Raw LLM Responses

LLMs are the type of AI people are most familiar with that do predictive text. GANs are another type; sometimes combined with LLM functions these are the ones used to produce images. Image Tracking AI has less flashy quick consumer use cases but arguably more industry use cases, provided it gives a reliable output. Basically the type of training data fed into these models is images and videos so that the model learns to broadly classify objects. "This is a nose, this is a mouth, etc." The most common consumer use case here is certain implementations in self-driving cars. "This is a pedestrian, this is a stop sign." I'm curious how subjective the conclusions they're trying to draw from this are. Discontent, boredom, or inattention seem like the type of things that could much more easily be picked up by human monitoring than by using a tool like this. If they're trying to catch more objective mistakes (Hand was over here at this point while part needed was here) then it might do okay at that if they don't want to pay a person to watch. Obligatory obviously still gross though.