By and large, the professional handling of huge data collections is regarded as a fundamental ingredient of the progress of machine learning and of its spectacular results in related disciplines. While the recent debate on how rogue AI may arise seems to be quite controversial because of influential different positions, there is a growing agreement on risks connected to the centralization of huge data collections. In this talk I sustains the position that the time has come for thinking of a new learning protocol where machines conquer cognitive skills in a truly human-like context that is characterized by environmental interactions. This comes with a specific restriction on the protocol where the agent is supposed to gain those cognitive skills without storing the information acquired from the environment. While data is processed at a certain instant with the purpose of contributing to update the current internal representation of the environment, the agent is not giving the privilege of recording the temporal stream. Basically, there is neither the permission to store the temporal information coming from the sensors nor their internal representations. Hence, no data collection can be created from machine environmental interactions.
I argue that facing this challenge requires new foundations on computational processes of learning and reasoning that might open the doors to a truly orthogonal competitive track on AI technologies that avoid data accumulation by design, thus offering a framework which is better suited for privacy issues. I give the big picture of a novel approach to face the proposed challenge which is based on computational laws of learning inspired to principles of theoretical physics.
Join at: imt.lu/conference