Turing’s Cathedral

Just wrapped it up, and posted some thoughts to Goodreads (I’m thinking of automating that process here — let me know if any of this blog’s few readers would find that irritating). The short version: it’s a great history that’s marred by a bunch of fairly silly futurist speculation. I’d be curious to hear if anyone else has read this thing.

the myth of American decline

Being the crotchety old man that I am, the time I spent this evening on my gym’s treadmill left me feeling cantankerous. I had been watching Jeopardy, and all of the categories seemed horrible, dagnabit. Back in my day we didn’t have questions about sitcoms! No, it was all Latin, and poetry, and similarly high-minded pursuits.

Then I got home and remembered I had a bunch of code left over from when we built this thing. See, there is a terrifying website called j-archive.com. It’s maintained by former players, and it comprehensively chronicles every game of Jeopardy.

It’s possible to scrape this site to reconstruct games, which is what I did for the Cordray infographic. With this as a starting point, figuring out the percentage of categories devoted to television versus weightier topics was a relative cinch. I was absolutely confident that I would find a line snaking smoothly upward. Here are the regular expressions I used:

RE_TV = re.compile(r'(T\.?V\.?|TELEVISION|SITCOM)', re.I)
RE_BIBLE = re.compile(r'BIBL(ICAL|E)', re.I)

And here’s the graph that resulted (normalized by total number of categories in a season):

Gotta say, I didn’t see this one coming. I guess the nerds are (mostly) all right after all. Alex Trebek’s still kind of a supercilious asshole, though.

Anyway, I’m open to other suggested analyses. Lay ‘em on me.


Apple didn’t trick Google into making better iOS maps

Matt has a post up arguing that today’s debut of a seemingly-great Google Maps iOS app represents a strategic success for Apple. The basic idea is that Google used its head-start on mapping to artificially retard the iOS maps experience, in an attempt to give Android a competitive advantage. Now, with Apple competing, they’ve been forced to deliver a first-class, free maps experience to Apple customers.

I think this is a valid description of the present dynamic, but not of how we got here — the account of the underlying strategy is a bit too 7-dimensional-chess-y for my tastes.

What really crystallized my thinking on the maps issue was a conversation with Eric Gundersen of Development Seed/Mapbox. Over beers at Townhouse, I asked him and his colleague Alex Barth what the hell was up with Apple’s maps strategy. Why set up a gigantic and difficult new internal practice? Their lack of expertise meant they had low odds of success, it would be horribly expensive, and Google’s model of offering free computing services to keep users in their ad ecosystem seemed like it was pretty compatible with Apple’s needs.

Eric’s response was pretty simple: “The future is mobile, right? Mobile’s about location. If you want to own that, you need to own the map layer.”

I think that’s right. This mapping fight isn’t about iOS or Android specifically, but rather a play to avoid dependence/achieve dominance over an increasingly vital (and potentially expensive) informational asset. If your software giant steps into the ubiquitous computing/augmented reality era wholly dependent on a mapping provider with monopoly power — which Google was on track to be — you’d be pretty well boned.

So they made an investment (though a rather chintzy one, by some accounts), took a reputational hit, and have their fingers crossed that iOS’s popularity will subsidize the development of their mapping stack into a competitive informational asset.

Google, meanwhile, is facing its own headaches.  It’s only in the last year that they’ve tried to monetize GMaps in a big way, and almost immediately they were forced to drop their prices due to credible competition from upstarts like the aforementioned Mapbox (which is pretty awesome, by the way, and deserving of a lot more fawning profiles than I’ve seen so far — if you want a DC software startup with a credible plan for world domination, they’re the ones you should be talking to, not the guys selling coupons).

So Google’s clinging to its vision of a consumer mapping monopoly, I think, by focusing on the quality of their offerings. It’s a credible approach: OpenStreetMap still has licensing problems, everyone hates Apple Maps, and Google’s privileged access to oceans of information about consumers’ preferences and desires really does give them a competitive advantage. They have geographic information that no one else seems to have, and ambitious engineering projects that are going to be very, very difficult for competitors to replicate without paying Google for GIS privileges.  So they’re working to build an unapproachably awesome map stack, and retaining users is part of that — as Apple Maps have amply demonstrated, there’s no substitute for users using your product and slowly helping to refine it.

Matt’s right that this dynamic is good for iOS users and consumers in general. It’s competition! But I think the competition is happening over the map layer, not the handset OS layer. Geographic information isn’t just a handset feature, it’s a potential monopoly. These guys legitimately want to crush each other; it’s not just brinksmanship in service of preserving a cross-firm subsidy.

a poor use of my time

Still, the lingering shreds of my 14 year-old self couldn’t help wasting a few hours writing a convenience utility for extracting snippets from the Simpsons for throwaway-gag social media use. I did this despite realizing that, yes, quotation is a low and basically irritating form of humor — it’s basically the same as the bully with the audio-playing jacket in Back to the Future II . Still, if you’ve got a library of video you want to pull snippets from, perhaps you’ll find it useful.

Important notes/caveats to this important work:

  • Having to upload to an FTP endpoint sucks. Using a video service would be great, not least because they’d handle the tedious ffmpeg tweaking I wasted a bunch of time on. And, in fact, I had this working with YouTube. But their copyright infringement detection algorithms are too good. It’s a shame; I feel that quotation of this kind is fair use.
  • Quicktime is a real jerk, and ffmpeg is a mystery. For an embarrassingly long time I couldn’t get Apple’s default OS X codecs to play the H.264 file I was making (VLC played it no matter how badly I mangled the parameters, of course). Using the .mov contained was the trick. Bah.
  • Is there really no URL shortener that will work without an API key? Weird. Weird and stupid.

Mostly this saves me a minor amount of trouble — the command line is faster and more flexible than the VLC/SimpleMovieX/CyberDuck workflow I used to employ. But my real motivation has more to do with a pie-in-the-sky featureset I’ve daydreamed about for a while:

  1. Enter text phrase
  2. Search database of extracted subtitles for timestamps and surrounding text.
  3. Select desired quotations.
  4. Search for moments of audio silence surrounding the window indicated by the rough subtitle timestamping.
  5. Potentially repeat the process with video scene transition data.
  6. Plug results into today’s script, automating the gap from remembering a line to pulling the video for it.

Again, a huge waste of time. I don’t even have a torrent with subtitles yet! And I have a ton of projects I ought to get to before then, not least of all my mom’s website. But if I were a collective of infinite monkeys, I’d certainly tackle this. Hell, one could conceivably connect it to work, if you ignored C-SPAN’s copyright and pulled all their video and transcription.

A more tractable next-step is probably adding animated GIFs as an output option.