Podcasts

May 11, 2017 — November 4, 2024

content
music
Figure 1

Podcasts are a thing. Since being an academic has destroyed all joy from reading the written word, audio is my remaining narrative pleasure. (C&C audiobooks).

Podcasts set the bar for shallow engagement with science, history and current affairs. “I heard in a podcast that…” denotes that I heard something interesting but failed to treat it as important enough to follow up. This is probably what I should expect from an art form designed to keep me exactly engaged enough to do the dishes without either breaking those dishes or getting bored. Friends don’t let friends get away without fact-checking their podcasts.

It seems for now that podcasts open the door to some odd and interesting voices who entertain and occasionally inform me cheaply, and in a way that seems somehow less pathological and/or sleep-disrupting than prestige TV shows.

1 Podcast Lifecycle

There is a common lifecycle of podcasts, which is that any given podcast tends to start scrappy, peak, then become a boring parody of itself. I think they share this with other serially-delivered art forms, such as TV shows and comic books. Presumably, this is because it takes a while to build up an audience, and then target and profile that audience, then the funding model dictates that there must be low-variance predictable content thereafter, because that is the easiest format to monetise the audience. As a result, shows overflow the banks of their concept, or dilute it until the flavour is weak. Or maybe they get persistently better but the audience tires of the format?

There are exceptions, or at least ones where the lifecycle is so slow that we can pretend they are exceptions; I do not have enough data to distinguish between the hypotheses. Some shows sidestep the lifecycle by committing to a fixed number of episodes.

Even at the peak of any given show, podcasts have highly variable killer-to-filler ratios, because not every show is relevant to every listener. Accordingly, I annotate podcasts by how frequently they seem to me to be killer, as a hint as to how many episodes you might wish to audition before giving up and deciding my recommendation is not for you. A low killer-to-filler ratio is not an indication I think a podcast is bad, but rather that it is not laser-focused on my interests alone, which is allowed I suppose.

Observationally, the more a podcast apes the radio and commits to a regular schedule, or commits to advertisers and needs to make up volume, the more likely it is that any given episode is unexciting filler.

All of which is to say, this list is very likely to be outdated

2 Podcast Discovery

Since I wrote this, Apple’s podcast app and Spotify’s recommendation algorithms have become popular. I use neither.

Not everyone wants to hear podcasts about everything. What are you looking for? Entertainment? Titillation? A road into some area of knowledge you are not familiar with? News about your industry? A strong but baseless sense that you sound erudite at parties, without requiring substantive effort or transformative understanding?

It is hard to search for that last quality, which is my personal favourite, but here, try a podcast search engine, Listen notes. Open Culture’s podcast list also includes excellent recommendations. It skews a little virtuous and improving for my smutty lowbrow tastes, but I got some good ideas from there.

For now, I am dumping some names which I will link to and expound at some hypothetical time in the future when I have leisure, but for now, I can at least remember that I need to reference them.

3 Stats/ML/AI

  • The Gradient / Podcast - The Gradient. My current champion. If you are deep into specialist research topics, this is hard to beat, although by definition not everything will appeal to every speciality. 60% killer
  • Data Skeptic (feed) founded by Kyle Polich but now with a large crew, does Socratic-method introductions to useful concepts in machine learning. 80% killer.
  • Machine Learning Street Talk (MLST); Basically what it says on the tin. Tim gave me good book recommendations at NeurIPS. 50% killer.
  • The TWIML AI Podcast (feed) is interview-led. 50% killer. Something about host Sam Charrington’s manner enthuses me. He does not seem to dedicate so much of his time to appearing cutting-edge (although the content is cutting edge), but rather is curious and good at humbly asking the dumb questions that I wish I was better at, and finding interesting people to ask them. 30% killer. I skip the purely industrial/implementation-focused episodes.
  • Talking Machines (feed) by Neil Lawrence and Katherine Gorman. Boasts a mix of high-grade science journalism about machine learning, useful insights for practitioners and social and philosophical concerns. 90% killer. On hiatus
  • Betancourting Disaster Opinionated, practical podcast on the frontiers of Bayesian inference. 70% killer. Subscriber only. On hiatus
  • Lex Fridman Podcast (feed) 40% killer. Some good interviews with thinkers in machine learning plus also a ragtag cast of other internet provocateurs of more disputed acclaim, and a side-order of sociological/philosophical/political/economic speculation. I have tired of this podcast. Nothing against the guy, but I’ve used him up.

There are some other podcasts that I have auditioned, but none with so high an ROI for my own purposes as these. Ones that failed to make my list did so because

  1. they skew shop talk and industry trends rather than foundational research, or
  2. they lack the tightness of editing and production values to stand out in the competitive world of tightly-edited podcasts.

This is not to say that those other ones will not appeal to you, dear reader, just that they did not appeal to me so much that I diverted work time into linking to them.

Further, this is a growth area and there are more that I have not yet given a proper listen. Here are some more podcasts auditioning for a spot on the list.

4 Culture, economics and punditry

“Serious podcasts” for wonks and entrepreneurs.

  • The Joe Walker Podcast: I’m not sure how to even classify this. I don’t want to define it purely by oppositions to other podcasts. And yet, one must find a shorthand. If you’ve ever heard an interesting interviewee on Lex Fridman and wished that they were asked fascinating, incisive questions instead of vague stoner questions, this is the podcast for you. 100% killer thus far, even on matters I thought no-one could interest me on; for example, I have successfully gotten five people from very different backgrounds to gush enthusiastically about listening to his least fancy episode, a four hour interview with a retired Australian bureaucrat.
  • Sean Carroll’s Mindscape (feed) 50% killer. Although no-one seems to agree about which 50%, which I suppose is a recommendation.
  • Wil Anderson is a comedian, but his Wilosophy Podcast interviews are serious, and very well done. More Australia-specific than most here. 70% Australian killer.
  • Your Undivided Attention Podcast - Center for Humane Technology: Previously, “how does social media send us nuts?”. Now, “how does AI send us nuts?” 80% killer.
  • Uncomfortable Conversations with Josh Szeps: An Australian free speech podcast. Interview style is … actually great. Szeps can get the best version of his guests out of them, even guests who I would have thought had not enough to say to be worth the treatment. 80% killer, worth the occasional culture-war trigger phrases.
  • The Dream (feed) 60% killer. Essays in the economics of contemporary hustle culture, hope, scams and mortality.
  • Das geheime Kabinett (feed). So weit 60% killer. Vielleicht wäre es 100%, wenn ich Deutsch besser könnten.
  • You Are Not So Smart (feed). Packages Dunning-Kruger theory of mind results in a productive and useful way.
  • Odd Lots (feed) is Bloomberg’s economics podcast about and interesting parts of business. 70% killer.
  • The After On Podcast (feed) about the next big steps in technology and science.
  • The Money (feed) is the Australian ABC’s economics podcast. 70% killer.
  • Background Briefing (feed) I have listened to this show as long as I can remember; longer-form background journalism on contemporary flashpoints.
  • Stephen Fry’s 7 Deadly Sins (feed) is delightful in an erudition-porn kind of way, but more observational than analytical. 70% killer.

5 History

Many of these I listen to to put myself to sleep, because the great thing about history is I know the ending, so the stakes are decoupled from the narrative.

  • The Constant: A History of Getting Things Wrong. History of stupid stuff scripted and performed by an actual dramaturge. Just listen to it.
  • In our Time by Melvyn Bragg, best viewed via the Braggoscope
  • Rear Vision (feed) produced by the (Australian ABC) is genius. Each episode is historical context for one current event, kind of like we imagine the media should be.
  • Cautionary Tales (feed) is pop-economist Tim Harford’s version of Aesop’s Fables from economic history. The lessons drawn from each are not mutually compatible, but each is packaged in a useful way to discourage your colleagues from doing one particular non-obvious stupid thing.
  • Decoder Ring (feed). 90% killer
  • Tides of History (feed) was economic history podcast and now is some kind of paleo-history podcast. Still recommended for newcomers, but I think I wore this one out. Presenter Patrick Wyman has an entertaining CV with a rare cocktail of working proficiency in dead languages and MMA. 90% killer on the normal episodes, 50% killer on the interview episodes.
  • Stephen Fry’s Great Leap Years. 100% Killer. A fan’s history of telecommunication, which sounds terrible but is exquisite, unless you are the kind of person who likes to imagine that the current miracles of modern information technology are unremarkable, in which case what are you even doing on this corner of the internet? It has now been augmented/replaced by the Seven Deadly Sins podcast, which is confusing.
  • Rum, Rebels & Ratbags (feed)
  • The China History Podcast (feed). 80% killer. It took me a while to warm to host Laszlo Montgomery but now I am hooked. He is curious and enthusiastic about epic stories of the various peoples of China and their complicated history, warts and wonders both, and is just so warm in inviting you to share his wonder that I cannot help but listen. The world where we are all as interested as Laszlo is a better world.
  • Nice Try! (feed) is a history of attempted utopias
  • Damn Interesting (feed) is a history of small amusing episodes from history with good sound design.
  • Byzantium And The Crusades (feed) is a quixotic project to walk through the crusades with an angle that they represented a “Byzantine world war”. They document an interesting chunk of history from a commanding perspective.

6 Sound

I care a lot more about music than is normal.

7 Smut, jokes and fiction

My tastes run puerile. Everything here should be considered to have a content warning for swears, sexual content and lowbrowness.

8 Asia, especially Indonesia

When I lived in Southeast Asia, the podcast scene was not massive. But now it is.

I would like more podcasts from Indonesia and about Indonesia, but also the East-Asian/Southeast-Asian context generally.

9 Unfiled, auditioning

10 Podcasts everyone I know likes but I didn’t get into

Benjamin Walker’s theory of everything. Unable to listen without contemplating how much more fun this show would be for me to record than to listen to.

Night Vale. Enjoyed the merch but did not enjoy the listening required to earn the moral claim to the merch.

Joe Rogan Experience. I like the idea, but I do not have the time for this kind of marathon entertainment, for the same reason I do not have time for Netflix binges. I think part of the attraction is supposed to be entering a fugue state of interviewness. I would possibly listen to Joe Rogan edited highlights?

11 Make your own podcast why not

See podcasting.

12 Diversion: on hardball interview podcasts

Julia Black, How Lex Fridman’s Podcast Became a Safe Space for the Anti-Woke Elite seems … fair? Presents both some sus things and some good things about the guy. tl;dr it’s complicated. If you do not like Dr Fridman, you can read this negatively and if you do like him, you will not find anything especially damning here about his interests or incentives.

I would be more interested in an analysis of him as an example of interviewer trade-offs as regards a podcast. For example, notoriously, Lex does not push back on his (diverse and interesting) guests, even when they say very crazy stuff. Presumably this also attracts more interesting guests who feel they will be treated unfairly elsewhere. At the same time, many potentially interesting questions don’t get asked. Is his pushback as strong as it could possibly be while still attracting interesting people to interview? What is the optimal frontier of softball interview questions and interesting guests, and is Lex on that frontier?

I think there are likely multiple axes of interestingness: it seems that pushing back on controversial opinions, and asking interesting questions are both interesting things to do. They sometimes coincide but do not necessarily.

I suspect that Lex could up his oppositionality somewhat, or his depth of questioning a lot, and achieve a strictly more interesting interview without losing too many guests. But I do not run the podcast; so what do I know?