This AI Predicted the Pandemic in 2019

By Da Young Kang

Graphic by Senching Hsia

Whenever you hear about artificial intelligence (AI), you may think of it as a brand-new technology or even rocket science. However, studies about AI have been around for decades. Now, these algorithms which can think, reason, and learn like people are being adapted into the field of public health. The most recent and prominent example is “BlueDot” led by Dr. Kamran Khan.

Even before the World Health Organization (WHO) released its warning about COVID-19 to the public on January 5, 2020, BlueDot had predicted the impact of the novel virus on December 31, 2019.

BlueDot is a health monitoring platform in Canada, and its algorithm is based on AI. Established in 2014, the algorithm draws on vast amounts of data to predict the general scale and spread of diseases. Specifically, BlueDot gathers its data from news reports in 65 different languages, current lists of animal and plant diseases, official reports about the diseases released by governments, blogs, and most importantly, global airline ticketing data. However, it does not rely on social networks. According to the research paper submitted by BlueDot, its algorithm analyzed flight data from January to March 2018. By looking at the data on direct and indirect flights from Wuhan, China, and the final destinations of the passengers, BlueDot predicted the cities that would be first affected by the disease. With the analysis of epidemiologists, BlueDot correctly foresaw the virus hitting Seoul, Taipei, Tokyo, and Bangkok first. Additionally, it calculated whether the cities had the capacity to manage the outbreak based on seven factors: “demographic, health care, public health, disease dynamics, politics (domestic), politics (international) and economics.” These conclusions were then reported to “public health officials in 12 countries, airlines, and frontline hospitals.”

Surprisingly, information used to calculate travel patterns from over 4,000 airports is gathered in just a few seconds using two leading technologies: machine learning and natural language processing. These phrases might sound unfamiliar, but we encounter them every day. Machine learning technology in AI collects an extensive amount of data and finds statistical patterns in it. One easily found example is the YouTube algorithm. Its machine learning technology gathers information on your interests based on what your activity and suggests videos that are likely to interest you. On the other hand, natural language processing is a technology that works to understand human language by interpreting and responding in context. Familiar examples include Siri and Google Translate.

Other than predicting the cities that the virus would affect first and calculating their capacities to manage the pandemic, BlueDot also has access to the location data of millions of cell phones. This anonymized data was utilized to follow population movement from Wuhan. Furthermore, by tracking the mobile phone data, BlueDot can tell if people are staying at home and where people still gather the most. Based on fast and accurate conclusions made by BlueDot, Governor Gavin Newsom of California shut down many at-risk cities, thus preventing a catastrophic initial spread in his state; Newsom later testified to BlueDot’s efficiency. Moreover, considering that government information about diseases is often incomplete, public health officials can quickly receive accurate information using BlueDot and avoid relying on their respective governments to plan courses of action. Considering that transmission speed and accuracy of information are key in preparing for outbreaks, Dr. Michael Gardam of Humber River Hospital in Toronto, Canada expresses appreciation for BlueDot. He comments that such AI technology will allow frontline hospitals to be prepared by wearing proper protective equipment and avoiding missing the first case of any given infectious disease.


32 views0 comments