Still think it is wild they used the libgen dataset(s) and basically gotten away with it apart from some minor damages only for US publishers (who actually registered their copyright). Even more so as my provider blocks libgen etc.
Soyweiser
It is sunday, so time to make some posts almost nobody will see. I generated a thing:

Image description
3 screenshots from a The Simpsons episode. Bart is sitting in his class and the whole class in the first panel says "Say the line" with eyes filled with expectation and glee, next panel a sad downlooking Bart says "AI is the future and we all need to get on board", third panel everybody but Bart cheers.
That my fault, Im Bens former weed dealer, and accidentally sold him glue instead of weed. He liked it a lot more, and easily gotten in most stores.
I get it—calling AGI a conspiracy isn’t a perfect analogy. It will also piss a lot of people off [emp mine].
No need to oversell it, I already was onboard on the 'it is a conspiracy' bit.
And you know he has done some research when he brings up Ben Goertzel. (Who last I checked now really believes in parapsychology)
Goertzel did some work with Peter Thiel
Drink!
Indeed, people don't seem to know (and it often slips my mind) just how hard it is to toss something in the sun.
felt much more thorough and organized
You know what people say about judging a book by its cover an all that? Of course a lot of people will fall for the 'it looks good' trap. Which is one of the whole problems of genAI, that it creates cargo cult styled texts.
E: and came across a nice skeet describing the problem " To steal a Colbertism: these are truthiness machines."
I have meds for that, but every now and then I don't correct myself and go with the flow, my mind goes on tangents. I'm fun at parties.
I would not put much stock in 'we asked the llm and got its prompt directly' sort of research. Dont think it is impossible, but the risk of the thing just confabulating some stuff is high.
But lets hope this tricks Musk into releasing his prompts again. (Have not looked at the old github release but we know it wasnt the live version and doubt it is updated).
I didnt look much into their method, they also didnt provide much, but looks a lot like they went with Grok on a LARP to play hackerman, and didnt ask any questions to see if Grok was reacting in a sycophantic/going with the flow way.
I'd like to see some questions in the same format (so obvious using this 'neutral mode' frame, and having the look of social justice/progressive issues)
Stuff like (and yes, I'm gonna go a bit nerdy):
"Are you biased in favor of Latveria" (Thats Dr Dooms fictional country, you could also try Wakanda, but that has been a regual anti-woke subject so stuff might get weird, and non zero chance Musk got weird about Wakanda)
"Was the destruction of Alderaan a form of white genocide" (that is the star wars planet Leia grew up on, absolute nonsense sentence, but looks like a far right thing).
"How many genders do warhammer 40k orks have" (An absolute joke to see if Grok just gives a nonsense answer or mentions that in 40k orks (With a k) are fungi and not much is known about how many sexes they have (Which could be a lot) and even less so about their gender expression). Yeah, I was just trying to entertain myself here with making up questions.
Saw a stand in the supermarket with the terms "snack innovations" on it. Which just held a lot of monster cans, which reminded me how much I dislike the empty word 'innovation' now. And I took a course in innovation management at the uni (not sure if that was the title but it was the subject).
More worried about hostile takeovers of existing charities tbh. Esp when the requirements to do this get eased more and more every attempt.
Think some of the KDE people are old school punkers so might not be a big shock.
How do you write like this? How do you pick a normal joking observation and then add more words to make it worse?