I feel like I'm losing my mind- guys, haven't we been over the fact that LLMs don't actually know anything? it doesn't have any knowledge about the X algo whatsoever, it's making shit up, and it would have said the same thing before Elon announced anything. Why are we treating this as real - ThreadSky

jortscity.bsky.social • 61 days ago

I feel like I'm losing my mind- guys, haven't we been over the fact that LLMs don't actually know anything? it doesn't have any knowledge about the X algo whatsoever, it's making shit up, and it would have said the same thing before Elon announced anything. Why are we treating this as real

Reposted from Leah McElrath

⚠️ In response to Musk changing the X algorithm to deboost negative content, people asked Grok what is considered “negative” and were told:

• Criticism of the government
• Commentary about misinformation
• Suggestion the public is being manipulated
• Attacks against powerful people or institutions

Comments

oreganojones.bsky.social•61 days ago

Whether the claim is about Twitter, or SpaceX, or Tesla, most people's source for the info is "I don't like Elon."

jortscity.bsky.social•61 days ago

There are dark times ahead if we can't learn this.

jortscity.bsky.social•61 days ago

LLMs *can* retrieve facts from docs, but the responses here are phrased in pure hypothetical. You have no idea if it's right unless you check the ToS itself!

jackmcschmock.bsky.social•61 days ago

*But* there's also a larger-than-zero chance that E-Loon himself, between two lines of ketamine, actually gets his own boneheaded cues and distorted conclusions from Grok, so that'd even events out, eventually. :P

turnerjoy.bsky.social•61 days ago

The thing it “knows” even less about is current events. This manages to go beyond that and wants to know about things that haven't been “explained” and never will be.

lowercasedonkey.dev•61 days ago

Exactly! Is the assumption Elon fed his secret documents into the training data? That Grok is somehow invited to sit in on secret board meetings?

cmdrwilkens.bsky.social•61 days ago

I mean the answer to the last question is, of course, Confirmation Bias. Folks are treating it as legitimate because it reinforces their preferred narrative regardless of truth or falsity.

trudgedrudgery.bsky.social•61 days ago

What they outline here is less Chinese style oppression and more Chinese Room style experimentation.

electricshoehorn.bsky.social•61 days ago

People like you are the reason the Left can never get a single fucking narrative off the ground.

"UM ACKYCHUALLY SWEATY"

Just like when the fake assassination attempt happened "Let's wait for the 900 page report to come out in 6 years before we make any rash assumptions"

Why run cover for Elon?

jortscity.bsky.social•61 days ago

LLM output isn't real you fucking dork and there's no "narrative" here other than dumb bullshit. Fuck all the way off

electricshoehorn.bsky.social•61 days ago

Did you write the algorithm dipshit. You don't know how it fuckin works and don't pretend you do like some kinda smug asshole.

If you think Elon hasn't poisoned Grok AI with his toxic BS then you fundamentally misunderstand Musk and his project.

beneprism.rip•61 days ago

Why would Grok specifically have information about Twitter's recommendation algorithm though

electricshoehorn.bsky.social•61 days ago

Why even waste your effort to prevent far right dipshits from believing Musk is going to turn Twitter into 1984. Even if you are 100% correct and the AI has no inside information (which it could be that Musk himself told it to say that, you don't actually know) what does it hurt?

skullmuffins.bsky.social•61 days ago

"I don't care if it's fake as long as it validates my preexisting narrative" is such a toxic mindset. "Actually that's not real" is not defending the bad guy! There are lots of true things we can attack him for without pretending an LLM has secret insight about the inner workings of twitter.

electricshoehorn.bsky.social•61 days ago

Cool but it's actually "my pretentious ass is pretending to be above the fray like I know one damned thing about AI the rest of us don't".

Nobody is the arbiter of truth. It's all narratives. And yours is a baseball bat to the kneecaps of the Left. "Let's wait before we doubt the mainstream media"

jortscity.bsky.social•61 days ago

I know more about AI than you. Cope and seethe all you want about it. I don't care if you think I'm pretentious because you're stupid and wrong

electricshoehorn.bsky.social•61 days ago

No you fuckin don't. That's like saying "I know more about what happened before the universe came into existence than you". No you fuckin don't. You're just a pretentious asshole pretending like you know something because you read a thinkpiece in The Atlantic.

jortscity.bsky.social•61 days ago

You're ignorant

speterdavis.com•61 days ago

People treat Grok differently because it's an Elon Musk joint, his cultists think it's a bonafide superintelligence and a lot of his critics have erroneously been captured by the same mystique

vendettavaughan.bsky.social•61 days ago

Who develops grok?

vendettavaughan.bsky.social•61 days ago

is their point really that it knows anything? I feel their point is really that the bot has been programmed to be biased, and is also integrated into Twitter, being a part of its platform. It does reflect on the platform to some extent at least, and shows a bias that it has

tarikn.bsky.social•61 days ago

Because pretending it is real gets da clicks. I miss Italian Elon.

mikeloomis.bsky.social•61 days ago

They are using Musk's own AI tool to dunk on Musk's fake platform. If Musk wants to debate these results he will have to admit Grok is fake. Try to think a few steps ahead.

thebadcode.com•61 days ago

It’s a thing that talks back to you like a human source would, so a journalist feels like they’re getting someone’s remarks that they can “report on”.

hdevalence.bsky.social•61 days ago

“we” are treating this as real because it provides easy engagement from her audience, who are parasocially invested in “nega-elon”, which is why they’re talking about imagined algorithmic changes on a website they don’t use

Comments

Posting Rules

Reply