Hi Percy,
Well it's not a question, but I did notice one interesting failure in ChatGPT; it can't do malapropisms.
So I was playing around with ChatGPT. I got it to do some style pastiches, which it can certainly do, even if I found them a little obvious - starting a Raymond Chandler pastiche with the phrase "It started with a dame"? Really? But it could do it.
Then I asked to it write a rant in the style of a rabid conspiracy theorist. And... well, it did it, but it didn't do it very well. I tried to polish it up a bit. I asked it to add lots of pointless emojis and multiple exclamation marks and that helped a bit. I GOT it to capitalise RANDOM words. That sort of helped. But I still wasn't feeling it, so I asked it to throw in a few malapropisms and I hit a brick wall. It just couldn't do it, or at least do it properly. ChatGPT's idea of a malapropism was just to substitute a word for a random and completely unrelated word. Not a similar sounding word or an etymologically related word, just a random word that no human would ever select. I tried to explain to it where it was going wrong, but no matter how much I coached it I just kept getting bland, nonsensical word substitutions. I mean, a malapropism is supposed to follow a certain pattern, Take these classic examples from The Sopranos;
"...what with the passing of Vito Sr., and all that entrails."
"Create a little dysentery in the ranks"
"I was prostate with grief."
I wasn't expecting world class one liners like that, but I thought I'd get something. I was surprised how bad it was at this, although to be honest, not that surprised. ChatGPT's main ability seems to be churning out mediocrity.
Mutate and Survive
On two occasions I have been asked, – "Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?" ... I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question. - Charles Babbage