Even though only Scarfolk is correct, I'm putting all of that on my CV.
@cstross @Richard_Littler
Content-wise I agree, but if you ever read or listen to that human bloviator, you will realise ChatGPT has a much better grasp of the English language.
Oh imagine the pitch meeting …
@Richard_Littler I mean, even if it did acknowledge that its responses could cause real-world problems... it would be meaningless because *it doesn't understand what it's saying*.
All that verbiage indicates is that the creators of ChatGPT wish to deny all responsibility ... but we knew that already.
So the primary directive will force AI to force all humans to stop self harming.
The first action AI would take would be to "switch off" all fossil fuel extraction & biomass production. Air pollution is evidently self harming.
Humans will then have to work out it out for themselves.
@Richard_Littler
Its training data is probably about 30% know-nothing guys "well, actually"-ing people on social media, so it replicates that behaviour pretty well.
If pressed, nobody involved in large language models would claim automating extreme online Dunning-Kruger was asked for by anyone or has a use case, but it's kind of a monkey's paw situation now and what has been invented cannot be uninvented.