The Importance of Context in Machine Learning

The most rudimentary machine learning that I have worked with lacks context and it doesn't comprehend the problem right. We started using embeddings which seem to capture some notion of context but for me they are still transformations of input data. For example we were training a model to detect whether it was for a company that sells used clothes so the users of their application would take a picture send the picture to this company  and the company will firstly decide whether it is sellable whether it can be sold, and also how much would it be worth to sell a second hand piece of clothing. And actually we trained or many pictures and the model learned that pictures with better quality have clothes of higher prices