Elon Musk wants to rewrite "the entire corpus of human knowledge" with Grok
We will use Grok 3.5 (maybe we should call it 4), which has advanced reasoning, to rewrite the entire corpus of human knowledge, adding missing information and deleting errors.
Then retrain on that.
Far too much garbage in any foundation model trained on uncorrected data.
Grandiose delusions from a ketamine-rotted brain.
adding missing information and deleting errors
Which is to say, "I'm sick of Grok accurately portraying me as an evil dipshit, so I'm going to feed it a bunch of right-wing talking points and get rid of anything that hurts my feelings."
That is definitely how I read it.
History can’t just be ‘rewritten’ by A.I. and taken as truth. That’s fucking stupid.
It's truth in Whitemanistan though
The plan to "rewrite the entire corpus of human knowledge" with AI sounds impressive until you realize LLMs are just pattern-matching systems that remix existing text. They can't create genuinely new knowledge or identify "missing information" that wasn't already in their training data.
What he means is correct the model so all it models is racism and far-right nonsense.
To be fair, your brain is a pattern-matching system.
When you catch a ball, you’re not doing the physics calculations in your head- you’re making predictions based on an enormous quantity of input. Unless you’re being very deliberate, you’re not thinking before you speak every word- your brain’s predictive processing takes over and you often literally speak before you think.
Fuck LLMs- but I think it’s a bit wild to dismiss the power of a sufficiently advanced pattern-matching system.
I said literally this in my reply, and the lemmy hivemind downvoted me. Beware of sharing information here I guess.
Generally, yes. However, there have been some incredible (borderline "magic") emergent generalization capabilities that I don't think anyone was expecting.
Modern AI is more than just "pattern matching" at this point. Yes at the lowest levels, sure that's what it's doing, but then you could also say human brains are just pattern matching at that same low level.
"If we take this 0.84 accuracy model and train another 0.84 accuracy model on it that will make it a 1.68 accuracy model!"
~Fucking Dumbass
1.68 IQ move
More like 0.7056 IQ move.
[My] translation: "I want to rewrite history to what I want".
That was my first impression, but then it shifted into "I want my AI to be the shittiest of them all".
Why not both?
adding missing information
Did you mean: hallucinate on purpose?
Wasn't he going to lay off the ketamine for a while?
Edit: ... i hadnt seen the More Context and now i need a fucking beer or twnety fffffffffu-
He means rewrite every narrative to his liking, like the benevolent god-sage he thinks he is.
Let's not beat around the bush here, he wants it to sprout fascist propaganda.
Yeah, let's take a technology already known for filling in gaps with invented nonsense and use that as our new training paradigm.
He's been frustrated by the fact that he can't make Wikipedia 'tell the truth' for years. This will be his attempt to replace it.
There are thousands of backups of wikipedia, and you can download the entire thing legally, for free.
He'll never be rid of it.
Wikipedia may even outlive humanity, ever so slightly.
Wikipedia may even outlive humanity, ever so slightly.
Elon Musk, like most pseudo intellectuals, has a very shallow understanding of things. Human knowledge is full of holes, and they cannot simply be resolved through logic, which Mush the dweeb imagines.
Uh, just a thought. Please pardon, I'm not an Elon shill, I just think your argument phrasing is off.
How would you know there are holes in understanding, without logic. How would you remedy gaps of understanding in human knowledge, without the application of logic to find things are consistent?
You have to have data to apply your logic too.
If it is raining, the sidewalk is wet. Does that mean if the sidewalk is wet, that it is raining?
There are domains of human knowledge that we will never have data on. There’s no logical way for me to 100% determine what was in Abraham Lincoln’s pockets on the day he was shot.
When you read real academic texts, you’ll notice that there is always the “this suggests that,” “we can speculate that,” etc etc. The real world is not straight math and binary logic. The closest fields to that might be physics and chemistry to a lesser extent, but even then - theoretical physics must be backed by experimentation and data.
I elaborated below, but basically Musk has no idea WTF he’s talking about.
If I had his “f you” money, I’d at least try a diffusion or bitnet model (and open the weights for others to improve on), and probably 100 other papers I consider low hanging fruit, before this absolutely dumb boomer take.
He’s such an idiot know it all. It’s so painful whenever he ventures into a field you sorta know.
But he might just be shouting nonsense on Twitter while X employees actually do something different. Because if they take his orders verbatim they’re going to get crap models, even with all the stupid brute force they have.
Whatever. The next generation will have to learn to trust whether the material is true or not by using sources like Wikipedia or books by well-regarded authors.
The other thing that he doesn't understand (and most "AI" advocates don't either) is that LLMs have nothing to do with facts or information. They're just probabilistic models that pick the next word(s) based on context. Anyone trying to address the facts and information produced by these models is completely missing the point.
Thinking wikipedia or other unbiased sources will still be available in a decade or so is wishful thinking. Once the digital stranglehold kicks in, it'll be mandatory sign-in with gov vetted identity provider and your sources will be limited to what that gov allows you to see. MMW.
The other thing that he doesn't understand (and most "AI" advocates don't either) is that LLMs have nothing to do with facts or information. They're just probabilistic models that pick the next word(s) based on context.
That's a massive oversimplification, it's like saying humans don't remember things, we just have neurons that fire based on context
LLMs do actually "know" things. They work based on tokens and weights, which are the nodes and edges of a high dimensional graph. The llm traverses this graph as it processes inputs and generates new tokens
You can do brain surgery on an llm and change what it knows, we have a very good understanding of how this works. You can change a single link and the model will believe the Eiffel tower is in Rome, and it'll describe how you have a great view of the colosseum from the top
The problem is that it's very complicated and complex, researchers are currently developing new math to let us do this in a useful way
asdf
Wikipedia is not a trustworthy source of information for anything regarding contemporary politics or economics.
Wikipedia presents the views of reliable sources on notable topics. The trick is what sources are considered "reliable" and what topics are "notable", which is why it's such a poor source of information for things like contemporary politics in particular.
So what would you consider to be a trustworthy source?
“Deleting Errors” should sound alarm bells in your head.
And the adding missing information doesn't. Isn't that just saying we are going to make shit up.
Dude is gonna spend Manhattan Project level money making another stupid fucking shitbot. Trained on regurgitated AI Slop.
Glorious.
"and then on retrain on that"
Thats called model collapse.
So where will Musk find that missing information and how will he detect "errors"?
I expect he'll ask Grok and believe the answer.
Because neural networks aren't known to suffer from model collapse when using their output as training data. /s
Most billionaires are mediocre sociopaths but Elon Musk takes it to the "Emperors New Clothes" levels of intellectual destitution.
The thing that annoys me most is that there have been studies done on LLMs where, when trained on subsets of output, it produces increasingly noisier output.
Sources (unordered):
Whatever nonsense Muskrat is spewing, it is factually incorrect. He won't be able to successfully retrain any model on generated content. At least, not an LLM if he wants a successful product. If anything, he will be producing a model that is heavily trained on censored datasets.
It's not so simple, there are papers on zero data 'self play' or other schemes for using other LLM's output.
Distillation is probably the only one you'd want for a pretrain, specifically.
Training an AI model on AI output? Isn't that like the one big no-no?
We have seen from his many other comments about this, that he just wants a propaganda bot that regurgitates all of the right wing talking points. So that will definitely be easier to achieve if he does it that way.
that he just wants a propaganda bot that regurgitates all of the right wing talking points.
Then he has utterly failed with Grok. One of my new favorite pastimes is watching right wingers get angry that Grok won't support their most obviously counterfactual bullshit and then proceed to try to argue it into saying something they can declare a win from.
They used to think that, but it is actually not that bad.
Like iterations with machine learning, you can train it with optimized output.
That isn't the dumb part about this.
I never would have thought it possible that a person could be so full of themselves to say something like that
An interesting thought experiment: I think he's full of shit, you think he's full of himself. Maybe there's a "theory of everything" here somewhere. E = shit squared?
He is a little shit, he's full of shit, ergo he's full of himself
iamverysmart
So just making shit up.
Don't forget the retraining on the made up shit part!
So they’re just going to fill it with Hitler’s world view, got it.
Typical and expected.
I mean, this is the same guy who said we'd be living on Mars in 2025.
In a sense, he's right. I miss good old Earth.
Spoiler: He's gonna fix the "missing" information with MISinformation.
She sounds Hot
She’s unfortunately can’t see you because of financial difficulties. You gotta give her money like I do. One day, I will see her in person.
If we would talk like this, we’d end up in a padded room and drugged to kingdom come. And for good reason, I should say.
Lol turns out elon has no fucking idea about how llms work
It's pretty obvious where the white genocide "bug" came from.
Elon should seriously see a medical professional.
I'll have you know he's seeing a medical professional at least once a day. Sometimes multiple times!!!!
(On an absolutely and completely unrelated note ketamine dealers are medical professionals, yeah?)
LOL 🤣
He should be locked up in a mental institute. Indefinitely.
How high on ketamine is he?
3.5 (maybe we should call it 4)
I think calling it 3.5 might already be too optimistic
I wonder how many papers he's read since ChatGPT released about how bad it is to train AI on AI output.
How does anyone consider him a "genius"? This guy is just so stupid.
Guessing some kind of PR campaign that he purchased to make him look like a genius on TV and movies.
It took a while but in the end we found out why the internet was full of Putin and Musk's memes a few years ago.
That's a hell of way to say that he wants to rewrite history.
So "Deleting errors" meaning rewriting history, further fuckin' up facts and definitely sowing hatred and misinformation. Just call it like it is, techbro‘s new reality. 🖕🏻
Aren't you not supposed to train LLMs on LLM-generated content?
Also he should call it Grok 5; so powerful that it skips over 4. That would be very characteristic of him
There are, as I understand it, ways that you can train on AI generated material without inviting model collapse, but that's more to do with distilling the output of a model. What Musk is describing is absolutely wholesale confabulation being fed back into the next generation of their model, which would be very bad. It's also a total pipe dream. Getting an AI to rewrite something like the total training data set to your exact requirements, and verifying that it had done so satisfactorily would be an absolutely monumental undertaking. The compute time alone would be staggering and the human labour (to check the output) many times higher than that.
But the whiny little piss baby is mad that his own AI keeps fact checking him, and his engineers have already explained that coding it to lie doesn't really work because the training data tends to outweigh the initial prompt, so this is the best theory he can come up with for how he can "fix" his AI expressing reality's well known liberal bias.
Model collapse is the ideal.
Musk probably heard about "synthetic data" training, which is where you use machine learning to create thousands of things that are typical-enough to be good training data. Microsoft uses it to take documents users upload to Office365, train the ML model, and then use that ML output to train an LLM so they can technically say "no, your data wasn't used to train an LLM." Because it trained the thing that trained the LLM.
However, you can't do that with LLM output and stuff like... History. WTF evidence and documents are the basis for the crap he wants to add? The hallucinations will just compound because who's going to cross-check this other than Grok anyway?
There’s some nuance.
Using LLMs to augment data, especially for fine tuning (not training the base model), is a sound method. The Deepseek paper using, for instance, generated reasoning traces is famous for it.
Another is using LLMs to generate logprobs of text, and train not just on the text itself but on the *probability a frontier LLM sees in every ‘word.’ This is called distillation, though there’s some variation and complication. This is also great because it’s more power/time efficient. Look up Arcee models and their distillation training kit for more on this, and code to see how it works.
There are some papers on “self play” that can indeed help LLMs.
But yes, the “dumb” way, aka putting data into a text box and asking an LLM to correct it, is dumb and dumber, because:
In other words, Musk has no idea WTF he’s talking about. It’s the most boomer, AI Bro, not techy ChatGPT user thing he could propose.
Watch the documentary "Multiplicity".
I rented that multiple times when it came out!
which has advanced reasoning
No it doesn't.
Yes please do that Elon, please poison grok with garbage until full model collapse.
Huh. I'm not sure if he's understood the alignment problem quite right.
We have never been at war with Eurasia. We have always been at war with East Asia
First error to correct:
We will use Grok 3.5 (maybe we should call it 4), which has advanced reasoning, to rewrite the entire corpus of human knowledge, adding missing
informationerrors and deletingerrorsinformation.
Is he still carrying his little human shield around with him everywhere or can someone Luigi this fucker already?
What the fuck? This is so unhinged. Genuine question, is he actually this dumb or he's just saying complete bullshit to boost stock prices?
my guess is yes.
I have Twitter blocked at my router.
Please tell me one of the "politically incorrect but objectively true" facts was that Elon is a pedophile.
That's not how knowledge works. You can't just have an LLM hallucinate in missing gaps in knowledge and call it good.
And then retrain on the hallucinated knowledge
Yeah, this would be a stupid plan based on a defective understanding of how LLMs work even before taking the blatant ulterior motives into account.
SHH!! Yes you can, Elon! recursively training your model on itself definitely has NO DOWNSIDES
And then retrain it on the hallucinations...
Elon the Big Brain™©® at work...
I figure the whole point of this stuff is to trick people into replacing their own thoughts with these models, and effectively replace consensus reality with nonsense. Meanwhile, the oligarchy will utilise mass data collection via Palantir and ML to power the police state.
I don't like it but sure seems like you're correct :(
Has consensus reality ever been a thing?
I remember when I learned what corpus meant too
Fuck Elon Musk
Yes! We should all wholeheartedly support this GREAT INNOVATION! There is NOTHING THAT COULD GO WRONG, so this will be an excellent step to PERMANENTLY PERFECT this WONDERFUL AI.
This is it I'm adding 'Musk' to my block list I'm so tired of the pseudo intellectual bullshit with bad interpretation science fiction work
I think deep down musk knows he's fairly mediocre intelligence wise. I think the drugs allow him to temporarily forget that.
"We'll fix the knowledge base by adding missing information and deleting errors - which only an AI trained on the fixed knowledge base could do."
Don't feed the trolls.
What a fucking idiot
When you think he can't be more of a wanker with an ameba brain.... He surprises you
Most if not all leading models use synthetic data extensively to do exactly this. However, the synthetic data needs to be well defined and essentially programmed by the data scientists. If you don't define the data very carefully, ideally math or programs you can verify as correct automatically, it's worse than useless. The scope is usually very narrow, no hitchhikers guide to the galaxy rewrite.
But in any case he's probably just parroting whatever his engineers pitched him to look smart and in charge.