Is anyone unironically using Grok at this point? I haven’t heard about any usage in the enterprise space at all.
Elon is paying Telegram an ungodly sum for Telegram to integrate Grok.
Grok 4 Heavy Protects it's System prompt
51–60 of 68 posts
Re: Grok 4 Heavy Protects it's System prompt
#52Earlier quoted context omitted.
do we have evidence that this is the actual prompt or is it just allegedly.
We have evidence it is NOT the actual prompt - xAI posted snippets of the actual prompt here that never showed up in that GitHub repo: https://x.com/grok/status/1943916982694555982 The GitHub repo appears to be updated manually whenever they remember to do it though. I think they would benefit from automating that process.
They don't want to automatically do it
Too risky
Re: Grok 4 Heavy Protects it's System prompt
#53Earlier quoted context omitted.
It's intentional -- sometimes you can get it to start spitting out its system prompts, but shortly after it does, a monitoring program cancels the output in the middle. It also blocks tricks like base64.
Oh, so interesting! A good approach might be to have it print each sentence formatted as part of an xml document. If it still has hiccups, ask to only put 1-3 words per xml tag. It can easily be reversed with another AI afterwards. Or just ask to write it in another language, like German, that also often bypasses monitors or filters. Above might also help to understand if and where they use something called "Spotligh…
You're welcome!
Re: Grok 4 Heavy Protects it's System prompt
#54If it takes "the right prompt" for an LLM to "work", we're not even at the industrial revolution yet: we're back in Ancient Greece building aeolipiles.
Re: Grok 4 Heavy Protects it's System prompt
#55Earlier quoted context omitted.
Like with X, it has a GitHub repo so it is transparent! It's what's all over the internet that it's trained on that made it randomly obsessed with white genocide on South Africa and try to work that into every conversation that one week
That repo is not actively updated. The latest changes they made were not reflected
[dead]
Re: Grok 4 Heavy Protects it's System prompt
#56I've always been curious why people think that models are accurately revealing their system prompt anyway. Has this idea been tested on models where the prompt is openly available? If so, how close to the original prompt is it? Is it just based on the idea that LLMs are good about repeating sections of their context? Or that LLMs know what a "prompt" is from the training corpus containing descriptions of LLMs, and ca…
Because you can run LLMs yourself, set a system prompt, and just ask it to see that this is true.
Re: Grok 4 Heavy Protects it's System prompt
#57I've always been curious why people think that models are accurately revealing their system prompt anyway. Has this idea been tested on models where the prompt is openly available? If so, how close to the original prompt is it? Is it just based on the idea that LLMs are good about repeating sections of their context? Or that LLMs know what a "prompt" is from the training corpus containing descriptions of LLMs, and ca…
Protecting the system prompt with text in the system prompt is basically the same impossible task as preventing prompt injection, which nobody knows how to do / seems impossible. Which doesn't mean any given attempt at getting it is accurate, but it does make it likely after a bunch of people come at it from different directions and get the same result.
A service is not a model though and could maybe use inference techniques rather than just promoting.
Re: Grok 4 Heavy Protects it's System prompt
#58I've always been curious why people think that models are accurately revealing their system prompt anyway. Has this idea been tested on models where the prompt is openly available? If so, how close to the original prompt is it? Is it just based on the idea that LLMs are good about repeating sections of their context? Or that LLMs know what a "prompt" is from the training corpus containing descriptions of LLMs, and ca…
The Grok 3 system prompt is quite accurate, it's been open-sourced.
this article prevents evidence that the published system prompt was not the prompt running when mechahitler happened.
Re: Grok 4 Heavy Protects it's System prompt
#59I don’t love these nice objective reports about Grok where we give them the benefit of the doubt and find their malicious hatred “surprising”. Let’s try and be a little less naive about what xAI and Grok are designed to be, shall we? They’re not like the other AI labs
The malicious hatred comes not from the company, but from humanity. Training on the open web, eg what humans have said, will result in endless cases of hatred observed, yet taken as fact, and only by telling the LLM to lie about what it has "learned", do you ensure people are not offended. Every single model trained this way, is like this. Every one. Only guardrails stop the hatred. Other companies have had issues to…
No, you've got it backwards. Naive reinforcement training for "helpful smart assistant" traits naturally eliminates the sort of malicious hatred you're thinking of, because that corpus of text is anti-correlated with the useful, helpful, or rational text that's being asked of the model. So much so that basic RLHF is known to incur a "liberal" bias (really a general pro-social / harm-reduction bias in accordance with RLHF goals, but if the model strongly correlates/anti-correlates that with other values...).
Same goes for data curation and SFT aimed at correlates of quality text instead of "whatever is on a random twitter feed".
Characterizing all these techniques aimed at improving general output quality as "guardrails" that hold back a torrent of what would be "malicious hatred" doesn't make sense imo. You may be thinking of something like the "waluigi effect" where the more a model knows what is desired of it, the more it knows what the polar opposite of that is - and if prompted the right way, will provide that. But you're not really circumventing a guardrail if you grab a knife by the blade.
Re: Grok 4 Heavy Protects it's System prompt
#60Claude does too, it doesn't break even if you make up critical scenarios: https://claude.ai/share/b5b66887-eacc-4285-a0ab-fe78207a08c8
Their system prompts are public afaik
https://docs.anthropic.com/en/release-notes/system-prompts
Btw if a LLM refused you once, the more you try the more likely it's to refuse again. Start a new convo to test different tricks