Artificial intelligence (AI) is slowly but surely breaching the mist of the hype cycle, and becoming an integrated part of many aspects of our work and home lives. In 2020 for example, the same health issue was on everyone’s minds: the coronavirus. Which prompted //next columnist Markus Sekulla to ask, “So why not develop AI that supports the fight against COVID-19?” Thankfully, scientists were already on the case.
The coronavirus has the entire world in its grip. What makes the virus particularly dangerous is that victims become contagious before any symptoms appear. So although we feel fine and can go about our daily lives in the initial phase, we could already be transmitting the virus. So much from me, the non-epidemiologist. Regularly testing people who aren’t symptomatic does not seem to be feasible. Yet thankfully, a solution to this problem may already be at hand, and it is both simple and efficient.
Researchers at the Massachusetts Institute of Technology (MIT) have developed AI that can recognise, at an early stage of the illness, whether a person has been infected with the virus or not, just based on their cough. The team recorded 200,000 forced coughs from people, of whom 2,500 were proven to have been infected with COVID-19 – both symptomatic and asymptomatic. And the result: a 98.5% probability of correctly identifying a person with COVID-19 symptoms, and even a 100% probability for asymptomatic infections. And what's more, the application works by telephone, which is important in these #stayhome times.
In summary, the test-without-test method is ...
If you think that this is not a useful application of AI, please leave a comment below and tell us what is. For me, this is one of the best AI use cases of the year. And the details of the study are also interesting, since vocal chords can apparently communicate a lot more than just words (or even coronavirus infections). Computers can easily detect people’s sentiment, and other demographic characteristics, and analyse them for medical purposes.
For more information, please see the >>> press release from MIT.
Text: Markus Sekulla