I asked chaptgpt to explain the idea of principal parts in relation to ancient greek verbs. It answered, but gave the present infinitive as the second principal part. That’s different from Mastronarde. But it can tell you who Mastronarde is.
I asked another question: what is the deictic iota. Chatgpt seemed to confuse the deictic iota with the iota subscript.
I judge that chatgpt is now unready to give grammar advice on ancient Greek. However that it could report even wrong answers is pretty impressive, like an overconfident D student afflicted with the Dunning Kruger effect. But not at all like a good teacher, who makes on-the-spot corrections when students have gone astray.
From previous sessions with chatgpt, I gather that its information hoard is fixed. It cannot learn from an interlocutor’s corrections.
like an overconfident D student afflicted with the Dunning Kruger effect
My feelings exactly! I once asked ”What are the weaknesses of the oral-formulaic theory in Homeric scholarship?” The answer wasn’t exactly wrong, but it wasn’t very good either.
But isn’t that impressive in itself? ChatGPT can’t replace an Albert Einstein, a Leonardo da Vinci or a Martin L. West, but then again - most us can’t! Most of us are unable to write anything but platitudes and excruciatingly bad poetry, just like ChatGPT! It can’t beat a truly gifted human being, but it does an excellent job at being an average Joe. Maybe we shouldn’t call it artificial intelligence - but rather, artificial mediocrity?
But isn’t that impressive in itself?
Absolutely! I hope I didn’t sound too negative. chatgpt takes my questions and gives answers relevant to the question, in English I can understand. That is a wonderful achievement. It may be that teaching it codified principles–for example grammar precepts, engineering standards, safety regulations–is less difficult.
From previous sessions with chatgpt, I gather that its information hoard is fixed. > It cannot learn from an interlocutor’s corrections> .
There is ample evidence on these boards that many humans are like this too.
But I wonder if it’s true that chatgpt cannot learn from the questions? I asked a number of questions about accents and when I pointed out that it had made mistakes it acknowledged this and corrected it .
Here’s a blog post about what ChatGPT actually is that I found interesting. (But please note that I don’t know much about AI.)
https://acoup.blog/2023/02/17/collections-on-chatgpt/
The blog contains lots of other material about the classical world and ancient history that might interest folks here as well (how textiles were produced, logistics of ancient warfare, how fortifications work, how ancient Sparta was actually a horrible place etc,).