Learning Abstract Representations through Lossy Compression of Multi-Modal Signals
This work addresses the challenge of forming abstract representations to facilitate generalization in multi-modal AI systems, presenting a novel method but with incremental impact on the field.
The paper tackles the problem of learning abstract representations for open-ended learning by treating it as a lossy compression of multi-modal signals, showing that this approach naturally extracts representations that retain shared information across modalities while discarding modality-specific details.
A key competence for open-ended learning is the formation of increasingly abstract representations useful for driving complex behavior. Abstract representations ignore specific details and facilitate generalization. Here we consider the learning of abstract representations in a multi-modal setting with two or more input modalities. We treat the problem as a lossy compression problem and show that generic lossy compression of multimodal sensory input naturally extracts abstract representations that tend to strip away modalitiy specific details and preferentially retain information that is shared across the different modalities. Furthermore, we propose an architecture to learn abstract representations by identifying and retaining only the information that is shared across multiple modalities while discarding any modality specific information.