backtop


Print 26 comment(s) - last by Shadowmaster62.. on Feb 5 at 1:11 PM


  (Source: tecflap.com)
The software collects information from news article archives and other data sources to predict the future

Using old news articles and a form of Wikipedia, new software is capable of predicting the future's events.

The software, which was developed by both Microsoft and the Technion-Israel Institute of Technology, takes a look at archives from The New York Times and studies related data on the Internet in order to form predictions about what will happen next in certain parts of the world -- such as disease, violence and a large number of deaths.

The system's sources include The New York Times' archives from 1986-2007, DBpedia (the information in Wikipedia constructed using crowdsourcing), WordNet (helps software understand what words mean) and OpenCyc (provides a database of common knowledge).

What the system does is study news reports, then uses outside data for context. For example, the system saw reports of droughts in Angola in 2006. From studying data on the Web, the system knew that droughts can lead to cholera outbreaks in the country. The system further researches the country's location, population density, GDP, whether there was a drought the year before, proportion of land covered by water, etc.

After collecting said information, and studying yet another report from NYT saying that there were large storms in Angola in early 2007, the system predicted the cholera outbreak. Less than one week later, reports of cholera had appeared.

“I truly view this as a foreshadowing of what’s to come,” said Eric Horvitz, codirector at Microsoft Research who led the study with Technion-Israel Institute of Technology's Kira Radinsky. “Eventually this kind of work will start to have an influence on how things go for people.”

When testing the software, Horvitz and Radinsky found that it was correct between 70 and 90 percent of the time. The team said the software could use some extra work in terms of greater accuracy, but once that is complete, it hopes the system can be used to help organizations tackle world problems.

While some predictive tools are already in use, this particular software uses 90 data sources total, making it a more "general purpose" tool.

Source: MIT Technology Review



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

By kattanna on 2/4/2013 12:28:56 PM , Rating: 2
quote:
Less than one week later, reports of cholera had appeared.


thats a simple chain of cause and effect

whats being described here is the same as if I made a "prediction" that my thumb would hurt after hitting it with a hammer.

or.. I "predict" that after the next school shooting there will be another cry for gun control, since it always happens in that manner, cause..and effect.




By daboom06 on 2/4/2013 12:39:00 PM , Rating: 2
they're building a bridge over some complexity. fortune telling is only mystical because the string of cause and effect on the way to a distant event is extremely complicated. they'll might find online that cholera mutates and becomes super dangerous once every 30 years so when there's a flood coinciding with mutated cholera, the movement of people will be slowed by some amount and civil unrest is likely. all they're doing is extending normal cause and effect a little into the gray area (which is only called gray because not all the causes are known, so to speak).

learn more about which causes matter and what their effects are, and then start charging a dollar for palm readings.


By augiem on 2/4/2013 2:05:42 PM , Rating: 2
Sure, but over time as they gain more data to work with, systems like this may be able to detect deeper and more complex patterns of cause and effect that humans have missed. I'm sure the technology will be applied to all sorts of fields such as market research to spot trends, predict what would sell best at the moment, etc. At some point, to widen the scope, you probably want to plug in many more data sources such as random internet chatter (FB, Twitter), phone conversations and text messaging/email content, TV/movie/music consumption figures, shopping and e-commerce data, job creation/loss, political sentiment and swings, etc. Yes, I know marketing reasearch companies already do all this to a degree, but it would be interesting to see what it could come up with using such a wide variety of data sets.


By ppardee on 2/4/2013 2:10:31 PM , Rating: 2
Everyone predicts the future. You have a model of what you expect the world to do after a cause and effect chain and you'd be startled if it didn't happen. The future is only the effect of causes in the past/present.

You're using your understanding of the past to predict what will happen in the future.. the trick with the computer is that it has the capacity for a ton of information and can more accurately predict what will happen.

After the next school shooting, there will be a cry to disarm the people, which will be successful and 15 years later we will have a dictator that sends people to concentration camps and there won't be anything we can do about it because we have been rendered impotent by fear-mongering fools.

This is the kind of prediction we're looking for. It will take the most likely scenario, then the most likely scenarios after that and build a very long complex chain of events. It is no different than weather forecasting... the farther out you go, the less accurate they will be. It doesn't make it not predicting the future because you can do it on a small scale.


"It seems as though my state-funded math degree has failed me. Let the lashings commence." -- DailyTech Editor-in-Chief Kristopher Kubicki

Related Articles
Microsoft Predicts the Death of Blu-ray
September 23, 2010, 10:33 AM













botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki