Words & Stuff

I got curious about why HAL 9000 sings “Daisy” (actual song title: “Daisy Bell) in 2001.

It turns out that it's because Arthur C. Clarke saw a 1962 Bell Labs demo of an IBM 704 singing “Daisy”.

That wasn't the first electronic speech synthesis; the Voder was invented in the 1930s. But the Bell Labs demo may've been the first electronically generated singing.

I wonder what led the Bell people to pick that particular song. I especially wonder whether it's because Daisy's last name in the song is Bell.

YouTube has an audio recording of the demo, accompanied for some reason by a still image made to look like an old movie. The voice doesn't start until a minute in. It's pretty good speech synthesis; not as good as, say, Siri, or Google's synthesized voices, but not as much worse as I would've expected, given that it was fifty years ago.

Speaking of which: in response to the query “sing me a song,” Siri will recite the beginning of the chorus of “Daisy”—but won't actually sing it. So fifty years on, we still don't have singing computers in daily life. Another failure of living in the future, like jetpacks and aircars. I'll have to console myself by listening to the Dictionaraoke version of “Video Killed the Radio Star” again.

(Nitpicky number details: The YouTube video says the demo took place in 1961 (rather than 1962) and was on an IBM 7094 (rather than 704). I tend to believe the Bell Labs official website over a random YouTube video, but I don't know for sure. Also, Wikipedia says that the 7094 wasn't introduced until 1962; if that's true, then the video can't be right about both the year and the model number. But I haven't actually researched any of these numbers.)