In my weird poetry tests, Google's Bard loses to Bing and ChatGPT-4 by a lot. It can't do a sestina at all, even after being asked about the form. And the poetry is super boring Bing and ChatGPT-4 are almost right, but forget the pattern of the poetic form right before the envoi

Mar 22, 2023 · 3:26 AM UTC

8
7
1
125
Bard also fails the Lem Test SciFi author Stanislaw Lem wrote of two rival constructors of robots. One creates a robotic poet & the other challenges it to write an impossible poem, which it does (the English translator did a great job). Here is the original, Bard, ChatGPT & Bing
5
3
1
38
Replying to @emollick
also, when you call them out on this, they agree, then miss again. :)
1
Yeah, it doesn't seem to learn as much as Bing, and the transparency of searching on Bing is much better (though negotiating with the chatbot can be a bit weird, since you don't want to antagonize it or it might stop responding)
3
Replying to @emollick
Not weird tests in fact.
2
Replying to @emollick
are you of the opinion @emollick that the "forgetting the pattern" relate to the size of the # of tokens that the gpt is allowed to process? on the api gpt3.5 chat side it's easy to specify, but I have no idea what the gpt4 chat model uses by default in their ui.
Replying to @emollick
I have the same overall conclusion from my early tests of AI research that's not in the training data and ethics.
Replying to @emollick
twofergoofer.com/blog/bard a fairly embarrassing showing by Google's Bard in solving rhyme-based puzzles (96 / 100 for GPT-4, 0 / 100 for Bard

Bard is much worse at puzzle solving than ChatGPT - Twofer Goofer Blog

We're testing Bard's ability to solve Twofer Goofers

twofergoofer.com
4