Cross posted from my home base.
My friend Steven Kotler, a man of prodigious turnout, recently wrote a fascinating story about a former FBI agent who left the bureau for the ACLU. The agent, Mike German, was a once-politically conservative crack undercover agent, compiling evidence in two of the most successful domestic terrorism cases in history.
What forced German into the arms of the agency's rival? In German's last undercover case -- a conspiracy between natural pals: Islamic fundamentalists and American neo-Nazis -- the bosses tried to use his evidence improperly. He cried foul. They tried to transfer him. He said no. Thereupon began his descent into dreaded Whistlblower status, which if you work for the military or intelligence community means that you are not a patriot adhering to rule of law, but a crazy person. (Hey, it worked for the Soviets!) The spy who was left out in the cold found a new home at the ACLU, where he uses his insider knowledge to hold his former employer's feet to the fire. Surprisingly, German points out the similarity between his unlikely career paths:
After two years on the outside, he decided the best way he could
continue being useful was to join the ACLU. "With the FBI, I saw a lot
of very brave people do a lot of very dangerous things. At the ACLU, I
see the exact same thing--but these people aren't out there with flak
vests and big guns taking on the bad guys. They don't have the weight
of the federal government behind them. They stand alone, unprotected,
saying this will not happen on my watch. I wanted to be a part of that."
Among German's projects at the ACLU is leading the charge against "fusion centers." These are repositories for "information sharing," meaning places where the government tries to collect data on its citizens from public and private sources -- doctors, credit card bills, financial information, shopping records, travel arrangements, gambling habits -- and look for patterns. This is the data mining that we heard so much about just after 9/11. Computers were going to help us find the next twenty hijackers. Or, as Rumsfeld would have put it, data mining would help us know those infamous unknowns. Remember all this? The Pentagon called it Total Information Awareness at first, and, reassuringly,their logo was an eyeball shining a light ray out of a pyramid, with the inscription Scientia Est Potentia. Well, it turned out people didn't like the idea of their scientia as the government's potentia.Total Information Awereness was disbanded, but in name only. Data mining lives on as a key technological holy grail for the intelligence community. Now, after revelations about Bush's illegal warrantless eavesdropping and the revised FISA, data mining's inherent dangers to privacy seem even more notable. Especially because, as German points out, data mining doesn't work:
"The whole data-mining model doesn't work," he says. "We're
sub-contracting to companies who want to solve terrorism with
technology. It's pure snake oil." Explaining this, German likes to
point out that pro football teams use much narrower data sets to scout
rookies than the government does hunting terrorists, but they still
draft a guy like Ryan Leaf. "If this kind of predictive analysis really
worked, these companies would be selling their services in Vegas or on
Wall Street. Vegas and Wall Street aren't buying it--only the government
I did an interview in 2003 with Usama Fayyad, one of the world's data mining experts. Fayad is a data mining booster, with his own consulting company, DigiMine. He was also editor in chief of Data Mining and Knowledge Discovery, the primary technical journal on data mining technology. And he was very candid about the limitations of that technology for intelligence gathering:
UF: That's true, and let me give you a tougher scenario. It's one thing
to rate a credit card transaction for its likelihood of being
fraudulent. But as soon as you start looking for groupings of people,
like terrorist networks, which is what Total Information Awareness
needs to do, you now have an exponential problem. Because if there are
N entities, there's an exponential number in N of possible
subgroupings. You're looking for ten people, you don't know which ten,
among millions, and that's an absolutely astronomical number of
BLVR: That's a whole lot of evidence extracting and link discovering.
UF: Yes, but it's not impossible. It's just very challenging. It's
never been done on this scale. It's like putting a man on the moon. If
they really want to get it done, it will require a lot of resources,
the best people in the country, and so on.
That was in 2003. And German is saying the same thing today. Fayad isn't saying its impossible. He just says it would require a commitment like the Apollo project. Which begs the question or whether those resources wouldn't be better spent on developing human intelligence rather than spying on our citizenry?