Occasionally, the computer jumped and made a foray into science fiction. This time, scientists, with the participation of Microsoft, created a technology that can swallow vast amounts of data in a single view: predict the future, at least to some extent.
Microsoft and researchers at the Technion (Israel Institute of Technology) have united in a very special project. They drew on multiple databases to cut information and find cause and effect relationships between some of them. The New York Times in particular has opened the door to its archives which are spread over 22 years. Other bases have been added, including DBpedia, WordNet and OpenCyc. A total of 90 sources have fueled the algorithms researchers.
But for what? Although the technology is still in its infancy, the tests are encouraging. Solely on the basis of incomplete data, the system was able to predict cholera epidemics that hit in Angola in 2006 and in 2007. In the first case, the system had made the connection between drought and the increase of the bacteria. In the second, it was because of violent storms.
Eric Horvitz, co-director of Microsoft Research, said that through the tests, the system showed that predictions were correct in 70-90% of cases. A high enough score but concerns for now a limited number of events.
However, the results are encouraging for the future because that is where the goal predict some major catastrophic events, whether epidemics, natural and other causes. The idea is that the relief, international aid agencies involved can prepare in advance. Similarly, people could be warned of an increased risk, which would take extra precautions, especially when the events are sanitary type.
For now, the project will continue only on a research plan. No matter today to market the product because many improvements are needed. Accurate results with the climbs of the information provided, and the selected sources were not necessarily designed for such use. Moreover, according to Horvitz, it will be interesting to find data going further back in time.