John Poindexter's career has played out in the headlines: Iran-Contra conspirator, the Pentagon's Big Brother in chief, godfather of a futures market to predict terrorism. But there's an alternate reality: The 67-year-old retired admiral is the only serious technologist ever to reach the highest circles of power in Washington. He's a Caltech PhD who two decades ago dragged the White House into the digital age, plugging in everything from fiber-optic video to email. He uses Groove Networks' Workspace to keep in touch with friends and rhapsodizes about encryption like a cypherpunk. In the first interview since Congress forced him to step down last summer as head of Darpa's Information Awareness Office, Poindexter speaks out about privacy, sim terrorists, and Iraqi WMD.
WIRED: What was it like being grilled by Richard Feynman for your PhD?
POINDEXTER: I was scared out of my wits.
What was the topic?
Electronic shielding by closed shells in thulium compounds. I'm afraid it doesn't really translate into English.
Good practice for some of the political buzz saws you've run into since?
It's easy to be a critic. We live in an information society. Corporations and governments have mountains of data that power our economy and give us the highest standard of living in the world. The question is, How do we manage information intelligently to preserve our freedoms, protect our way of life, and advance civilization?
"Knowledge is power" was the IAO's official motto - that spooks a lot of people.
Knowledge is power, for good or evil. The issue is giving goodness the edge. We can't eliminate evil; we can recognize it and try to deter it by ensuring that those doing evil are detected and punished. This applies to the terrorist and to those who would abuse data.
The program's goal was to "revolutionize" the US government's ability to identify terrorists.
You can't take an existing system and dramatically change its capabilities overnight. You start by creating a small, experimental network, running it in parallel with the "normal" system, and then introducing new ideas and capabilities. We had real users from the intelligence community working with a combination of real and synthetic data.
Synthetic data?
It's a little like the Sims - you create a virtual world that has real addresses, real airports, but is populated with imaginary people. We built them by taking a list of all the last names in the country and then adding first names at random. Then we had them take trips. We had a team of a dozen people who came up with scenarios. You introduce terrorists into your world, and then you start looking for ways to pick them out from the data.
And you succeeded?
In a very preliminary way, with a lot of human help, yes, we did.
Your critics never relented on privacy questions.
Advocacy groups want to stay in business, so it's in their interest to paint a dire picture.
Is privacy a right?
It's certainly not a constitutional right. It's an individual right that has to be balanced with concern for the common good. Privacy has to be relative to other objectives - for instance, security. The greatest threat to privacy is terrorism. How much privacy was there in Afghanistan under the Taliban?
Are we managing that balance well today?
Not at all - in a lot of ways we have the worst of both worlds: no security and no privacy. There are at least 50 federal laws and regulations regarding the handling of personal information. Programmers call that spaghetti code.
You were accused of building giant data banks of private information.
Nothing I worked on had to do with collecting data - we have plenty of that in this country, probably more than we need. Our focus was turning it into useful information. You leave it where it is - because of the cost of moving it to a central location, the difficulty of keeping up with technology, and the US citizen's basic distrust of the government.
So how do you persuade people that having the government peer into their lives is a good idea?
Most people don't understand what we were trying to do. Too many opinions are formed based on sound bites from those who yell the loudest. One of the things we were working on was a "privacy appliance" that would conceal a person's identity until a case could be made against them. Congress killed that, too.
The technologies you used include Groove Networks' very trendy collaborative software
You don't collaborate because it's faddish - you do it because there's always ambiguity in the data and you need diverse viewpoints to try to decide what it all really means.
For instance, Iraqi WMD?
That's a perfect example. There were obviously different perspectives, but did they find their way to the decisionmaker? And in such a way that he could understand what the different interpretations were and how they were arrived at? I don't think that happened.
So, a Groove space for the president?
At some point, a US president will be in a Groove space or something comparable, sure. Maybe not this next time, but four or eight years from now we'll elect someone who grew up on the Internet and is more willing to sit at a keyboard and do things on his own.
Al Gore!
God, I hope not.
Contributing editor Spencer Reiss (spencer@upperroad.net) also writes about MIT's new Gehry-designed geek paradise on page 192.
John Poindexter
credit Photo by Michele Asselin
