ATHENS – The progress made in the field of Artificial Intelligence (AI) and the challenges scientists are currently facing dominated a lecture delivered by Constantinos Daskalakis, Professor at MIT’s Electrical Engineering and Computer Science Department, at the Eugenides Foundation on Tuesday evening.
In his speech, Daskalakis explained that AI is largely based on the data available in humanity’s “digital footprint” and the way this is processed by the various algorithms.
However, he referred to a massive “credibility issue” with respect to the technology, arising either from incomplete or unrepresentative data or from a misuse of statistical methods. After all, his research on AI focuses in part on how to avoid adopting the stereotypes and prejudices contained in the data from which the AI learns.
AI’s lack of credibility, however, should not be a deterrent to its use. Instead, Daskalakis is in favour of introducing computer science classes in elementary school. “You have to know the processes that are going on and how the algorithms work. Otherwise you cannot be considered a complete human being and a responsible citizen in the present day,” he said, and added: “That’s why I want to help the general public understand what processes may be behind the technology they use.”
In addition, according to Daskalakis, AI is intended to help people and not become a hindrance to human development. “The person of the future and the present uses technology as an aid. To carry out calculations, which computers are better at doing,” he said, pointing out: “The human brain is one of the most amazing computers. With AI, we want to relieve [the human brain] of trivial tasks and free it to do more creative work.”
Daskalakis explained that science has made great progress in the area of reproducing mental processes, such as voice and image comprehension and game play. After all, more and more people are doing voice searches on their cell phones or using smart assistant devices, “chatting” and choosing what music to listen to or buy online.
Progress in text comprehension, translation and composition is only moderate, however, while the results with respect to long-term planning, transferring knowledge and general intelligence skills, such as those needed for a robot to learn to ski, for example, are disappointing.