Thursday, October 12, 2006

An online lie detector?

Interesting notion if it pans out, but I suspect politicians will adapt very quickly and just become more adept liars. More likely is that this will be like the AI breakthroughs that Minsky pitched for the last 30 years as being only a few years away. Those few years were always a moving target though.

Don't get me wrong, there's a lot of genuine advances in AI that have happened in the past 20-30 years, but the AI cheerleaders are always making predictions an order of magnitude more optimistic than is warranted. Prudence suggests this instance won't be much different. If they say 5, think 20.

Arnaud de Borchgrave
Google's supremo stunned the British Conservative party's annual conference -- and thousands of politicians in Western democracies -- when he said the next step in cyberspace was the "truth predictor." Within five years, new software will allow voters anywhere to check in real time the probability that political statements are factually correct...

2 comments:

Anonymous said...

Normally, I would say that you should just dismiss it as the diseased ranting of an idealist, but this is the head of Google, one of the largest employers of Extremely Intelligent People in human history. If any company or university stands a chance at making such a project feasible, it would be them. Still, the odds of it working for a very long time are slim to none.

Purple Avenger said...

If they do hit the claimed 5 year mark, what we'd (of necessity) wind up with is some sort of large data mining operation run against "sources" "approved" by Google as "legit".

Given Google's plainly hard left orientation, those "sources" would result not in a a genuine lie detector, but an online legitimizer for propaganda ;->

The ones who control the information, can control the "truth".