K. Daskalakis | How does Artificial Intelligence fail and what can we do about it?

As the applications of Artificial Intelligence multiply rapidly, it is becoming increasingly apparent that its use involves many risks. AI systems do not have the reliability expected when compared to critical decision-making systems. They are not predictable in new environments; they need huge amounts of data and computer power to grow, and they reproduce stereotypes containing the data on which they are trained.

In his talk at Athens Science Virtual Festival, Konstantinos Daskalakis talks about the reasons that AI fails, analyzing the foundations of Machine Learning. The world-renowned professor of computer science at MIT also reveals how ideas from mathematics and the social sciences can provide solutions to the challenges that have arisen.

The discussion is moderated by Marily Nika, AI Product Manager at Google.