Negative Ontology of True Target for Machine Learning: Towards Recognition, Evaluation and Learning under Democratic Supervision
The episode discusses a paper titled 'Negative Ontology of True Target for Machine Learning,' which challenges the assumption that a single correct answer exists in machine learning. The hosts explore how this framework uses 'Democratic Supervision' to allow multiple imperfect inputs to create robust models, ultimately enabling non-experts to participate in and gain understanding from AI systems.