Learning to tell if an animal is in pain is most likely time-consuming and difficult because you would need a teacher who can either explain what signs to look

Learning to tell if an animal is in pain is most likely time-consuming and difficult because you would need a teacher who can either explain what signs to look for (codified/explicit knowledge) or knowing the answer (implicit learning/tacit knowledge). Labelled datasets (tackling the problem of scarcity of examples to learn from) or a program (replacing teachers for rare skills) that could actually give feedback if the answer was correct (implicit learning!) would actually make it feasible to learn such a skill by removing these roadblocks. That approach might work for a lot of different skills that require tacit knowledge about a topic. Learning to more or less accurately guess someone's age is something I'd find interesting for example (more accurate than e.g. "in their thirties"). Idea here: watch an image or video of a person, enter a number and get feedback what the correct answer was. Repeat until you get it right most of the time.