Weekend Reading & Selected Links

6 min read

Happy weekend! Here are some links to things I've been reading that you might also enjoy:

  1. My 2023 retrospective podcast episode. In this episode, the tables were turned and a listener, DJ Thornton (an economics PhD student from Sydney), interviewed me. At the bottom of this email, I've included six excerpts from the conversation.
  2. 'Is the west talking itself into decline?', a new FT article by John Burn-Murdoch.
  3. Australia's greatest cultural export?
  4. If you haven't yet, it's worth checking out Perplexity – a search engine that uses AI. I've found it useful, and it's given me information (for an essay I'm writing) that I wasn't able to elicit from GPT-4.
  5. 'The Iron Law Of Evaluation And Other Metallic Rules', by Peter Rossi (1987).
  6. Former guest of the pod Timur Kuran's take on recent events at American universities.
  7. The best sets of (unpublished) lecture notes that can be found freely online.

Have a great weekend,‌

Excerpts from my 2023 Retrospective

1. What percentage of good interviewing is just good scholarship?

DJ THORNTON: What percentage of good interviewing then, would you say is just good scholarship, good preparation? 

JOSEPH WALKER: I think it's very high. For me, the main lever for the quality of an episode is just: how much prep can I get done? Maybe, I don't know, 80%.

2. Why did I change the podcast's name?

THORNTON: On August 7, Joe, you posted a Twitter poll asking your followers whether you should change the name of the podcast to something less ridiculous than 'The Jolly Swagman Podcast'. Loaded question, by the way. Almost 70% of the 431 people who responded said, “No, leave it as is.” 20% said, you should change the name. 10% were not sure. But you changed the name. What's going on there? 

WALKER: So I'm trying to persuade the marginal listener to subscribe to the show. And there was probably a massive selection effect where the people responding to that poll were existing listeners who had some kind of attachment to the name, some kind of status quo bias. But I don't care about them. Well, I do care about them, a lot. But I don't care what they think about the name, because they'll keep listening to the show in all likelihood. I'm trying to convince the person listening from, I don't know, America, who is finding the show for the first time and is like, “The Jolly Swagman? What the f--k is this?”, who is rightly confused, or was rightly confused, about the old name.

3. How I covered AI in 2023

THORNTON: We've already mentioned AI, and AGI should make its way in there, too. What would be some of the other themes that you would say have run through the year? Some of the dominant themes. 

WALKER: Firstly, I don't feel like AI was a massive theme. I'm not sure what you think, but I feel like this was kind of a blind spot for me in the show in 2023. Relative to how hyped and important it was as a story in the world, I don't think I gave it much coverage, which was a somewhat deliberate decision, to just hedge against that...

THORNTON: This is good. So, firstly, I suppose it sort of depends on how we define what a theme is. AI or AGI came up in six of your episodes this year. 

WALKER: Okay, so I was relevant. 

THORNTON: You were relevant. So Kahneman you asked whether AI systems will reduce bias. 

WALKER: Oh yeah. And will they do it consistently or affect some biases more than others. 

THORNTON: Yep. And then in your Palmer Luckey interview, you asked him about VR in a post AGI world, and the risks of a country like China getting AGI. 

WALKER: And will AGI actually deliver a more final kind of means of experiencing and communicating to us than VR.

THORNTON: Yes. And then in the Richard Rhodes episode, you drew this parallel between the development of AGI, and the making of the atomic bomb. You didn't really talk about it, but it did come up. 

Obviously, you did talk about this with Stephen Wolfram, in terms of the implications of computational irreducibility for AI. But it sounds like we've got another episode coming that will flesh that out a little bit more. 

Then it also came up in the Peter Singer episode, because you asked him about the ethics of superintelligent beings. 

WALKER: And if AI take over and cause human extinction, should we just kind of fade into history and wish them well, from a utilitarian standpoint.

THORNTON: Yeah. And he said yes. 

And then, of course, in your last episode, you talked kind of at length – well, I mean, Stephen Pinker and David Deutsch talked about this at length.

So I think it was maybe more relevant than you think. 

WALKER: Yeah, I think you're right. Although I still feel as though I didn't give it the truly rigorous coverage that it deserves, including and especially in the Pinker and Deutsch episode. 

4. The dyads theme

THORNTON: So let's maybe talk a little bit about dyads and pairs. So you have this running hypothesis that pairs can advance science –

WALKER: Or any creative field. 

THORNTON: Yeah, okay. Or any creative field, in a way that individuals cannot, and that groups of three or more also cannot, because they can kind of bounce ideas off of each other. I know you like the book Powers of Two

WALKER: I actually don't like that book, but it's a good reference book.

THORNTON: Okay, well that's good, because I want to ask you kind of where did this hypothesis come from for you, and where does it sit now at the end of the year, having talked with a couple of different people about it. 

WALKER: Yeah. I mean, the conversation with Kahneman about it was pretty special and touching. Like the way he talked about Amos and the emotion that he spoke about him with. 

So originally, my thinking on this came about through contemplating examples of pairs or noticing lots of examples of very fertile pairs – maybe the first of those was Kahneman and Tversky – and that was actually long before I encountered Josh Shenk's book Powers of Two

I think now, my thinking on it is maybe a little more developed in that I've tried to think of, like, okay, how do you model this? Or at what points are pairs the optimal creative unit? So that's where I'm at now, and I'm intending to write this up into some kind of blog post or essay that I'll publish on my website in the new year. 

5. On reading Thinking, Fast and Slow, and interviewing Daniel Kahneman

THORNTON: When you read Thinking, Fast and Slow for the first time, it took you about three months to read that book. Why did it take you so long? 

WALKER: It's just really dense. 

THORNTON: Okay. I mean, there were lots of other dense books that you were reading at the time, kind of going back, and they were only taking you sort of one or two weeks. Was there anything special about this particular book? 

WALKER: I wasn't reading it continuously. I was probably reading other things as well, and context-switching with other books. But I think from memory I must have finished it sometime in 2017? 

THORNTON: Yeah, I think it was 2017, January to April or something like that. 

WALKER: Yeah. I think I remember wanting to actually internalise it, and so I spent a lot of time just re-reading stuff, making sure I'd understood it. And it's a dense book. So that's probably why. 

I mean, to my credit, I dare say most people who own that book haven't read it or haven't finished it. I actually made the effort of finishing it. 

THORNTON: Well done, well done. Have you wanted to interview Kahneman since then, or has that been kind of on your list for a long time? 

WALKER: Yeah, I think I might have emailed him in the early days of the podcast and then maybe roughly every year since then. 

THORNTON: So the persistence pays off. 

WALKER: Yeah. And in a way, it's kind of good that he didn't agree initially because it would have been a radically different interview, a much worse interview, if he'd come on the show in the early days. 

6. My biggest podcasting lesson from 2023

WALKER: So I feel like this is the biggest thing I've maybe not necessarily learned – because I always knew it subconsciously or implicitly – but maybe this is the biggest lesson that's been reinforced to me as an interviewer: you have to morally deserve the guest's best material. So you have to convey status to the guest, whether that's in how you come across in your emails or how you carry yourself at the beginning of the interview, how you demonstrate the level of research you've done through your questioning and the context and preface that you attach to your questions. Showing them that you've done a lot of work and you're the kind of person who deserves their best information is a really important but underappreciated way to make interviews go well. That was something I reflected on a lot this year. 

THORNTON: Yeah, it's funny, we talked about this when we talked about me doing this interview, which is that it kind of seems like a big signalling game. 

WALKER: It is.

THORNTON: You have to prove to them that you're worthy of their time. Or, I mean, you're saying not just worthy of their time, but worthy of good answers. 

WALKER: Worthy of good answers, exactly. Yeah, podcasts: it's just all one big signalling game.