What components are involved in learning what may likely work for somewhat similar tasks?