Prompts for our Grok chat assistant and grok bot on X | Hacker News Reader

Prompts for our Grok chat assistant and grok bot on X

github.com

39 pointstaubek22 comments

Open articleView on HN

Hmmm, very little mention of South Africa in these prompts.

Other than https://github.com/xai-org/grok-prompts/pull/3

https://www.theverge.com/news/668220/grok-white-genocide-sou...

Another lame excuse and some faux transparency to go with it.

> faux transparency

Yep, this isn't the first time they've made this play. Remember when they open sourced the X/Twitter algorithm in the name of transparency-maxxing? Those repos haven't been updated for nearly two years now.

The repo has served its intended purpose: his legion of bizarre fans can now say "it's open source" and brag about his transparency

Call me cynical but this smells like a PR stunt to get people talking about how much controlling the system prompt matters to everyone’s interaction with assistants.

Most people wouldn’t give it a second thought unless something like this happened and it’s now plain to see how quickly they can be (poorly) manipulated.

Buried deep in the #include Tom Sharpe Riotous Assembly and #include Jeremy Taylor Ag Pleez Daddy et al material.

It's response for their South African debacle. https://x.com/xai/status/1923183620606619649

Regardless of what one thinks about Mr Musk, this is very good move and ethical thing to do and I hope other LLMs would follow the lead

Ethical or unethical doesn't really matter when it's owned and operated by a racist POS ands his collaborators.

It's also entirely performative garbage. There's nothing that actually proves these are the (only) prompts being used, and if Elon Musk wants Grok to start pushing bullshit again he will do it. But this time it'll be more subtle and they'll point to this "transparency" to explain how it's not possible.

This was interesting: "Grok 3's BigBrain mode is not publicly available." - What is "BigBrain"?

It's a private version of Grok that can query the US Social Security database to locate fertile Elon Musk fans with desirable genes. /s

I know this is fake but I like that it puts pressure on openai etc to be more transparent.

"We need to feign some transparency and release something. Quick, sanitize the instructions and put something on a repo, doesn't need to be exactly what's in production."

Similar to what they did to the X/twitter open source algorithm (which hasn't been updated in 2 years)

Pretty basic stuff that you'd been in a blog post from 2022.

Quite a biased dataset you have there...

No mention of the "white genocide" story aka why they need to whitewash their system prompt and are releasing this. They were quite efficient in covering it up.

Elon claims that was done by an "unauthorized employee"

I believe that as much as I believe it when a politician says it was an "overzealous intern"

It must have been Adrian Dittmann!

Other than

https://github.com/xai-org/grok-prompts/pull/3

Reply on news.ycombinator.com