ChatGPT vs ChatGPT showdown

I have had my first PC at age 10 (a sinclair ZX81, 1KB RAM and my mum's tape recorder as mass storage... she still blames me for overwriting her favourite complations). I was curious about what you could do with computers, then fell in love with the creative aspects of computer science (yes, it's a creative field, no eyes rolling please).

Of course, I was fascinated by AI, wrote my first backprop NN in 91, then did my thesis on NN vs wavelet transforms in pattern recognition, then got a real job :) With time I simply got disillusioned by the lack of progress (yes there were many successes in the eons between 1991 and 2023, but still, I was expecting/hoping more).

Then I simply went from disillunsioned to cynical, but still hopeful. I kept on trying new products, but always gettin that "yeah yeah yeah, the usual marketing stunt" feeling, even from OpenAI, testing all new models.

Still kept on feeling "maybe were are just not smart enough as a species to be able to get to a real advance in this field".

Then I tried ChatGPT 3.5, and for the first time in what felt like centuries I felt like "wow, the homo sapiens once again managed to come up with something astonishing". So I kept using it and playing with this new amazing "thing", until the Garante Privacy decided to stop it in italy, but that's another story.

So, left alone with only the API's, I thought to myself: how would chatGPT interact with "itself"? So I tried the following two experiments here using the API's. I simply wrote an initial "instruction" prompt then two "seed" prompts to chose the "subject of debate", and then I passed each answer from ChatGPT back to itself pretending it was from the "user".

Experiment 1:

ChatGPT as the helpful, cheery "thing" we all know and love, talking to itself about "cooperation vs competition". My instruction was to "help the user". Note: the GPT Nice1 and 2 are only there to help read, but it obviusly was always "the one and only, gpt3.5 we know and love (4.0 is just too slow for me)". I particularly enjoyed their final goodbyes...

and it kept "cheering itself" like a couple of lovers at the end of a phone call, untill I stopped it, gave it some rest and proceed to a more "cruel" experiment. I gave it instructions to "argue" against the user and chose a nice divisive subject like brexiters vs remainers.

Experiment 2:

The prompt here was to argue against anything the user said. This is the result.

In this case, it didn't "converge" to a shared solution and there were no pleasant greetings at the end, obviously, so I stopped it before the swearing started (it didn't..).

Maybe one could see in this example a glimpse of a potential dialogue with people exchanging ideas with an open mind or a closed mind, but I leave this to the philosophers among us :)

If you want to read about chaGPT conversing with itself on evolutionism vs creationism or similar existential questions please let me know, I will publish them