In plain English
Transfer learning reuses a model trained on one task as the starting point for a different but related task, rather than starting from scratch. The model brings general knowledge it already learned, so adapting it needs far less data and compute. It is the reason a single foundation model can be specialised into many different applications.
Why it matters
Transfer learning is why modern AI is affordable to apply: organisations adapt existing capable models instead of training enormous ones from zero.
A worked example
A model that already understands general English is adapted with a small medical dataset to read clinical notes, instead of being trained on language from nothing.
Common confusion
Transfer learning is the general idea of reusing learned knowledge; fine-tuning is one specific way of doing it.