CONCEPT
Going Beyond the Information Given
Bruner's 1957 name for the mind's defining capacity—to infer the unstated, project structure onto fragments, and treat particular instances as cases of something general—which turns out to be the central aspiration of machine learning, achieved in a manner that is simultaneously impressive and fundamentally different from the human version.
“Going beyond the information given” is the title of a 1957 essay by Jerome Bruner and the through-line of his thought: the signature of mind, the thing that distinguishes a consciousness from a recording device. A creature that only registered what was present to its senses would be a camera. A mind takes the input and builds—inferring a whole object from a partial glimpse, projecting a category from a handful of instances, reaching from what is present to what is implied, what is absent, what must be the case. This projective, constructive, leap-making quality of thought was, for Bruner, what made cognition cognitive and what the cognitive revolution had to explain. It is also, by an irony he could not have foreseen, the central technical aspiration of
large language models: a model is trained on a finite corpus and judged by whether it