hoos

(online)

Justin Jacoby Smith is an organizer, web geek, Buddhist, and poet.

Goog-411, Privacy & Artificial Intelligence (Google is Skynet)

Did you ever use Goog-411? I did. A lot. Google's short-lived directory assistance was easy to use, faster than my carrier's 411 operators, and best of all: totally free. When I got lost on the way to the new Chinese restaurant off 183 (this was back before phones with GPS, kids), I just called Goog-411--say the name of the place, have it repeated back to me with startling accuracy no matter how mangled my pronunciation, and be connected to the restaurant within seconds. I even remember thinking at the time "why can't I use this to dictate stories yet?"

I was really disappointed when Google abruptly decided to end the service, and I remember being kind of puzzled at hearing why--

"The reason we really did it is because we need to build a great speech-to-text model that we can use for all kinds of different things ..." she said.

"The speech recognition experts that we have say: If you want us to build a really robust speech model, we need a lot of phonemes, which is a syllable as spoken by a particular voice with a particular intonation. So we need a lot of people talking, saying things so that we can ultimately train off of that."

Goog-411 was never a public service--it was a creative way to get millions of volunteers to train Google's speech recognition engine. Now, with one of the largest pools of voice data in human history, Google's voice recognition is among the best in the world.

Cohen says that building just one part of the speech-recognition system required "roughly 70 CPU-years" of computer time. Google's cloud of processors can do that amount of crunching in a single day. "This is one of the things that brought me to Google," Cohen says. "We can now iterate much more quickly, experiment much more quickly, to train these enormous models and see what works."

While Google was never secretive about the real purposes behind Goog-411, they weren't fully up front either: as a regular user of the service I had no idea that my voice was being added to a massive database to teach & improve their new software. Yes, the technology has been put to use in increasingly impressive ways--Google Voice Search doesn't feel that far from the Star Trek computer--but the reuse of customers' data in ways they weren't made aware of speaks volumes about Google's priorities. Privacy & full human disclosure will always come second to their desire to strengthen the machine.

The Slate article points out that speech recognition is just one of Google's many projects working toward applied artificial intelligence. With the global data cloud at its disposal, I can't help but think that the first enlightened machine is gonna wink to life soon somewhere deep in a Google data center.