On January 9, the World Health Organization announced the outbreak of the coronavirus in China. As of January 27, the official number of people infected is at least 2,700 and the death toll is ten. However, the CDC, the U.S. disease control and prevention center, released information about the outbreak of the dangerous 2019-nCov a few days earlier, on January 6. Even more surprising is the Canadian Health Monitoring Platform's reported coronavirus outbreak on December 31, 2019. But how did you do it?
How does artificial intelligence predict epidemics?
We often talk about how scientists around the world are concerned about the growing risk of an epidemic of a particular infection. Climate change is making the planet warmer, and insects such as mosquitoes that transmit malaria and Zika fever can move to new regions where it was previously much colder. However, mosquitoes are not the only threat. Scientists have recently discovered a number of old viruses that have been trapped in ice for 15,000 years. Meat consumption in turn increases the number of animals on farms, which means that viruses can mutate constantly.
Read more interesting articles about developments in the world of popular science and high technology on our channel in Yandex.Zen
For this reason, scientists today use the AI-driven BlueDot algorithm, which displays news about animal and plant diseases in foreign languages, as well as official statements to warn the public in advance not to visit dangerous areas such as Wuhan. In the case of the corona virus, the experience of 2003 is important – during the SARS outbreak when the Chinese government hid information about the epidemic for a long time. And the speed of reaction when an infectious disease breaks out is extremely important.
According to the Wired publication, BlueDot was launched in 2014 and received $ 9.4 million in risk funding. The company currently employs 40 people – doctors and programmers – who are developing an analytical disease monitoring program that uses natural language processing and machine learning methods to display messages in 65 languages as well as flight data and animal disease outbreaks. The developers make it clear that the algorithm does not take into account publications on social networks because this data is too confusing.
Do you think this algorithm can successfully predict the spread of infections in the future? Let's discuss AI skills in our telegram chat and in the comments to this article.
After the automated data selection has been completed, the results of the algorithm are checked by epidemiologists. The report is then sent to public health officials in dozens of countries (including the United States and Canada), airlines, and hospitals that may contain infected patients. If spread, the BlueDot algorithm could correctly predict that the Wuhan corona virus would first appear in Bangkok, then in Seoul, Taipei, and Tokyo. In addition, the algorithm has successfully predicted an outbreak of the Zika virus in South Florida in the past.