Discovering containment: from infants to machines
This addresses the challenge of unsupervised concept learning in AI, which could improve machine learning by mimicking infant development, but it is incremental as it builds on existing cognitive science.
The paper tackles the problem of how infants learn complex concepts like containment without supervision, contrasting with current supervised AI models, and aims to understand the trajectory of early learning.
Current artificial learning systems can recognize thousands of visual categories, or play Go at a champion"s level, but cannot explain infants learning, in particular the ability to learn complex concepts without guidance, in a specific order. A notable example is the category of 'containers' and the notion of containment, one of the earliest spatial relations to be learned, starting already at 2.5 months, and preceding other common relations (e.g., support). Such spontaneous unsupervised learning stands in contrast with current highly successful computational models, which learn in a supervised manner, that is, by using large data sets of labeled examples. How can meaningful concepts be learned without guidance, and what determines the trajectory of infant learning, making some notions appear consistently earlier than others?