Artificial intelligence or the computing power of computers to help humans surrounded by coronavirus. Professor of "Technology Management" at the INSEAD business school, Theodoros Evgeniou recently signed an article noticed in the Harvard Business Review on the role of artificial intelligence in the battle against Covid-19 and government strategies for deconfinement. Interview with an international specialist in algorithm prediction.
How could artificial intelligence have predicted the extent of the Covid-19 pandemic?
THEODOROS EVGENIOU. The mathematical models of epidemiologists are often used to predict the emergence of diseases worldwide. This is perhaps one of the rare cases where AI and the data could not have predicted the arrival of this massive epidemic a year in advance.
But AI can now help manage the different aspects of this pandemic. It started with the very rapid sequencing of the DNA of the virus or the creation, in February, of new diagnoses of the disease by scanners for example. Finally, technologies based on artificial intelligence, such as digital tracing or rapid screening tests, have proven themselves in Taiwan or South Korea.
A more exotic project has started in an engineering school in Switzerland, where a computer with "Machine Learning" can detect the virus depending on how a patient coughs and the sound it emits. It is not 100% safe but, combined with other technologies, it provides additional information about the disease.
How is AI used by health authorities?
It accelerates the analyzes of infected patients, but not only. AI also provides information that guides political decisions.
Newsletter - The essentials of the newsEvery morning, the news seen by Le Parisien
Your email address is collected by Le Parisien to allow you to receive our news and commercial offers. Find out more
The algorithms used by governments and health agencies take into account data, such as the medical record or the places that patients have visited, and distill the most important information of a citizen or a community to keep only the essential. Data is useless unless it influences a decision. This may be which drug route to explore or who is most likely to be infected.
Sequencing the DNA of patients also helps to determine who is most vulnerable to Covid-19. This information will indicate to the authorities which type of population should be protected as a priority and placed in isolation. In South Korea or Singapore, this teaching made it possible to quarantine the people who were most likely to be contagious and to transmit the virus.
Analysis of the data can also determine which profiles of infected people are more likely to die or suffer the most severe symptoms, and thus encourage the authorities to protect them as a priority.
What can be the role of analyzing data from a population in building collective immunity?
Collective immunity is the result of a political strategy and the only alternative to mass screening and total suppression of the virus in the population. Artificial intelligence can help do this, in the most secure way, by rationalizing it.
First, you want to make sure that the first infected are the least likely to develop severe forms of the disease. We already know that this corresponds to 99% of the population who will be either asymptomatic or with a milder version of the disease. The most fragile should therefore be isolated for a while in order to be the last in contact with the virus.
This part of the population would then have few opportunities to contract the disease, since it would almost no longer circulate in an immune population. We don't yet know who the 99% least risk is, but we already have the data to predict the remaining 1% who should be isolated and protected longer. But it is up to governments to choose this tactic or not.
How can AI help in the next phase of deconfinement?
With David R. Hardoon and Anton Ovchinnikov, we explain in our article on deconfinement that governments make bets with two options on the table, based on mathematical models of prediction whose effectiveness it is impossible to predict at this moment.
The most optimistic is a scenario, like in Taiwan, where there are not too many new cases and where the patients are easily identified and isolated. This involves real-time testing to keep the infection rate up to date and monitor it. So use a technology that analyzes population movements, test results and data continuously. If the infection rate rises too quickly, then consider re-containment.
The other strategy, which I think is the most likely, is to look for a form of collective immunity where the processing of data can help the government decide on a prudent relaxation of the constraints according to the risks for the population. . AI can predict whether there is an urgent need to protect based on, for example, chronic diseases like diabetes or hypertension, age or gender.
For this there are risk assessment instruments with models created by MIT or at the University of Oxford, in which it suffices to enter databases to establish forecasts. It is like when a bank determines a rating level for a loan. The more precise this risk assessment, the more careful and effective the deconfinement.