In the example provided in this article, ChatGPT's response opens with an incorrect phrasing that represents an utter lack of understanding of the situation.
I am writing to request that you consider disputing the parking fine with number 4[...]3 that was issued to me.
This is just getting the premise of the letter incorrect and while a human could definitely parse out the meaning here, the model is getting the actual request quite a bit wrong just in the opening. The person who is sending the letter wants to dispute the ticket, but it is written as though they would like their regional council to dispute it, despite being the issuer of said ticket.
This is a pretty basic language error to be making. This isn't nonsense in that it's just spitting out gibberish, but it is nonsense in that it isn't proper clear communication for the result being sought.
Also, by the time you have written everything into the prompts here, you've basically written the second e-mail, so I'm not clear what time this is saving you.
I'd like to see people blogging about more complex examples than these, because this kind of thing is borderline form letter (I'd like to dispute X, for Y reason).
I bet 90% of humans would make a similar phrasing error quite often, and they would rightly not care because it doesn't affect the ability of a reader to comprehend the text.
I wouldn't be happy if I'd hired a lawyer and they made that kind of small mistake, but it's not even going to register compared with the level of grammar used in the average parking fine complaint.
90% of humans don't speak English, so I'm not sure that's a compelling comparison case.
If we're talking 90% of English writing people, I'd take that bet any day just on the basis that most of the communication I've ever seen isn't written so passively.
This is a pretty basic language error to be making. This isn't nonsense in that it's just spitting out gibberish, but it is nonsense in that it isn't proper clear communication for the result being sought.
Also, by the time you have written everything into the prompts here, you've basically written the second e-mail, so I'm not clear what time this is saving you.
I'd like to see people blogging about more complex examples than these, because this kind of thing is borderline form letter (I'd like to dispute X, for Y reason).