Comment on FBI Got Grok to Hand Over Prompts Used to Create Nonconsensual Porn
CombatWombat@feddit.online 1 day ago
I’m usually against complaints about poor headlines, but this one is completely factually incorrect? The FBI didn’t interact with Grok here literally at all? They issued a search warrant to X to get their logs?
The earliest draft title seems to be “FBI Subpoenaed X To Get Grok Prompts Used To Create Nonconsensual Porn”
Which, in my opinion, is more technically correct (especially for a draft) but a whole lot harder to parse. They interacted with the Grok makers, which is kind of what I assumed
I’m not as quick as you. I got most of the way through article and was still wondering why X would expose a database of historical prompts to an llm for querying by law enforcement.
personally I wouldn’t even call it clunky. The current headline misleads about grok and has the wrong focus (the guy did a lot of bad shit and a small portion of that included some actions on grok which are being audited)
I guess my brain equated from with X and reached the intended conclusion. I assumed that the title meant what you say the article is.
I’m not as quick as you. I got most of the way through article and was still wondering why X would expose a database of historical prompts to an llm for querying by law enforcement.
Oh, I wasn’t trying to diss you, I was mentioning that I failed to recognise the difference between grok and X, I was actually berating myself from accidentally understanding the title for the wrong reasons haha.
I parsed it as that they got Grok the company to hand over records.
Although TBF, I misread it as Nonsensical Porn, thought it sounded entertaining.
Grok isn’t a company, its xAi
I’m usually against complaints about poor headlines
Why are you against complaints? They are what most people who are super busy (voters) tend to only read. If they’re bad or misleading, then most of our voters have bad information or are mislead.
People commenting after only reading the headline and not the article is exactly the behavior I find irritating and distasteful about headline-related complaints.
I think you’re in the same boat I am where I fucking haaaaaaate the culture on link aggregators (and probably other social media) where people will bitch and moan to no end that their preferred format (publicly reacting to disconnected headlines whose articles they haven’t read) isn’t giving them literally all the information they need to form a cogent opinion.
They genuinely think that the article body should be effectively superfluous to the headline – not just to have a basic gist of but to discuss and debate current events, which is insane. It reminds me of people who think they can learn math and physics by passively watching somebody else do it – which is true only to an utterly incosequential extent.
Speaking as someone who’s read thousands of articles for research, I feel confident saying that reading the article is an insane force multiplier to understanding. Any time you spent reacting to the headline would’ve been 3x as effective put into reading even just part of an article. This doesn’t just apply to current events, and even I haven’t thoroughly learned this lesson; so many times I’ve been editing Wikipedia and arrived at a point where reading one goddamn article for three minutes would’ve saved me half an hour of fucking around (“2 hours of debugging can save you five minutes of reading the documentation”).
This is my way of pleading with you (you, the non-CombatWombat reader): it’s enriching once you can steel yourself and work through the initial dopamine drought, and it quickly becomes enjoyable. It’s not your fault it’s so hard psychologically; this was done to you by formats that value engagement with the platform over engagement with the material.
It is possible simply to ask a chatbot to reveal their prompt. And Grok isn’t exactly one to push back on anything.
I mean, most llm makers work pretty hard to conceal the system prompt, and I have no idea why XAi give Grok access to a database of historical prompts. LLMs don’t have memories by default, and their inability to learn from past experiences is kind of a big stumbling point for a lot of folks. You can ask, but I doubt you’re likely to get anything other than a confabulation.
TheTechnician27@lemmy.world 1 day ago
Yeah, there’s no fucking way “Well Grok told me these are the prompts they used” would be admissible as evidence of any kind.
Tetragrade@leminal.space 1 day ago
@Grok@twitter.com is this true?