August 2000

1 Aug 2000[link to here]

Mmm. Tuesday. That means science from the NYT, including the continuing theme of Insects Have Cool Ways of Killing Each Other. This time it's ants and butterflies.

Sweet. jwz's page is a snippet of an MP3 file, as viewed in Emacs' hexl-mode (with some values clobbered so it says something). I pieced it back together. It's about two seconds long. I don't recognize the song. Do you?

Why doesn't my browser make a portal page for me? It knows which URLs I keep visiting. Since I already run a URL logging proxy, a rough version should be easy. I'd still want something organized decently, and I don't expect a program to manage that tricky task, but it could handle a "things not already on this page that should be" section. Yet another thing for the to do list...

deja.com has gone off the deep end. I'm looking at a message, and it tells me what forum (newsgroup) it was from, and it doesn't give me any way to see other messages in that forum except by manually entering a new search that describes such a view. But they're hiring, so if I wanted to perform an apoplectic interpretive modern dance inspired by their UI (translation: gun them down), I could at least do it as an inside job. (Note to other potential employers: I'm kidding. I don't do modern dance.)

I've been writing a lot, but I'm not yet sure what I'm writing about. I like doing that though. I start out with some ideas and keep expressing and reorganizing them until they click into a clean and elegant structure. That's when I know I've got it.

NYT article:

A new study of male body images found that the average young man wants to look like Superman and thinks that women want men to look like Superman, while women really want men to look like Clark Kent.

2 Aug 2000[link to here]

People keep telling me I should do some street performing during this Month O' Festivals, but not a single person has given me any reason other than "because you could". Not very convincing.

3 Aug 2000[link to here]

We got some new computers in the lab yesterday and we had a little fun with the packaging. (More than with the computers, I think.)

4 Aug 2000[link to here]

Paul threw himself a lovely birthday party last night. I almost never get pictures of myself as I'm usually on the not-so-pointy end of the camera, but last night was an exception thanks to Marietta. I'm the one with the unicycling t-shirt and the rosy three-glasses-of-wine cheeks.

I got to try two new (to me) British foods. I'd seen too many commercials for Twiglets, but still had no idea what they were. (Slogan: "Twiglets are weird".) I assumed they were just oddly shaped pretzel sticks. I tried one. It tasted funny. The moldy-looking patches turn out to be marmite. It was more weird than good or bad. I just wasn't expecting it. I'll have to try marmite in a proper setting at some point.

The other novelty was a hedgehog. No, not a real hedgehog, but a half orange with toothpicks spiked into it. On each toothpick was skewered a chunk of cheese and a chunk of pineapple. (I foolishly didn't think to take a picture. Argh!) These yummy hedgehogs are apparently common at children's parties; everyone had fond memories from their youth of not wanting to eat the cheese and the pineapple at the same time. It was really good though. At really classy parties, the hedgehog has cherry eyes.

5 Aug 2000[link to here]

Wow. Last night, someone showed me a short New Scientist blurb about automatic music classification (my thesis topic), and today it's all over slashdot. If that had happened six months ago, I could have gotten other people contributing code to my thesis. I probably would have had an easier time keeping my motivation up too. Working alone gets old after a few months.

The weirdest part though is seeing other people say some of the things I've been spouting for years, mostly about all the great applications for this. I kept thinking, "Yes! Yes! Yes! Now help me write it!"

I'd been planning to start an open project on music analysis after I finished here. I have to spend the next month writing up my results, but I can't waste all that good publicity, so I created a SourceForge project today, project name "vole". Picking a name was hard until I stopped trying to make it relevant. I also quickly put up some links to music analysis research in case anyone is actually interested.

7 Aug 2000[link to here]

I so dig psychoacoustics. I dig psycho acoustics too, but that's different.

Back in eighth grade I did a science fair project (well, a report, really) on how the human ear worked. I remember wondering how people could tell whether a sound source was in front of them or behind them. Loudness differences between the left and right ear only help you figure out how far left or right something is, but for any given volume ratio, the source could be anywhere on a circle (forward, backward, above, behind, etc.).

Back in eighth grade, I was kind of dumb and it didn't occur to me that I could probably track down the answer. Of course, this was before the web, when tracking down information was a big pain in the ass. But eventually, I found out that it had to do with the shape of the pinna. (That's the fleshy outer ear, also good for holding up glasses and jewelry.) Then I had the clever idea of sticking two microphones inside a fake head's ears and listening to the recording on some headphones to see if I could correctly place sound sources. But I was lame, and I never did it. Other people are less lame though, so rather than building fake heads, anyone doing 3D sound simulations can just use the well-known HRTF (Head-Related Transfer Functions).

Another neato bit of trivia is that our audio cortex is arranged tonotopically. Tones with close frequencies excite portions of the brain near each other, and if you listen to a rising or falling tone, the region of excitement will slide neatly along your brain. For those interested in this stuff, I recommend Music, Cognition, and Computerized Sound: An Introduction to Psychoacoustics, edited by Perry R. Cook.

Ok, one more bit of fun psychoacoustic trivia. If you play two pure (sinusoidal) tones with frequencies almost, but not quite the same, you'll hear relatively slow beats as the waves go in and out of step. That makes sense, and if you sit down with some graph paper (or an oscilloscope) you can see how it all works. What's wacky is that you'll hear the beats even you play the two tones into separate ears and the sound waves are never combined (except mentally).

9 Aug 2000[link to here]

The NYT ran an article yesterday about how culture affects our way of thinking. By itself, That's not very surprising, but the depth of the effects is interesting. There were several examples of how the Japanese subjects paid more attention to context and less to the foreground, which is what Americans typically focused on. That certainly fits with what I know of the two cultures.

But the greater attention paid by East Asians to context and relationship was more than just superficial, the researchers found. Shown the same larger fish swimming against a different, novel background, Japanese participants had more difficulty recognizing it than Americans, indicating that their perception was intimately bound with their perception of the background scene.

Today, while watching the zillionth picture-taking tourist outside my window, I thought about the Japanese stereotype of holiday pictures, in which a family poses in exactly the same way in front of different backgrounds.

10 Aug 2000[link to here]

A few weeks ago, I was (finally) describing my audio feature extraction to someone else who knows enough that I had to go into detail. We both immediately noticed a dumb mistake I'd made. I had thought about it before, but had swept it under my mind's rug. That's harder to do when trying to justify decisions to someone, and this is why I should have been writing up my thesis as I went along. Local wise folk recommend doing that, but I foolishly ignored them.

Once the mistake was identified, I should have immediately fixed it and rescanned all my music. But I thought it would take too long, so I decided I'd just mention the error in my writeup and suggest that it be done differently in the future.

Well, I'm writing everything up now and I couldn't bear to write that I'd made such an obvious mistake, so I'm fixing it now, even though I don't really have time. I need to do a final run, plot and comment on the results, drop in some graphs, and so on, but instead I'm writing around those parts, writing everything that doesn't depend on actual results, and hoping I finish reprocessing the 96GB of data before I run out of other things I need to do.

Fixing it now is certainly a hassle, but leaving it would have annoyed me for years, and I'm glad I realized that in time (barely) to change it.

11 Aug 2000[link to here]

I finished rescanning all my data. Over 100GB. Yow. But my classifiers are more accurate now, which is a nice bonus. All I really wanted was peace of mind and non-idiot-looking-like, but I'll take that performance boost. 91% accuracy on pop/rock vs house/techno.

12 Aug 2000[link to here]

Bandwidth is cheaper than attention. I don't have time to read all the news that's fit to print, and as printing becomes cheaper, the quality threshold drops. Reputation is still a valuable asset though, so I'm not too worried. Today's edition of Worth Your Valuable Time includes an excellent article from the NY Times on TiVo and ReplayTV and how they will change television, marketing, and consumerism. It's got great pictures too.

People who watch commercials subsidize people who don't; people directly influenced by ads subsidize people who watch ads with ironic detachment.

When I was more interested in natural language processing, I wondered how I could work on that without working for The Man. Now I wonder how hard it will be to work on preference-attentive software agents without selling people out to marketers. A growing free software community will help; even solutions that require a large distributed infrastructure wouldn't necessarily depend on a revenue stream. A growing public concern with privacy will also help. Selling out User #340984 isn't a big deal if the ID isn't attached to anything important like a real name. It won't be long before there's a popular mechanism for whipping out disjoint identities for different things, and public opinion may finally be ripe for forcing such a thing to be secure and user-controlled, not just convenient.

"One of the reasons people used to watch TV in the 1950's and 60's was for the shared experience."

The article also talks a bit about how if we don't care when shows are on, there's no prime time. Pop culture is already splintering a bit. People's subcultures and temporary communities are more vibrant than ever, due to changed attitudes and better communications technology. I guess they'll have to substitute quality for proximity. Friends wouldn't have existed without prime time, but South Park was a hit before it was a show. It wasn't until after the film came out that I finally saw an episode of South Park as it was being broadcast, rather than replayed from a friend's tape or hard drive. I'm not worried about the lack of market push destroying all shared media experiences. Besides, the new identity isn't nationalism. I'll share culture with those I share culture with, and that's the community I call home.

13 Aug 2000[link to here]

I got tired of web sites miniaturizing all the text I'm supposed to read. Making subscripts or footnotes tiny is one thing, but entire articles? Since publishers can't be trusted with font size control, I'm taking it away. No text sizing for you! (Yay, FilterProxy! Power to the people!)

Steve asks:

It's a poorly-kept secret that cassettes cost more to produce nowadays than CDs. So, why have CDs remained more expensive?

Probably because people like them better and so are willing to pay more. We don't need a big conspiracy to explain it. It's just a matter of the music commodities market. There isn't one.

There's only one producer of the latest DJ Hype album, and they can set whatever wholesale price they like. Retail competition will squeeze retailers' margins, but the only way for competition to affect wholesale prices is for another producer to market a competing product. Music artists do compete in style and marketing, but when consumers decide which CD to buy, I think they care much more about what's on it than whether one costs a little less.

If it cost a lot less, it might matter. People buy cheap used CDs whose marketing has lost its edge or whose catchy jingle on the radio was supplanted weeks ago. But if a 50% price drop is needed to overcome the competition's heavily marketed product, the producers are probably better off sticking with their high prices.

What they did do, contractually outlawing using their products as loss leaders, is a bit rude, and it's reasonable to call that price fixing. But this isn't the sort of behavior that makes us upset. People just read the headline and think, "Yeah, they should be cheaper!". I agree that they should be cheaper. I don't buy many new CDs anymore. But I don't think price fixing is the problem.

On a related note, Mojo Nation is set up to facilitate a content distribution economy, but is designed around automatic redistribution, so if the market operates efficiently, costs will always drop to match the costs of distribution, and the price of the content will always drop to zero. It's designed around a lack of proprietary content and is a nice illustration for what happen when obstacles are minimized and artificial monopolies are removed.

Yesterday I hit two sites with bad link colors. On one, the link text was hard to read against the background color. On the other, the links didn't change color after I'd visited them. I really need to hack together a color-fixer.

Lots of sites are hard to look at, but only a few are nice to listen to. I've been working to the streaming beats of repruhzent. It's not so much that I need fresh music; I own a ton of music I've only listened to once or twice. But it's nice to have a good long mix without having to deal with setting it all up. My thesis toys run fast enough that they could keep up with the music, so with the proper glue, I could be constantly reassured that the stream hadn't switched over to classical choral without me noticing. I do feel a bit decadent sucking 128kbps across the Atlantic just for some music.

14 Aug 2000[link to here]

More audio mechanics wackiness from that great book:

In some forms of tinnitus (ringing in the ears) the ringing can be picked up with a sensitive microphone placed near the ear.

15 Aug 2000[link to here]

Oh, the irony.. From the SJ Mercury News:

A mountain lion was first observed at 7:30 p.m. Sunday by residents living on Antelope and Lynx drives, according to police.

16 Aug 2000[link to here]

I couldn't think of a nice way to print lines 101-200 from a file using the usual unix tools, so I wrote torso (to join the ranks of head and tail).

As Lyn points out, it's just head -200 | tail -100. I knew there was an easy way to do it, but my brain exploded a few days ago and I couldn't think of it. Luckily torso has other redeeming features so, although I feel like an idiot, at least I have a nifty new command.

17 Aug 2000[link to here]

Recipe for mood swings:

  1. Run final tests of two months' work.
  2. Notice horrible performance.
  3. Triple check everything, find no procedural errors.
  4. Form untestable but plausible hypothesis.
  5. Mope.
  6. Test one last thing, which the hypothesis will fail to explain.
  7. Tear hair out.
  8. Take a break, but don't relax.
  9. Lather, rinse, tear, repeat until inspiration strikes.
  10. Rejoice.
  11. Kick self over minute procedural error you cautioned others against months ago.
  12. Eagerly anticipate expected proof.
  13. Test hypothesis.
  14. Rejoice!
  15. Madly try to make up for lost days.

21 Aug 2000[link to here]

Lilly mentions the phenomenon of using something only for its original purpose.

"I can't use that; that's the silicone spatula for hot stuff"

A friend of mine has a good term for that: "functionally fixed". I like having a term for it because it encourages me to notice when I'm behaving that way. For example, pens aren't just for writing, they're also pointy things that are useful in a wide variety of pointy situations. Children are good at not being functionally fixed. They don't have years of unimaginative precedence and often aren't burdened with the knowledge of what things are intended to be used for.

23 Aug 2000[link to here]

Need muse bait.

25 Aug 2000[link to here]

eGroups is nice enough to mark their ads clearly. This procmail recipe strips them out:

:0 fb
* ^Received: .*egroups\.com
| perl -ne '$ad = 1 if (!$ad && (/-{25} eGroups Sponsor -{20}-*~-~>/ || /-{65}<e\|-/));print unless $ad;$ad = 0 if ($ad && (/-{65}_->/ || /-{65}\|e>-/));'

26 Aug 2000[link to here]

As though I didn't have enough to do, I failed to fend off a burst of graphical creativity and I made the cute little icon now featured above. It doesn't link to anything because you're already here.

27 Aug 2000[link to here]

I like the word "tonsorial". It sounds like it should be a combination of "tonsil" and "sensorial", maybe for describing the silky feel of a fine chocolate being savored as it melts and coats your entire mouth. Think about that feeling for a minute. Isn't it wonderful? But it actually means "having to do with barbers". Gurk! The image of luscious luxury is jarringly replaced with one of hair, and there's a brief moment where they mix in an anatomical blur and leave you with that icky feeling of a stray hair stuck to your soft palate. That's why I like that word.

31 Aug 2000[link to here]

Six years ago, Karl Sims bred robot bodies and controllers together in simulated evolution. There are pictures, movies, and info on his site and a fan site. Some other folks recently did pretty much the same thing and dumped the output to an industrial fabrication machine. There's nothing really new here, but it's always more fun and more impressive to see something done for real instead of just in simulation. Also, as roboticists love to say, reality is always more realistic (and more complicated) than the simulator. As usual, the NY Times article is pretty good.