In fact, I’ve found GPT surprisingly candid and useful in explaining areas where it’s still limited, and in identifying its mistakes and why it made them. It’s almost ‘superhuman’ in that regard, as I’ve found it rare as hens’ teeth for people to both admit an error and then surgically explain where they went wrong. Meanwhile, I agree it has an hallucinogenic (“BS”) aspect, but that’s only one side of it. I find it’s also super-good at summarising things based on its impressive ‘search and synthesize’ capabilities. When working with it, I just try to keep things as concise and technical as possible, not pushing it in to its more wonky areas. Frankly, so far it’s been one of the most useful tools and aids I’ve ever used, altho I certainly couldn’t say the same about others, like Gemini.

Yikes, I certainly don’t envy that grading challenge, though. :S