How the United States spies on us all

curtiss_leung at ibi.com curtiss_leung at ibi.com
Fri Jan 8 09:26:34 PST 1999


John Kawakami writes:


> Put another computer on the job, and you can scan those recordings for
> key words, like "hash" or "got the stuff" or "beat up". You can monitor
> conversations in real-time these days. I'll assume that PC can handle
> the full load, working 24 hours a day. This computer is responsible for
> flagging potentially important messages and copying them for a monitoring
> employee.
> Let's say that there are N interesting messages flagged per day, per 70
> people. Assume messages are 5 minutes long, or 12 messages per person
> per day.
> Let's say that there are N interesting messages flagged per day, per 70
> people. Assume messages are 5 minutes long, or 12 messages per person
> per day.
> One agent, working 8 hours monitoring people, can monitor 72 messages.

Assuming the computer could handle the processing load you describe (storage being another matter -- see below), I think you're underestimating greatly the time required for analysis. For someone working 8 hours to analyze 72 messages, they have only just under 7 minutes for each message, or less than two minutes for analysis once playback time is subtracted. Now consider that people in talking on the phone won't be completely explicit when referring to people, places, and events (I'm more likely to say "Remember the fucked up thing that happened a couple of weeks ago with Mary and whats-his-name?" instead of "Remember when Joe Smith insulted Mary Jones in our company at the Brass Rail on December 17th?") and I think you'll agree that analysis has to be a very time consuming process; analysis times will likely be at least twice the length of the message itself, and that's probably a best, not an average, case.

Another problem that comes to mind is the storage capacity needed for surveillance on this scale. Speech falls within a narrower bandwidth than the sounds in general so some savings could be achieved there, and data compression could help more, but I have the sneaking suspicion that the amount of opitical or magnetic storage needed to hold a large city's voice messages over the course of a month would probably be some fantasically huge number.

-- Curtiss Leung



More information about the lbo-talk mailing list