This approach relies on computers the same way a weather service does, to simulate digitally systems that are not in themselves digital. It excels in finding various sort of similarities among patterns, recognizing repeated (or almost repeated) patterns, and filling in missing parts of incomplete patterns.
NFAI typically learns from examples (but not in the same way humans do). Their architectures are inspired by the structure of the brain, but more deeply, by the importance and ubiquity of non-formal pattern reasoning.
NFAI is a very grab-bag term, it practically means anything that isn’t GOFAI.