The Sestina Test for GPT-4 confirms that Bing is running on the same model. Overall, a cool sestina, but, like Bing, starts to forget the pattern & messes up the three lines at the end. I suspect Bing, with internet access, will still be better than ChatGPT-4 for many tasks.
I have been applying the Sestina Test to every AI over the past year, to no avail. But Bing is sooooo close to being able to write one of the most difficult poetic forms. It adds one stanza and messes up the envoi a bit. Still pretty amazing that it is able work with the form.

Mar 14, 2023 · 8:20 PM UTC

Replying to @emollick
Is there any reason to pay for ChatGPT+ to get GPT 4 there?
Replying to @emollick
You think even with the 32k token limit of gpt4? If it doesn't know something you can always copy/paste it to it for context. With the unlimited conversation lengths I think that could make a big difference.
Replying to @emollick
This was the signature form of my professor John Ashbery. en.m.wikipedia.org/wiki/John… He helped me get publish in American Poetry Review with my sestina Thamos, King of Egypt.
Replying to @emollick
Bing‘s character and message limits, and the lack of a thread history still makes ChatGPT my favorite for tasks that don’t need a live web search. Better use both in different situations.
Replying to @emollick
The Bing implementation is still really, really bad. Built in voice recognition is strangely terrible, and it insists on doing web searches for almost everything -- greatly breaking the "quick flow" that makes ChatGPT so good.
Replying to @emollick
I guess there is one thing I can do that GPT-4 can't! I'm also glad you appreciate the art of the sestina.