Pre-Crime Detector

[sstein]

http://www.newscientist.com/blogs/shortsharpscienc...

So, in tests where people are told to act shify the machine detected people >> acting << shifty, it didn't actually detect anyone who was going to commit a crime...

article said:

Some subjects were told to act shifty, be evasive, deceptive and hostile. And many were detected. "We're still very early on in this research, but it is looking very promising," says DHS science spokesman John Verrico. "We are running at about 78% accuracy on mal-intent detection, and 80% on deception."

So thats a 78% false positive rate... brilliant! So great at detecting actors, unproven to detect terrorists.

Fox's take on it

http://www.foxnews.com/printer_friendly_story/0,35...

burns said:

Burns noted his team's goal is to "restore a sense of freedom." Once MALINTENT is rolled out in airports, it could give us a future where we can once again wander onto planes with super-sized cosmetics and all the bottles of water we can carry — and most importantly without that sense of foreboding that has haunted Americans since Sept. 11.

They just don't get it.

-

Stuart

Edited by sstein on Friday 26th September 21:34

[Ry_B]

78% accuracy meaning 22% false readings surely?

[Jamm]

Sounds like some Minority Report bullst to me.

[sstein]

Ry_B said:

78% accuracy meaning 22% false readings surely?

It means the 22% of people it didnt detect was correct, the 78% were false positives. They were NOT terrorists. No terrorist has ever walked through the machine

The people didn't actually have any mal-intent.

It's like me writing a program that analyses hand-writing and asking you to write a letter while pretending to be a terrorist. If the machine identifies you as a terrorist, that's a failure. You are not a terrorist, you are someone pretending to be a terrorist

The difference may be subtle, but important!

-

Stuart

Edited by sstein on Friday 26th September 21:47

Edited by sstein on Friday 26th September 21:48