Nectarine
Infamous OneLiner
Rapture
(21:44:04)
but atm I'm too shy to test gpt5. well. lets jump into the cold water. half of the internet ppl say it's very dissapointing. the other half says it's genious. huh.
mirrorbird
(21:43:59)
just trying to help you understand how it works, Rap, there is nothing similar to basic where you could say LET NAME="RICO":LET AGE=18 -- it doesn't have boxes for facts about the world
Rapture
(21:43:28)
haha
mirrorbird
(21:43:14)
sleep well and don't get solar burned
Rapture
(21:43:05)
gn8 fara 
Rapture
(21:42:59)
well that hallucination i spoke about (mixing up the SPECIES, eh RACE, not the name) is only occuring after very VERY long texts. and I write VERY long and complex texts. memory is full pls start a new fresh chat, works good, well.
faraday
(21:42:45)
time to say GN8 
mirrorbird
(21:41:01)
-- but with billions and billions of words, so it's much more realistic-sounding. but it's really the same trick.
mirrorbird
(21:40:35)
you might like to play with this: https://projects.haykranen.nl -- it's a very simple thing that takes a text, and generates new text that "looks similar" (based on words occurring together). then you can imagine AI is like this, but --
mirrorbird
(21:39:59)
but a token isn't a semantic storage location, a token is just a word relative to other words.
