O futuro da pesquisa em inteligência artificial

Abstract

This paper examines the future of research in artificial intelligence, arguing that, in the near future, symbol-based techniques will continue to receive some attention, while machine learning techniques based on data processing will continue to grow explosively. Models produced through deep learning, in particular language models, will be applied to many sectors and will be significantly enhanced (in efficiency, in interpretability, in performance). Still in the near future, the social debate about artificial intelligence will take a more concrete form, demanding research effort related to regulation, social impact and job markets. Looking into a more distant future, we believe that mixtures between symbol-based and data-based methods will get more attention, while some bets, such as quantum computing, may take us to new performance levels.