Predicting Disease Outbreaks With AI
Virus control has been all over the media in recent weeks, specifically regarding coronavirus, its impacts on global society and the economic consequences. The Centers for Disease Control and Prevention (CDC) and the World Health Organization (WHO) have been releasing alerts and providing close to real-time updates on the spread both within the United States as well as internationally. But before those alerts and warnings went out a Canadian artificial intelligence (AI) organization, BlueDot, sent out the first warnings of the virus.
So how did BlueDot ‘blow the whistle’ on this outbreak before highly renowned programs such as the CDC and the WHO?
AI.
Essentially, they created a program that is able to process data, fast. They gather news postings (in several different languages), airline ticketing data, and even reports from plant and animal disease tracking. Once all this data is compiled, an algorithm is created that seems to be better at predicting or simulating the spread of disease for its private and government customers.
So what makes this model better?
Traditionally, tracking was done by recording when and where people would come in contact with and contract a given disease to help identify a source of the outbreak and then determine which regions or populations are most at risk for contamination. The system BlueDot designed works a bit differently, it is able to work similar to how a future cast in weather works but shows how the given disease is spreading through a population. This makes it possible to estimate how far the disease will spread, how fast, and where it originated.
How accurate is the algorithm?
Like most AI technologies, it’s all about the quality of the data it is being ‘fed,’ just like traditional forecasting methodologies. It uses structured data points such as the location of a given outbreak, the number of reported cases withing a given time frame, etc. With the rise of AI, the data points have expanded. With the use of social media sites like Facebook and Twitter, more and more data can be collected, but this is considered unstructured in nature. An example might be a conversation group or forum on Facebook where users are talking about symptoms.
The issue with social media data and predicting disease outbreaks is you may miss entire regions who have limited or no internet access. This could mean the models are making flawed assumptions or forecasts.
Is AI the future of stopping a disease from progressing or becoming a pandemic?
Yes and no. AI seems to show great promise in a lot of industries, even in disease control and prevention, but it’s nowhere near perfect. Even if the technology was perfect, it still poses some ethical issues. When using computerized models, we pose the risk of eliminating entire demographics and creating a false positive as far as a forecast goes. There’s no doubt that AI technology is an extremely powerful tool, but it doesn’t seem likely it will replace the traditional combination of statistical analysis and epidemiology anytime soon.
Here at Extract, it is our priority to utilize and harness technology to safeguard information and allow employees freedom from manual and repetitive workflow tasks. Our automated redaction software uses machine learning techniques to read your documents like a human would, and find the information you need redacted. If you’d like to learn more about how we can help you automate your redaction or indexing needs, please reach out today.
Sources Used:
https://gcn.com/articles/2020/03/04/coronavirus-ai.aspx
https://www.nextgov.com/ideas/2020/03/predicting-coronavirus-outbreak-how-ai-connects-dots-warn-about-disease-threats/163475/
https://www.cdc.gov/coronavirus/2019-ncov/index.html
