Luís Correia, LASIGE’s integrated member, was one of the speakers of the final workshop at the Gulbenkian’s Programme New Talents in Artificial Intelligence (AI), 4th edition, last September 9, 2021.

His presentation “Artificial Intelligence – the importance of knowledge context”, has begun with the approach to the debate about the possibility of having, since the field of AI emerged, intelligent machines.

It was also expressed that it is evident that AI has had resounding results in specific areas (games, data mining, etc.), but each of these systems is limited to the specific problem that was created. A central aspect of these AI limitations is the poor articulation between symbolic knowledge and reasoning and also between the knowledge-based on data and uncertainty. Currently, this is perhaps the point where there is a greater expectation about the possibility of significant advances in AI. However, AI has always offered several examples where these two types of knowledge are articulated, from research problems in state space to recent machine learning techniques such as reinforcement learning. The most recent models and more promising ones also show this articulation, still in a primeval phase compared to natural intelligence.