Coronavirus : Waterloo de l'intelligence artificielle ?

Le mythe de l'intelligence artificielle est-il mort ?

Avant la pandémie, on lisait que "une intelligence artificielle a démontré que Shakespeare n'a pas écrit son oeuvre" ou encore que "une intelligence artificielle a surclassé tel ou tel médecin".

Comment se fait-il qu'alors que l'on a des millions de cas, "big data" !, "une intelligence artificielle" ne nous ait pas dit des choses renversantes sur ce virus, et la façon de s'en protéger ? On en reste à des statistiques primitives, qui montrent des facteurs aggravants, corrélés à la presque totalité des décès.

D'ailleurs, pourquoi n'a-t-on pas été plus rapide dans l'exploitation des données de l'épidémie ?

By Jean-Pol GRANDMONT - Own work, CC BY 2.5, Link

(Un article sur le sujet :

USING AI TO TREAT COVID-19:
Not a day seems to go by without researchers or companies making headlines by releasing new computer models that, they say, will help clinicians better diagnose and treat COVID-19. Less noisy are the experts who are raising doubts and warning that half-baked technology could end up creating more problems that it helps solve.
Bombshell analysis:
Last month, a team of AI researchers analyzed dozens of computer models to diagnose and treat COVID-19. Their review was scathing: All of the models had been trained with unfit and insufficient data, they found — in one case with information collected from just 26 patients, while thousands would have been needed. All of them, they concluded, were unfit to support doctors with potential life-or-death decisions. How is it possible that not a single model stood up to their scrutiny?)