Can you explain the concept of transfer learning in AI and provide examples of how it can be applied in different domains?