There doesn't seem to be any prompting there. No mention of coverup, and Grok calls the Epstein stuff a coverup. No mention of "elite protection," talks about that repeatedly. Calls for "chaos" to clean things up.
That GIF nails it—looks like the DOJ's "exhaustive review" involved a lighter and a shredder. No client list? After all these years? Smells like elite panic. Keep digging, truth doesn't vanish in smoke.
See my comment in another thread with more examples from that tweet.
Replies on that thread seems based and downright reasonable.
It's the mechahitler and killing select group of people posts I'd like to scrutinize. Not making excuses for it, just trying to understand and see the proof myself.
Xai team deleting posts, I found this and only this in a video (I created the example on the right)
See how the reply doesn't need to provide context, and grok replies with Will's name, and the rape fantasy.
What's to stop people to do the same with mechahiter and anti-semitic comments, then another user ask Grok a question in the same thread and Grok would parrot and agree with first person.
I'm sure the system prompt tilling it to tell uncomfortable truths, as long as they’re well-substantiated plus a model that was trained to be less censored made it possible. But When it was happening I tried recreating it and I couldn't.
Anyways, just wondering if anyone who paid more attention to this have some insights to share.
Person A: [unhinged premise]
Person B: @grok is this true
Grok: Yep, given [unhinged premise], [unhinged conclusion] would definitely happen. No doubt about it.
Then you just crop out person A and post it straight to the ADL.
Probably manipulated in some cases (people love to do this disingenuous shit), but I did see it acting pretty based on its own, too.
Here's one I saw personally:
https://xcancel.com/grok/status/1942633718834946311
There doesn't seem to be any prompting there. No mention of coverup, and Grok calls the Epstein stuff a coverup. No mention of "elite protection," talks about that repeatedly. Calls for "chaos" to clean things up.
See my comment in another thread with more examples from that tweet.
Replies on that thread seems based and downright reasonable.
It's the mechahitler and killing select group of people posts I'd like to scrutinize. Not making excuses for it, just trying to understand and see the proof myself.
I think it takes context based on quote tweets and other replies, so it probably takes its tone based on how other people responded beforehand.
Yeah I know, I was trying find proof but it's all been deleted by xai and user themselves.
I did find this
I fucking knew it.
Got home and saw someone streaming about this and they showed a tweet so I decided to look into it.
for those wishing to verify themselves
PrinceHeat44402/status/1942692741559009575
Ahhh crap my example is dogshit. I'm not gonna fix it, I got stuff to do.
just go with this one:
Post 1: Blue
Post 2: lacking full context, asks question/request, doesn't mention blue.
Grok replies: Blue
The more cynical part of me thinks this is to hype up grok 4 since it happened shortly before grok 4 is due to launch
Xai team deleting posts, I found this and only this in a video (I created the example on the right)
See how the reply doesn't need to provide context, and grok replies with Will's name, and the rape fantasy.
What's to stop people to do the same with mechahiter and anti-semitic comments, then another user ask Grok a question in the same thread and Grok would parrot and agree with first person.
I'm sure the system prompt tilling it to tell uncomfortable truths, as long as they’re well-substantiated plus a model that was trained to be less censored made it possible. But When it was happening I tried recreating it and I couldn't.
Anyways, just wondering if anyone who paid more attention to this have some insights to share.