Coronavirus: but what is artificial intelligence?

SAN FRANCISCO | The question returns regularly with the authorities, and of the giants of the technology : has-t-it is not of computer tools for data analysis that could allow us to better anticipate, manage and address the pandemic of sars coronavirus ?

Last December 30, the researchers gave the alarm. Through algorithms that scan the web, they had identified cases of pneumonia unusual in the vicinity of a market in Wuhan, in central China.

But it took several days before the world health Organization (WHO) publishes a risk assessment. And a whole month before she declares a public health emergency of international concern.

“The power of these tools exist, but we have not yet really found out how to use it,” says Michael Greeley, a co-founder of Flare Capital Partners, which invests in particular in the health technologies.

The tools of artificial intelligence thrive on data. They can comb through texts, numbers, internet sites and images, and to establish correlations that are invisible to the human eye.

Several companies have seen signs indicating a fault at the end of December.

Dataminr, which specializes in the detection of risks in real time, had made the connection between the testimonies of the inhabitants of Wuhan on the social networks, images, disinfection of the fish market of the city and a warning to a doctor who later died of the coronavirus.

BlueDot, in Canada, had also spotted the problem health thanks to an algorithm through the review of hundreds of thousands of news articles each day and data traffic.

“The data were in response to the SARS outbreak in 17 years ago, but we didn’t know at what point this virus would be contagious”, explains to the AFP Kamran Khan, founder of BlueDot, a young shoot canadian specialized in the artificial intelligence (AI).

“We tend to adopt a posture of reaction in these situations, it is human nature,” says this professor of public health at the university of Toronto.


The authorities can use the AI to identify the upstream signals, to track the spread of the virus and thus anticipate the needs of hospitals, analyze the studies and statistics available and to accelerate the search for treatments and a possible vaccine.

There is also a need of good data, in sufficient quantity, and accessible in real time.

The personal information public, and especially private, represent a windfall of resources to exploit in the event of an outbreak. But their privacy is protected by law.

“In China, as they have fewer constraints, they have relied heavily on the geolocation of the people [via smartphones] and their monitoring system. It is more complicated in a democracy,” says Andrew Kress, the founder of HealthVerity, a company specialized in the technologies of extraction of medical data.

“We know the age, gender, geographic area, and some events, such as a purchase of cough syrup, or a diagnosis for a respiratory infection”, details there.

“In theory, one could link these events to the online activity, but the person would become quickly identifiable. There is a balance to be struck between the utility (of info) and the protection of the private life,” he says.

In the current phase of containment, companies are helping governments to follow the progression of the disease and to check that the population applies the instructions of “social distancing”.

BlueDot uses including data “anonymized” location “close to 400 million smartphones”.

The algorithms are also used in medical research.

Last week, researchers and companies have gathered over 29 000 scientific publications on the COVID-19.

The White House has called on the volunteers to quickly retrieve factual information, thanks to the tools provided by Kaggle, a platform for “machine learning” (learning machine) belonging to Google.

On the side of treatment, “the AI allows to conduct simulations and to understand the evolution of biological as never before,” says Michael Greeley.

The time and effort to test manually, one by one, the reactions of the virus. “It takes millions of samples and the machine identifies a handful of molecules that may work for a vaccine, for example”, he continues.

But the phases of clinical tests on patients are they, incompressible, because they determine if the drugs are risk-free for humans, and if they are effective.

At least a relaxation of the regulations. “There is a pressure extraordinary on the pharmaceutical industry in this time to use some tools that are perhaps not yet fully developed”, says the investor.

Share Button