Three Years Old, and an Idea for a Podcast

Happy birthday, Math ∩ Programming!

Today marks the end of the third year I’ve been writing Math ∩ Programming, and I’m excited to keep it going as I start my research career. In the last year I’ve started a secondary writing blog for some smaller, less technical bits, mostly to get thoughts out of my head. And while I could use this anniversary post to preview future Math ∩ Programming posts or review old favorites, I’ll instead share an idea that has been bouncing around my head for a few weeks. I’d love to hear your feedback in the comments.

I listen to podcasts and radio shows a lot, mostly storytelling and interviews. And they’re always bringing on these fancy-sounding people who write books on the New York Times Bestsellers list and who often have very interesting things to say. When discussing science they can often convey the ideas to the clueless listener, usually because it’s experimental science that’s naturally easy to understand (state the setup, state the results, hypothesize about the implications). But almost unilaterally there’s nothing substantive about math. All the mathematical content is popular math, how beautiful is \pi and such; math education, which I love to read and talk about but is common; or math history, which I’m not as interested in. And when there is some breakthrough, like Grigori Perelman solving a Millennium prize problem, the focus is entirely on the person and not the achievement. This isn’t specific to podcasts, but all news. I just happen to prefer my news in podcast form.

And so, aside from the myriad of excellent technical blogs by active researchers, what is there really that conveys the excitement I experience in theoretical computer science? There are publications like the ACM SIGACT monthly newsletter, which has a ton of book reviews and a handful of technical columns. Unfortunately it’s hidden behind a paywall, which basically immediately excludes it from being accessed by anyone not already embedded deep in academia. That being said it often has really interesting pieces like a poll by Bill Gasarch (2002, 2012) of researchers and their opinions on P vs NP. It’s really interesting to see just how much people differ on their desire to see other parts of mathematics incorporated into its resolution.

So if you don’t want to pay the ACM for a monthly newsletter, what can you do? Many of these ideas and opinions don’t exist in textbooks, and textbooks can be dry and bad at conveying why things are interesting or exciting. There are abstruse technical papers that you have to finish a graduate degree before you can even parse what’s being said. And then there are talks, which vary in quality almost as much as prose in technical papers do.

I recently came across a paper by Ryan Williams, a prominent researcher in circuit complexity. Roughly, when you study circuit complexity you try to understand which problems provably require big circuits to solve, and you study those proof techniques. It sounds boring but it’s interesting for three reasons: it’s extremely hard, there are many “embarrassing” open problems, and many of these problems imply wonderful things like P \neq NP. I actually get really excited by circuit complexity.

Anyway, this paper was titled “A Casual Tour Around a Circuit Complexity Bound,” in which Ryan reflects on the path which led him to one of the biggest breakthroughs in the last five years in circuit complexity. His writing is more or less informal (it was published in the SIGACT newsletter, though I had to access it through arXiv), and it focuses heavily on the big picture. It struck me as mostly how to think about circuit complexity. This kind of thing is truly invaluable for a graduate student and anyone, I imagine, trying to learn more about circuit complexity. Honestly, I’d love to see more of this in academic literature. Often papers are expanded from relatively simple principles into a mess of technical details, and reversing this process is slow and difficult.

But even besides these huge breakthroughs there are often really great ways to explain new problems and solutions. For example, this paper of Andrew Drucker, titled “High-Confidence Predictions under Adversarial Uncertainty,” starts with a really easy to understand setup:

A frog wants to cross the road at some fixed location, to get to a nice pond. But she is concerned about cars. It takes her a minute to cross the road, and if a car passes during that time, she will be squashed. However, this is no ordinary frog. She is extremely patient, and happy to wait any finite number of steps to cross the road. What’s more, she can observe and remember how many cars have passed, as well as when they passed. She can follow any algorithm to determine when to cross the road based on what she has seen so far, although her senses aren’t keen enough to detect a car before it arrives…

[Even if we assume the cars arrive according to a fixed probability distribution,] the frog may not have a detailed idea of how the cars are generated. It may be that the frog merely knows or conjectures some constraint obeyed by the car-stream. We then ask whether there exists a strategy which gets the frog safely across the road (at least, with sufficiently high probability), for any car-stream obeying the constraint.

This kind of story is better than coffee at keeping people awake during talks!

And so, I have been thinking a lot about what a podcast about theoretical computer science might entail. I imagine it going something like this: every episode is a half-hour conversation with a prominent researcher. The discussion would cover something about past work, something about future ideas of what’s important and a high level idea of the burgeoning techniques, and overarching questions about how one approaches research. Computer science is particularly interesting because most graduate students know enough to start working on open problems in their first year (so the topics are more accessible than, say, algebraic geometry), and because basically all of the theorems with names are named after people still active in the research community. Moreover the format of a podcast would require the interviewees to phrase their research in a way that doesn’t require a chalkboard or notation.

The hope is to popularize theoretical computer science by assuming some modest level of technical proficiency and to give access to anyone who wants to listen; say, an advanced undergraduate, an early graduate student, or a redditor who likes to argue about the Turing test. Moreover, if I were to actually run such a podcast, I could fill the listener in with additional details in a preface. The podcast could be a resource for undergraduates who want to explore the landscape of research topics before applying to graduate school, for graduate students who want to learn to think like a researcher or hear the variety of views out there, for researchers to advertise their favorite topics and get people to read/cite their papers, and for me to meet and interact with all of these great people. My advisor seems to know everyone and their lemmas, and more and more I’ve been finding myself wanting this as well (I’m just so terrible with names!). I’ve had enough conversations with researchers to know they have heaps of interesting things to say. And I travel enough to conferences and workshops. I imagine it wouldn’t be too hard to orchestrate a 30-minute conversation with one or two people. I’d just have to work up the courage to ask 🙂

What do you think of the idea? Would you listen to a theory podcast? Do you have a good idea for a catchy name? Are you a theory researcher who would like to have a conversation with me the next time we’re at a conference together? *fingers crossed* I’d love to hear from you.

Advertisements

24 thoughts on “Three Years Old, and an Idea for a Podcast

  1. Analog Dialog? Q.Idea? Kunversation? Podcastrination? I know, I’m terrible with names. 😀

    But I’d listen to the podcast, and wish you fun and luck, if you decide to go ahead with the idea.

    Like

  2. This sounds like a wonderful idea! I’d totally like to see it happen! 😀
    For a name… How about something like “turing complete”?

    Like

  3. YES YES YES.

    Just a word of advice, don’t overcommit. It will probably be more work than you imagine, so setting a conservative schedule of (say) just one interview a month will be more sustainable than the frequency of a typical podcast. I’d hate to see such a good idea get abandoned like most podcasts do. Oh, and get a good mic/equipment so it doesn’t sound like an underwater recording 🙂

    Like

    • My idea is more to do a series of interviews during conference season and release them slowly over the months. And I already have nice portable recording equipment that I’ve used for music 🙂

      Like

  4. Look forward to listening Jeremy! A great idea for an unfilled niche.
    Pick the level you want to pitch at, the cognoscenti won’t like it dumbed down and the novice will turn off if they lose the thread. That seems to me to be the toughest choice?
    You’ll need more than coffee to keep the majority of people listening if you play unbroken interviews with theoretical mathematicians? More editing involved but better if the presenter pre-conveys the ‘gist’ and uses the interviewee’s own words to elucidate, i.m.h.o. I suppose it depends upon whom your audience is, I guess?
    Listen to your favourite science presenter to analyse what makes their communication effective. I offer Dr. Norman Swan, presenter of ABC Radio Australia’s ‘Health Matters’ as a veteran whom does it very well. I’m sure there are many more, even better examples.
    [audio src="http://mpegmedia.abc.net.au/rn/podcast/2014/06/hrt_20140609.mp3" /]

    Like

    • I’m actually continually surprised by how eloquent a lot of the researchers I talk to are, and how willing a general mathematical audience is to listen to a “dumbed down” talk, though we might mean slightly different things by “dumbed down.” Often talks aimed at early graduate students have the largest audience, especially when given by a prominent researcher.

      Like

  5. This sound like Peter Seibel Coders at Work: Reflections on the Craft of Programming, http://www.codersatwork.com/, but framed as a podcast.

    When I wear younger I was interested in complexity theory and topology. Lovasz, coloring, random graphs, perfect graphs, approximate algorithms, heuristics, average complexity, Grobner bases, commutative algebra, Bayer theorem (graph coloring = solving algebraic equations), clique, relaxation, linear programming, fixed set point. I was thinking a way of generalizing the theorem for the existence of solution of univariate polynomial to multivariate polynomials (a hundred year theorem), Hilbert zero theorems, Richard Stanley work in combinatoric and polytopes, and many more topics that are very interesting but NP-hard is hard for a reason.

    After all those buzzwords, Hope to hear your podcast soon. Apart from math I am always trying to improve my English what seems sometimes infeasible, maybe I am trapped in some local minimum.

    Like

  6. First time commenting here.

    I love your blog, but most of the times I don’t read everything untill the end due to lack of time.

    A podcast would be a wonderful idea. It would be awesome to listen to your stuff on the bus.

    Hope to have news soon.

    Like

  7. Please, PLEASE do it!
    I’ve never commented here, but I have been following your blog for a while and I’m in love with it. I have also looked quite a bit for podcasts on math and computer science, but there’s nothing around. It would be amazing.

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s