Why Generative Methods Are Still Missing Something to Become Responsible

AI tools, including generative ones, have undeniably taken the world by storm, and their outcomes often leave us in awe. Nevertheless, there are some fundamental issues pertaining to veracity and impact assessment that continue to concern prospective users. This talk comments on Responsible AI in a multi-dimensional approach and delves into the complexities surrounding the apparent oversight of the human factor in these impressive adaptive algorithms. Furthermore, it explores how such a responsible approach can be applied to tackle the challenging task of classifying information, currently utilizing ensembles. The diverse aspects of fake news (and many other hard problems), for example, necessitate much more apt approaches, and the talk analyzes the potential benefits of Responsible AI ideas, specifically its reliability and self-mitigation abilities.