Raw LLM Responses

The problem is ChatGPT can be and regularly is wrong. They deliver these incorrect statements as if it's fact. Therefore, a bit of common sense is also extremely necessary. And well... the people working on these AI programs don't really care about that - in fact, they really don't care about anything other than their technological arms race and making absurd money from it. They pretend they don't earn any money at all to anyone poking their nose into it, selling their specialised tools to big business (tools designed to cut through the noise) while also letting kids and greedy techbros goof around with their older models for free (thus destabilising the creative arts industries, because it allows people to bypass needing any real skill to create anything). Leaving AI to mark an exam could cause similar problems - especially as the AI 'learns' from those inputs. What could be a regular mistake across all exam papers might cause the AI to change the right answer to that incorrect one, because it learns from correlation. It doesn't check to see if that correlation is correct because it's often not equipped to do so. Oh yeah, and don't tell me that anyone wanting to save money on a marking tool isn't going to use the noisy free model with all its flaws instead. Because you know they would. Favouritism sucks. I know that because I was a victim of it. But this could be potentially worse, as it could result in the severely unqualified being in jobs they shouldn't be - and this is already an issue, you absolutely don't want more of them.