Matt's Thoughts In Between - Issue #75

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

OpenAI, Microsoft and the end of the world

OpenAI announced a $1bn investment from Microsoft. It says it will use the money largely for computational power, which is increasingly the core enabler for machine learning (see previous coverage). Interestingly, OpenAI says it’s raising capital specifically to avoid having to develop a commercial product, which will allow it to focus on competing with Google to develop artificial general intelligence (AGI).

This is the second large investment for OpenAI since it announced its change in structure back in March. There’s been quite a lot of skepticism from leading figures in Silicon Valley and/or machine learning - from those who see the move as pure PR to those who are highly critical of its non-profit to for-profit transition and its increasingly non-open approach. Do read this thread from AI researcher Stephen Merity, which makes that argument.

I'm sympathetic to both sides here. On the one hand, even if compute were free, AI is a high-end talent game, and - as we discuss in the next section - it's very hard to employ top talent today without equity compensation. OpenAI has done a decent job of threading the for-profit/not-for-profit needle. On the other, whenever OpenAI makes a big announcement, I go back and reread this excellent and anxious post by Scott Alexander from OpenAI’s launch in late 2015. It strikes me that, if you’re worried about AI safety, Alexander’s concerns, especially in the final section, have held up a lot better than Musk and Altman’s reassurances.

We're all (human) capitalists now

A fascinating new economics paper, which I came across via Tyler Cowen, looks at the phenomenon of “human capitalists” - high-skill workers who receive a big chunk of their income via equity. This includes startup founders, of course, but also executives in any company with a stock compensation plan. The important idea is that, in terms of traditional economic analysis, these people act like labour, but make money like capital.

This is important for understanding the modern economy. According to the paper, human capitalists receive 45% of their income from equity and own ~7% of public companies. Taking this into account, the well documented decline in labour’s share of income since the 1980s drops by 60%(!) and the share earned by high skill workers increases from a third to a half. The authors find, unsurprisingly, that high skilled workers are a complement to physical capital - i.e. human capitalists do better as investment increases - and so this group has benefited from the falling cost of capital since the crisis.

Reframing an argument I’ve made before in these terms, ambitious individuals will increasingly want to be human capitalists. One feature of the digital economy is that (pace arguments about the cost of compute above), the upfront capital cost of starting a company has fallen rapidly. This means power has shifted from those who can write cheques to those who can write code. One of the big questions of the next few decades is how scalable this is. Can everyone be a human capitalist? Organisations like Lambda School think so. If they’re wrong, the result will be more inequality and dislocation.

The most interesting person in British politics

Boris Johnson became Prime Minister of the UK this week. Perhaps the most interesting thing he did in the first days of his premiership was appoint Dominic Cummings, the architect of the Leave vote in the Brexit referendum (and Benedict Cumberbatch’s character in the film) as a senior adviser with real power in Downing Street.

Whatever your politics (and I don't share Cummings') Cummings is a fascinating figure, perhaps the most interesting person in British politics today. As well as Tim Shipman’s superb book on the Brexit referendum, All Out War, which looks at Cummings in detail, I recommend this short profile of him by Labour adviser John McTernan and this Twitter thread by notorious Gordon Brown aide Damian McBride. And, to complete, the everything-is-related circle, here’s a photo of Cummings in an OpenAI t-shirt.

This shouldn't be a surprise; Cummings is perhaps the deepest thinker on the implications of technology of anyone engaged in high-level politics in the West. His blog is extraordinary for its engagement with cutting edge ideas from tech and beyond. Start here and don’t miss his series on the referendum. What does this mean for British politics? It’s hard to say (though this is the best guess I’ve seen). Cummings is, of course, a Brexiteer through and through - but no fan of the Conservatives’ Hard Brexit wing in Parliament. His goal has always been not just Brexit, but a wholesale reform of Britain's establishment institutions. It will be an interesting ride.

Quick Links

Everything is a machine learning problem. An apparent breakthrough in translating ancient language Linear B.
New problems in human science. A lot of the time, people are just screwing with researchers...
No mules are available right now. Amazing "Google Maps of the Roman Empire"
A clean meat bubble? Beyond Meat's market cap surpasses Campbell's
Turning points in internet history. Extraordinary thread on how Blockbuster nearly beat Netflix (via Michael Orland)

Your feedback

Thanks for reading and Thoughts in Between. I'd love it if you’d share it with a friend who might enjoy it. And feel free to hit reply if you have any comments - or talk to me on Twitter.

Until next week,

Matt