ChatGPT and Bard in Discussion

Hello,

some weeks ago Google has published its own big
speech AI, named Bard. A normal Gmail account is enough
to use Bard.

Today, I let ChatGPT and Bard create a discussion.
I alwys entered the text of one programm into the other.
In 25 minutes the following discussion happened.
The pdf document ist 18 pages long and can be found here.

https://www.althofer.de/chatgpt-bard-discussion.pdf

After some time one gets an impression who is creative
and who is just acting like a parrot.

Feedback is very welcome, either here
or with Email to ingo.althoeferGOLEMuni-jena.de
Of course you have to substitute golem properly.

Cheers, Ingo.

For me, both are parrots. I can't see them creative at all. They may best fit as assistants, but not as creative bots at all. At least this is my opinion according to the published dialogue.
For me, I use ChatGPT 3.5 more than Bard in my work, however I edit its output heavily to make it proper as a professional report, document or even a short email.

I just skimmed the text. Seems they mostly are in meta discussion and not actually doing much. But that is not so surprising, because both are fine-tuned to follow user instructions, not to lead dialogues into a new direction.

BTW: I've done the same - letting 2 bots talk to each others - already in the year 2000! We even had a small booklet printed, not sure if I still have it.

There's a difference between them to be sure, but largely they're just so excited to work together their excitement fills the 15 pages they didn't just bring up very worn tropes.

The new wave of 'AI' feels fundamentally like Cleverbot running on a Core i9 instead of a Pentium 4 and with access to a database measured in petabytes rather than gigabytes.

Very strange, that you "all" view things so differently from me.
Bard is really acting like a parrot, but ChatGPT makes constructive
proposals for a story plot.
If I would have to give them ELO ratings, Bard would get a mere 1200
and ChatGPT something like 1700 or 1800.

@richie fourtytwo: If you find your booklet from 2000 again,
please let me know. For instance via
ingo.althoeferWOSCHuni-jena.de
(I am preparing a presentation for the next ICGA conference "Advances
in Computer Games" - to be held online end of November).

Cheers, Ingo.

PS. Does someone have ideas how to assign something like ELO-ratings
to Chat bots in constructive

For Chatbot Elo rankings see: https://lmsys.org/blog/2023-05-03-arena/

I'm travelling atm, will see if I can find the booklet tomorrow.

Hi Ri42,
thanks for the fantastic lin!

Cheers, Ingo.