016, China was seen as a market and an inspiration (see, e.g., a16z’s famous piece on WeChat); and since 2017, China increasingly seems a competitor (TikTok, above all, but more broadly the realisation that market access will always be asymmetric) and a warning sign (especially the use of technology in Xinjiang).

In some ways this mirrors the shift in attitudes in the US’s political establishment. As I’ve said before, China hawkishness is a rare bipartisan issue in America (see also this new and very hawkish report from the US State Department). But nuance is important here. Politico’s China Watcher newsletter has an excellent piece on the need to revisit our China narratives. China is vast, complex and far from monolithic. Having an accurate model is a key first step - for both Silicon Valley and DC - as the West decides how to navigate relations with the world’s emerging superpower. 

Thank you for reading Thoughts in Between.

I’d love you to share the link on Twitter / Facebook / LinkedIn - or simply forward this email to a friend.

If you had feedback, feel free to email me or message me on Twitter.

Until next week,

Matt Clifford