Take a look at this article of Marianne Seem. "To date, there is no evidence of massive overall job losses caused by AI that some observers tend to anticipate based on studies (e.g. Frey and Osborne, 2017) that assess only the technological substitution potential of current jobs. It remains to be seen how far the previous experience of computer diffusion since the 1970s, which changed job tasks and affected various skill groups differently, but did not lead to higher aggregate unemployment".
