Peter N. M. HansteenJan 6, 2024
I just saw a post that referred to ChatGPT as "Mansplaining as a service", and it is so wonderfully correct - instant generation of superficially plausible yet totally fabricated nonsense presented with unflagging confidence, regardless of topic, without concern, regard, or even awareness of the expertise of its audience :D #chatgpt #mansplaining #GenerativeAiIsGoingGreat
Show threadAnti GravityJan 7, 2024
@pitrh
ChatGPT may not be capable of brilliant analysis but to me it’s great strength is that it seems to be incapable of lying or carrying implicit biases as do most other sources of information.
Show threadPeter N. M. Hansteen
@antigravy Can a not-conscious thing lie? It certainly generated made-up article titles it attributed to me. But somebody decided to call that phenomenon "AI halluscinations".
Jan 7, 2024 at 1:50pmWeb
Show threadAnti GravityJan 7, 2024
@pitrh
Sorry, I’ve been unaware of anything like that. I have only played around with the free version on my phone and it has been refreshingly transparent about its limitations. It makes sense that someone using it as you are describing would just delete any disclaimers. Plagiarists gonna plagiarize (and mis-attribute).
Show threadPeter N. M. HansteenJan 7, 2024

@antigravy Some limited experiments, lightly annotated: https://nxdomain.no/~peter/chatgpt_on_ipv6_and_openbsd_poetry.html, https://nxdomain.no/~peter/chatgpt_writes_pf.conf.html and finally where it got exactly one thing right (I did write The Book of PF, the other titles are things I am rather sure I have read but I did not write): https://chat.openai.com/share/cf07bdc6-4341-4a48-90e9-88c85bd1248b

This mainly demonstrates that the quality of the output depends critically on the quality and volume of input it has been trained on (aka classic GIGO, but now with billions of parameters)

That grumpy BSD guy: ChatGPT opines on IPv6 procastination, waxes lyrical over OpenBSD

Show threadAnti GravityJan 8, 2024
@pitrh
I have to admit that I know very little about internet protocol but I still find your probing of ChatGPT’s brain pretty entertaining.