ia-sante-mondiale-bordeaux-business

Model Covid-19 forecasts with AI

The great strength of artificial intelligence (AI) is raw data analysis (1). In the context of the health crisis, it has been heavily exploited in many ways. Thus, as of 31 December 2019, the emergence of the COVID-19 epidemic was detected by analysing messages on social networks (1). In addition, image analysis allowed AI algorithms to identify new factors. In continuity, it is also able to determine the complications of carriers of the virus. As well as detecting the body temperature of passers-by in public places (1).

AI to predict the unpredictable, a key asset in the event of a pandemic

Artificial intelligence technology has made it possible to put many tools in place. For example, the French Research Institute, Inserm, the world’s second largest biomedical research institute, is making its Clinical Research Navigator platform available in researchers (3). In this context, hundreds of research professionals will be able to use the NRN data for six months. A decision taken by Inserm to support researchers in research against COVID-19.

And for good reason, based on technology using artificial intelligence, the NRC is a key tool. Thus, it combines aggregation of content and decision support to enable healthcare professionals to react quickly. Handling more than 100 million medical documents (4), the platform offers a formidable readability of data. It allows for the most accurate modeling for relevant actions. These include data on clinical trials and patents, as well as content from the Web and social networks (3).

In addition, the use of the data allows AI systems to estimate COVID-19 figures. Life-saving information to anticipate epidemic peaks and hospital overcrowding. To make their predictions, experts can therefore rely on predictive modeling based on internet browsing.

young man doing internet research using artificial intelligence
AI is used on a wide range of subjects. Based on the analysis of recurrent behaviours, it needs consistency and data to interpret.

Artificial intelligence, a technology based on coherence

Although artificial intelligence provides researchers with useful data for crisis management, the system has limitations. The explanation is very simple: AI is based on a history to learn over a long time. Thus, under the Covid-19, innovation does not yet have enough hindsight to be 100% efficient (6). For example, some machines were used when they were designed for other diseases. While their use can provide valuable insights, they do not compress the time required to learn and develop a new AI technology. According to Dr. Hugh Harvey, “it took five years to validate AI systems that can detect breast cancer,” “there is no way that anyone could have done this work for Covid-19 in the last three months” (6).

At the same time, the health crisis has highlighted the difficulties for AI in adapting quickly to changing circumstances. Having had to assimilate a large amount of data that did not correspond to the usual pattern (7), many tools based on this technology were lost. This is an opportunity to remind people that artificial intelligence does not yet relate to human intelligence. It is based on algorithms that exploit the data we want to give it. Which need a certain regularity. Even machine learning tools can’t be brutally resilient. Thus disrupted, the tools based on artificial intelligence were then able to give inconsistent results.

For example, with containment and a net change in the consumer habits of the population,Amazon’s recommendation algorithm has been very annoyed (7). Consequence of an AI trying to do its job outside the scheme it knows.

Sources:

(1) Telecom Paris, “Fighting Covid-19: The Limits of Artificial Intelligence,” July 21, 2020

(2) The digital factory, “Artificial intelligence, the 6 major projects of Emmanuel Macron’s strategy” , 29 March 2018

(3) Mydatacompany, “Covid-19: Inserm evaluates an AI-based content aggregation tool”, 9 June 2020

(4) Expert System, “Inserm uses Artificial Intelligence from Expert System to support its researchers in Covid-19 research”, 4 June 2020

(5) Press release

(6) Esanum, “Covid-19, Artificial Intelligence Reaches Its Limits”, May 7, 2020(7) Big Data, “Covid-19: Our Behavior During the Pandemic Makes AI Bugger”, May 15, 2020

Abonnez-vous à la Newsletter