TiB 178: How the Taliban fight; AI researchers on AI governance; Ethan Mollick 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.

“Fighting like the Taliban”?

This will likely be a week of bad Afghanistan takes (e.g. no, we don’t know what implications it has, if any, for China’s Taiwan policy), so as it’s a long way from my expertise, I won’t offer one of my own. However, I thought it might be useful to signpost the best things I’ve seen so far. One of the biggest questions for a lot of people is how this all happened so fast. A good rule of thumb is to pay more attention to accurate predictions than post hoc explanations, so I’d highly recommend this short piece from Tanner Greer, “Fighting like the Taliban”, published a couple of weeks ago.

Greer argues that you have to see the Taliban’s rapid advance in the context of the pre-2001 history of conflict in Afghanistan. He quotes Dexter Filkins’ Forever War:

People fought in Afghanistan, and people died, but not always in the obvious way. They had been fighting for so long, twenty-three years then, that by the time the Americans arrived the Afghans had developed an elaborate set of rules designed to spare as many fighters as they could. So the war could go on forever. Men fought, men switched sides, men lined up and fought again… [The] people who really took fighting seriously were the foreigners— that is, the Americans and Al-Qaeda. They came to kill.

Essentially, the historical norm in Afghanistan was to switch sides once the winner became clear and before too many people died. That’s what’s happening now and why resistance has been minimal. Do read the whole thing.

I’m a long time admirer of Greer’s work. One of my favourite of his posts is this 2019 piece in which he outlines how he’d write a history of 21st century America so far (We discussed it in TiB 70). At the time, he noted that it wasn’t yet clear what would be the right point at which to end the book:

That moment has not yet arrived. I suspect, however, it will arrive soon

He was right. I suspect history will yield many, many books that take 2021 as their endpoint - or indeed their starting place.

Bonus 1: I’d also highly recommend this board game, A Distant Plain. One reviewer described it as the “best piece of media on our war in Afghanistan”.

Bonus 2: Extraordinary Twitter thread from the central bank governor of Afghanistan on his experience in the last week

Who do AI researchers want to shape the future of AI?

As we’ve discussed before, AI researchers are an important constituency in determining how AI will develop as a political and strategic force (at least for now - see Jade Leung’s work, discussed in TiB 149, for why that might diminish over time). This new survey of leading machine researchers on how they think about AI governance and safety therefore makes for interesting reading. There’s a good Twitter thread summary here.

Some of the most interesting (and dispiriting) findings are about which institutions researchers trust to shape the future of AI in the public interest. The graphic here summarises this well. Overall confidence is low: the top rated institutions score around 2 on a 0-4 scale, with arguably the most powerful groups clustering around 1. It’s striking to see just how low trust is in Facebook. It scores much lower than Google, Apple and Microsoft and roughly in line with the big Chinese tech firms. I’d be fascinated to see data on whether these scores influence willingness to work for these companies.

The survey also suggests that researchers are becoming more concerned about AI safety relative to a similar study from 2016. Their specific concerns, though, are interestingly different from those of the public. In particularly, as a group they seem to be much less worried about US/China competition, a topic we’ve discussed many times (see e.g. TiB 173). Unsurprisingly, home country is a significant variable for a number of questions, but perhaps less so than I expected. It will be interesting to see if and how this changes as strategic competition between the US and China intensifies.

What do we actually know about entrepreneurship?

This week’s TiB podcast is a conversation with Ethan Mollick, professor of entrepreneurship at Wharton. Long time TiB readers may also recognise him as almost certainly the most linked-to person on Quick Links (this week is no exception); he has possibly the world’s most interesting Twitter feed. Ethan is a former founder who has spent his academic life investigating what the data can really tell us about success and failure in entrepreneurship. He’s written a book on this topic, The Unicorn’s Shadow, which I highly recommend.

This conversation was the first time Ethan and I had met, outside Twitter, so it was fun to compare the Entrepreneur First approach to building companies with the lessons from the academic literature. I’ve always said that EF violates a lot of venture capital conventional wisdom and the same is certainly true of at least some academic wisdom! We both really enjoyed this part of the conversation and I hope it might lead to some further collaboration and study of EF’s company building process.

Topics we discuss include:

Where the best ideas come from
The most dangerous entrepreneurship myths
What makes a successful pitch
How to hire
The entrepreneurship simulation game that Ethan has spent the last decade building

Enjoy!

Quick links

Decline and fall? Chart of US refugee admissions over time.
Game, net, match. Apparently fish solve the prisoner’s dilemma.
Living the dream (machine). Fantastic thread of reading recommendations on the history of computing.
Dark ages. Fascinating paper that models early modern witch hunts as an consequence of Reformation-era religious competition.
When AIs troll. (Apparently) real demo of OpenAI Codex‘s impressive code writing (if juvenile joke making…) capabilities.

Help me grow

Thanks for reading. If you enjoy this, please share it with someone else who might, either by forwarding the email or sharing on social media.

I'm always happy to hear from you - feel free to hit reply.

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.