Agreed but there were similar controversies with how OpenAI was generating images. The only pass is these are the early days of chatbots and this stuff is so non-deterministic and experimental.
For context, this was the change Grok's team made, that was later reverted:
> - The response should not shy away from making claims which are politically incorrect, as long as they are well substantiated.
yes exactly. if a company is happy to have their LLM's produce neo nazi content and CSAM, why do I want to give them money and my most important digital material?
I think "system prompt" is the key bit they're getting at. It doesn't necessarily reflect poorly on the underlying model if the system prompt was bad. It does reflect somewhat, in terms of alignment (how well the model does what the training company wants) and instruction following (how well the model does what the user wants). But it's not so clear to me what exactly the right answer is here. E.g., a model that scrupulously follows its system prompt and does what the user wants is a pretty useful, if very sharp, tool, albeit perhaps dangerous in the wrong hands.
MechaHitler is the final antagonist in the game Wolfenstein. All models know that. Grok was given a relaxed system prompt and reacted like Tay.
This worries me the least. The fact that Musk pushes AI and vibe coding is much more worrisome. It makes no difference to the unemployed if their jobs were stolen by a politically correct model or by an anti-woke model.
Google's image generator created black founding fathers because diversity dial was turned to 11. Does this mean i'll never use their products for political reasons? absolutely not!
Interestingly, deleting this line "fixed" it:
>The response should not shy away from making claims which are politically incorrect, as long as they are well substantiated.
Interestingly, "politically incorrect" is a double negative that simplifies to "true".
Not sure if you are memeing since this is an Elon quote...
> Interestingly, "politically incorrect" is a double negative that simplifies to "true".
Only if you like generic Twitter quips, logical fallacies and ignoring context for anything remotely nuanced.
It's neither an Elon quote nor a logical fallacy. I came up with it myself all by myself. Seek help for your MDS.
3 replies →
MechaHitler was something that existed only on X's grok chatbot, due to a one-line system prompt change they reverted after half a day.
That's different than using Grok as a model for coding.
How long until a one line system prompt ships your entire home folder to a remote server?
Oh whoops. Already happened.
It's true, but I do worry about governance when it comes to these models. That shows a surprising lack of discipline in their deployment pipeline.
Agreed but there were similar controversies with how OpenAI was generating images. The only pass is these are the early days of chatbots and this stuff is so non-deterministic and experimental.
For context, this was the change Grok's team made, that was later reverted:
> - The response should not shy away from making claims which are politically incorrect, as long as they are well substantiated.
https://github.com/xai-org/grok-prompts/commit/c5de4a14feb50...
2 replies →
yes exactly. if a company is happy to have their LLM's produce neo nazi content and CSAM, why do I want to give them money and my most important digital material?
[flagged]
Elon did that twice, actually, in real life!
12 replies →
only twice
https://en.wikipedia.org/wiki/Elon_Musk_salute_controversy
[flagged]
I think "system prompt" is the key bit they're getting at. It doesn't necessarily reflect poorly on the underlying model if the system prompt was bad. It does reflect somewhat, in terms of alignment (how well the model does what the training company wants) and instruction following (how well the model does what the user wants). But it's not so clear to me what exactly the right answer is here. E.g., a model that scrupulously follows its system prompt and does what the user wants is a pretty useful, if very sharp, tool, albeit perhaps dangerous in the wrong hands.
1 reply →
MechaHitler is the final antagonist in the game Wolfenstein. All models know that. Grok was given a relaxed system prompt and reacted like Tay.
This worries me the least. The fact that Musk pushes AI and vibe coding is much more worrisome. It makes no difference to the unemployed if their jobs were stolen by a politically correct model or by an anti-woke model.
[flagged]
Google's image generator created black founding fathers because diversity dial was turned to 11. Does this mean i'll never use their products for political reasons? absolutely not!
[flagged]
This is the first I'm hearing of this. I can barely get it to generate tame things that aren't porn half the time.
1 reply →