From trivial representations to learning concepts in AI by exploiting unique data

The prospect of an AI-based revolution and its socio-economic benefits is tantalising. We want to believe that AI learns higher level concepts, but this is far from reality. Particularly in unstructured data such as images, AI extracts rather trivial notions even when provided with millions of examples. The solution of “the more plentiful and diverse the data the merrier” will not address this. Beyond the privacy and cost implications of amassing data in key applications (e.g. healthcare), some events (e.g. disease) can be rare or truly unique. However, the data inefficiency of modern AI also manifests in poorly leveraging the goldmine of information present in unique and rare data. The key research question we aim to answer is Why does AI struggle with concepts and what is the role of unique data? We propose to move away from trivial learning to a formal representation and learning of concepts that overhauls machine learning precisely by leveraging unique data. This paradigm shift will create unprecedented rewards compared to the current status quo