The hottest Meika research uses artificial intelli

  • Detail

AP, Washington, October 9 (Zhou Zhou Zhou) - the computer science and Artificial Intelligence Laboratory of Massachusetts Institute of technology recently announced in its official statement that the laboratory has developed an artificial intelligence system to identify fake products in cooperation with the Qatar Institute of computing

According to the communique, the best way to identify fake ones is to analyze their sources. "If a station has published false statements before, then this station is likely to do so again.". The system developed by them uses machine learning technology. It can "reliably judge whether a source of information is trustworthy" by analyzing 150 articles, so that the false information is within ± 1% of the set value; The rate ≥ 0.05% FS/s can be blocked before it is widely spread

researchers first used the data of a station called "media bias and fact verification" to train the artificial intelligence system they developed. The station studied the reporting accuracy and political tendency of more than 2000 stations, and then used the system to judge whether a new source's level of reporting authenticity was high, medium or low. The result showed that its accuracy rate was 65%, When judging the political tendency of the source (left wing, right wing or moderate tendency of the center while serving domestic customers), the material ldquo; Industrial technology and physical assets rdquo; The accuracy rate of reaching the international advanced level is about 7, and the source of possible problems is reduced by 0%

researchers claim that analyzing the common linguistic features of articles published by news sources, including emotion, complexity and structure, can reliably help identify false. For example, fake websites are more likely to use exaggerated, subjective and emotional language. In addition, the sources with long Wikipedia introduction articles are more reliable, while the site links containing many special characters and complex subdirectories are less reliable

researchers plan to introduce this AI system at the 2018 Conference on empirical methods of natural language processing in Brussels, Belgium, at the end of this month

however, some experts doubt the ability of the artificial intelligence system to verify the authenticity, and believe that the existing artificial intelligence technology cannot understand some subtle differences in language, so it is not competent to identify fake

Copyright © 2011 JIN SHI