Dept of Future Crime: AI algorithm gives computers ‘hindsight’ to anticipate the future

“If the servers and routing systems of the Internet could see and evaluate all the relevant variables in advance, they could more efficiently prioritize server resource requests, load documents and route visitors to an Internet site, for instance…”

From DATELINE ZERO NEWS: There’s been a lot of research being done to make computers smarter by creating an AI that can learn on it’s own. But Google has become particularly interested in programs that can learn from the fast, and “predict” the future.

Google already has a pretty good finger on the pulse of what’s happening now. Whenever anyone wants to know what’s trending at the moment, for example, there’s Google Trends, Google Discussions search, or Google Realtime. But Google would love to know what’s ahead of the curve; they want to be able to translate data into probability and use that to predict probabilities with greater accuracy.

Researchers are working on algorithms that would allow computers to learn from their past failures or mistakes in an effort to make smarter predictions in the future. Prof. Yishay Mansour of Tel Aviv University, who is heading this new program, refers to it as “minimizing virtual regret”.

Google isn’t alone in wanting such technology. The CIA and DARPA have also been interested in developing technology that would endow the U.S. Government with the ability to predict the future. In the last decade, shortly after 9/11, DARPA went to work on ways to efficiently predict another terror attack. One thing they looked at is how to utilize human intuition as an intelligence asset. The idea they came up with was to create and manage a prediction market; but Congress caught wind of that program and shut it down.

More about that in this related post: Pre-crime: Google, CIA seek the ‘Recorded Future’

Last year both Google and the CIA began backing a new company called Recorded Future. The company says its temporal analytics engine “goes beyond search” by “looking at the ‘invisible links’ between documents that talk about the same, or related, entities and events.”

Basically the 16-person Cambridge, Massachusetts, firm has developed software that combs the public Internet and watches discussions. Much like Web Bot, it looks at discussions of events, and takes note of emotional language. It can learn from discussions of events, before and after the event occurred, and learn from it. According to Wired: “The idea is to figure out for each incident who was involved, where it happened and when it might go down. Recorded Future then plots that chatter, showing online ‘momentum for any given event.”

“The cool thing is, you can actually predict the curve, in many cases,” says company CEO Christopher Ahlberg, a former Swedish Army Ranger with a PhD in computer science. This naturally makes Recorded Future attractive to Google Ventures, the search giant’s investment division, and to In-Q-Tel, which handles similar duties for the CIA and the wider intelligence community.

This month, Google announced that it will be funding another project aimed at developing an AI that can better anticipate the future … Continue reading, from the full article at Dateline Zero.