By feeding a supercomputer with news stories, it was able to help predict major world events. Great...AI with ESP.
The study used millions of articles that were fed into the SGI Altix supercomputer (known as Nautilus) at the University of Tennessee. With 1024 Intel Nehalem cores, the machine recorded processing power of 8.2 teraflops (a trillion floating point operations per second).
Information was taken from a range of sources, including the US Open Source Centre and BBC Monitoring (both monitor local media output around the world), as well as news outlets that published online versions and archives. With more than 100 million articles, they were analysed for two main types of information: mood – whether the article represented good news or bad news, and location – where events were happening and the location of other participants in the story.
Mood detection searched for words such as "terrible", "horrific" or "nice", while locational geocoding took mentions of specific places and converted them in to coordinates that could be plotted on a map. By analysing story elements, an interconnected web of almost 100 trillion relationships was formed.
Based on specific search queries, Nautilus then generated graph. For the different countries which experienced the "Arab Spring" revolts, in each case the aggregated results showed a notable dip in sentiment ahead of time – both inside the country, and as reported from outside. Aside from results charting ahead of the recent revolutions in Libya and Egypt, the system also picked up early clues about Osama Bin Laden’s location.
In the best case scenario for the future, your computer will order your pizza before you even know you want it. In the worst, well...use your imagination.
The study used millions of articles that were fed into the SGI Altix supercomputer (known as Nautilus) at the University of Tennessee. With 1024 Intel Nehalem cores, the machine recorded processing power of 8.2 teraflops (a trillion floating point operations per second).
Information was taken from a range of sources, including the US Open Source Centre and BBC Monitoring (both monitor local media output around the world), as well as news outlets that published online versions and archives. With more than 100 million articles, they were analysed for two main types of information: mood – whether the article represented good news or bad news, and location – where events were happening and the location of other participants in the story.
Mood detection searched for words such as "terrible", "horrific" or "nice", while locational geocoding took mentions of specific places and converted them in to coordinates that could be plotted on a map. By analysing story elements, an interconnected web of almost 100 trillion relationships was formed.
Based on specific search queries, Nautilus then generated graph. For the different countries which experienced the "Arab Spring" revolts, in each case the aggregated results showed a notable dip in sentiment ahead of time – both inside the country, and as reported from outside. Aside from results charting ahead of the recent revolutions in Libya and Egypt, the system also picked up early clues about Osama Bin Laden’s location.
In the best case scenario for the future, your computer will order your pizza before you even know you want it. In the worst, well...use your imagination.
No comments:
Post a Comment