When the 9/11 commission reported back on the intelligence failures that led to the attacks 16 years ago, they identified a key problem: America's spy agencies had collected so much useless, indiscriminate information (haystacks) that they couldn't find the useful, salient facts (needles) buried there.
But there is no procurement cycle for collecting less data. No one will pay Peter Thiel and the other Palantir investors for convincing their government customers to end dragnet spying and constrain themselves to Fourth-Amendment-compliant, lawful surveillance where there is particularized suspicion. No minor DoD functionary can write a million-dollar purchase order for less surveillance at taxpayers' expense and walk into a sweet job for the company that cashed the check in a year or two.
So the haystacks get bigger, and the needles keep getting harder to spot.
Last week, CIA deputy director for technology Dawn Meyerriecks announced that the Agency has involved itself in 137 AI projects, often co-developing them with Silicon Valley vendors who stand to make enormous profits by selling needle-identifying tools to complement the titanic haystacks other Silicon Valley vendors have helped the Agency to amass.
At a DC summit for other spy chiefs, we learned that the NSA and other US spy agencies are doing the same: shelling out large amounts of money to create unaccountable, fault-prone systems of algorithmic guilt, which expand upon the CIA's existing, notorious untouched-by-human-hands drone killing ("we kill based on metadata") and the no-fly list with its thousands of names of suspected terrorists (a list that has included babies and senior US Senators alike).
One contractor, Chris Hurst who stands to enrich his employer, Stabilitas, through this program, celebrates the fact that automating analysis will allow spy agencies to do more spying, buying more data-gathering tools from companies like his, begetting more contracts for more AI to analyze it.
The irony, of course, is that while AI isn't intelligent in any sense of the word, the corporations who've lobbied their way into billions' worth of procurements do constitute autonomous, potentially immortal artificial life forms, ones that colonize their human gut flora with an idea called "enhancing shareholder value" that causes them to act in service to the colony organism at the expense of their own species.
The challenge, US officials said, is gaining trust from the "consumers" of their intelligence product—like policy makers, the White House and top generals—to trust reports that have a significant AI component.
"We produce a presidential daily brief. We have to have really, really good evidence for why we reach the conclusions that we do," said Meyerriecks.
"You can't go to leadership and make a recommendation based on a process that no one understands."
Data swamped US spy agencies put hopes on artificial intelligence [Paul Handley/Phys.org]