Discovering containment: from infants to machines

Rapid developments in the field of automated learning have caused a major shift in the approach to the learning of intelligent systems, from explicit instruction to the automatic learning from a large number of labeled examples. Yet, current methods cannot explain infants’ learning, in particular the ability to learn complex concepts without guidance, and the natural order of concepts acquisition. A notable example is the category of 'containers' and the notion of ‘containment’, one of the earliest spatial relations to be learned 1–5 , starting already at 2.5 months 2,5 , and preceding other common relations (e.g., ‘support’ 6 ). Such spontaneous unsupervised learning stands in contrast with current highly successful computational models, which learn in a supervised manner, using large data sets of labeled examples 7,8 . How can meaningful concepts be learned without guidance, and what determines the trajectory of infant learning, making some notions appear consistently earlier than others? We present a model, which explains infants’ capacity of learning the complex concept of ‘containment’ and related concepts by just looking, together with their empirical development trajectory. Learning occurs fast and without guidance, relying only on perceptual processes that are present in the first months of life. Instead of labeled training examples, the system provides its own internal supervision to guide the learning process. We show how the detection of so-called ‘paradoxical occlusion’ provides internal supervision, which guides the system to gradually acquire a range of useful ‘containment’related concepts. The mechanism of using internal implicit supervision is likely to have broad application in other cognitive domains as well as artificial intelligent systems, because it alleviates the need for supplying extensive external supervision, and it can guide the learning process to extract concepts that are meaningful to the observer, even if they are not by themselves obvious, or salient in the input.