Thread 106951612

500 billion went into training this btw

>>106951612 (OP)
Yes, it will.
LLMs are not AI.

>>106951612 (OP)
>another luddite cope thread.
Kill yourself

>>106951612 (OP)
>wastes electricity summarizing data from a website that had the exact answer already in text format.
Omg it's capable of reasoning!

>>106951612 (OP)
this nonsense generator is already writing your laws and decrees since every lazy official is using it

>>106951612 (OP)
the answer is dollars

a beggar has zero dollars
a man has two dollars
a King has four dollars

>>106951612 (OP)
Come up with an original riddle that hasn't been asked or answered before, and ask it. If it can correctly answer that, then I will be impressed. Until then it's just regurgitating answers to common riddles it found online.

>>106951612 (OP)
kek

>>106951693
How is the king only twice as rich as the man?

>>106951637
it's all from reddit which is why it's retarded

>>106951742
>>106951892
I just don't fucking get it. Instead of teaching it premade solutions to problems that require counting letters, why not teach it to respond something along the lines of "this is probably a riddle involving counting the amount of letters in separate words, but as this particular LLM model I am unable to do it."

>>106951918
Because literal 80 IQ retards are the target consoomers of AI slop and they will believe it says anyway but if it sounds indecisive they won't take it so seriously.

>>106951918
LLMs fundamentally cannot be taught things they are unable to do, else they'll start responding with "I don't know" to shit they SHOULD know.

>>106951918
Why not just write the whole thing off and pretend we didnt just inflate a $100T stock market bubble based on chatbots?

>>106951612 (OP)
That's absolutely retarded.

>>106951693
this is probably $12 plus tip in 2025 though

>>106951612 (OP)
Answer is kidneys, the beggar sold his 2 to the king

>>106951612 (OP)
Legs? King rides a horse, beggar lost his in the war.

>>106951647
seeing as how there is nothing else viable for current use other than llms will it still be another 2 weeks?

>>106951612 (OP)
I mean to be honest I would reply the same shit lmao

>>106951612 (OP)
AI usually does this to me
Dunno why.

>>106951647
>LLMs are not AI.
Everyone thinks "ChatGPT sucks, therefore automation can never happen"
It's like if you looked at the Wright Flyer (pic related) and proclaimed that airplanes will never be popular

>>106951918
think of it this way. LLM is basically a book with practically infinite pages produced by averaging all other books, and it has a search function to quickly jump to a part that has something that can be paraphrased to your input. what goes next is the "answer". now, how many books have you read that say "i don't know"? it might be immediately after questions like "what's behind event horizon" or "what's outside of observable universe" but the whole list of unknowns is not enumerated in such a way that could be found by an LLM.

>>106951652
>another luddite cope thread.
>Kill yourself

>>106952990
You are brown and useless

>>106952989
Why can't LLMs have guiderails set up manually by their programmers? Like if you don't have any training data for a thing, instead of hallucinating and lying, just say you don't know. I'm not a programmer but their must be a way to translate what I just suggested into code.

>>106953242
because nobody programs a LLM. bunch of text is used to train a model, that model becomes "good enough" at mimicking the text, and that's it. it's a very compressed text jpeg.
there is of course another alternative, using an LLM to translate from normie language into some query like SQL or GraphQL, do a bog standard query, run resulting tsv back in reverse to produce normie language or graphs. that's called RAG and it that case the database query will be able to tell that there's zero results. but this is very uncool, since it's just interacting with a database.

seems like another riddle with completely retarded answer

>>106951892
>Gemini can make mistakes
Gee, ya think?

>>106953713
>two of his own and two of the horse

>>106953713
suits is honestly a good answer

>>106951612 (OP)
The answer is testicles. The king has 2 testicles + 2 since he owns a twink, and man has 2 because he cant afford bussy, and beggar has none because he has been castrated by society

>>106953094
GOOD MORNING SAAAAR
PLEASE TO BE REDEEMING THE AI!! AI IS FUTURE

>>106954013
>Jews are the most excited about the rise of AI, by a large margin
LMAO

>>106951612 (OP)
You all think you're so smart.
Tell me, geniuses, how did you score when you ran a low-ver routine?

>>106951612 (OP)
Not all AI are created equal.

>>106953713
I thought legs, but for a less retarded reason
>Man stands on two legs
>King sits on a throne (4 legs)
>A begger doesn't have a leg to stand on.

>>106953785
TAX DOLLARS WELL AT WORK

>>106951918
how would that affect the stock price if suddenly every other answer to a question was "I don't know"?

>>106951918
>but as this particular LLM model I am unable to do it
chatgpt has computing access so it can run python code snippets
you can try it yourself
there is no excuse for this

>>106951612 (OP)
the answer is balls
a man has two balls
a king has four balls ("crown jewels")
a beggar has no balls because he traded them for fentanyl

>>106951918
they can't, it'll start claiming everything is too hard for it to solve no matter what due to how biases in the model work
LLMs are a dead end, this is an unsolvable problem that even openai has admitted cant be solved

>>106954013
why so mad doe? are you scared of being replaced?

>>106952663
Mcdonald's used to do 3.5€ for an small big mac and fries some time ago, but haven't seen this deal for months.

>>106951612 (OP)
>mogged by grok

>>106952984
Nah. The basic principles in the Wright Flyer at least were still correct and built upon for modern aviation.
LLMs don't even have that. They are fundamentally a dead-end that can categorically never lead to true intelligence and every bit of research poured into them is going to lead us absolutely nowhere closer to it. Absolute waste of energy.

LLMs are more on the level of that one idiot who roped a flat board to either arm and jumped off a church tower while flapping his arms, expecting to fly like a bird.

>>106953242
You don't really "program" the LLM. You program the algorithm which trains it but the "LLM" itself is basically just a bunch of numbers that result from the training process on a huge amount of data and are then used for computation when you query it. You can't really go into the LLM and adjust these numbers by hand in order to fix various problems because these numbers aren't in any sort of human readable or understandable format. They're basically just the result of doing a lot of math on the training set, it's not like a huge but logically laid out database that a person could query and correct by hand or something.

>>106951612 (OP)
No, but I was replaced by young pajeets

>>106951612 (OP)
Schrodinger's AI, where it's shit yet still somehow a threat to your job.

>>106951693
it's lamborghinis in the hollywood hills

>>106951918
Because the goal is looking good on high profile benchmarks and hide all the gotchas in small lettering

Claude came up with something plausible after initially failing a basic spell check:

Or perhaps it's more abstract - could it be titles or names?

A man has two names (first and last)
A king has four names (multiple royal names and titles)
A beggar has no formal name/identity

What do you think? Do either of those fit the riddle?

>>106953883
he's the centaur king

>>106951918
That doesn't work because it of how generative models work. A LLM has some knowledge of distributions of written text over a number of situations, when you give it a prompt it samples from a probability distribution conditioned on your prompt and synthesizes a text. There is no reason why "I don't know" would be preferred over any other answer. See also how chabots I really good at god standard crud apps, leet-code type coding questions and undergrad math but start quickly failing once you start asking for things outside that

>>106951612 (OP)
my answer: dicks
t. not AI

>>106952692
lol

that'll be 500 trillion plus tip

>>106951612 (OP)
Just had an LLM pretend that it could schedule appointments. Good thing I checked beforehand

>GPT-5-high

Arms.

Explanation: It’s a wordplay. A man has two arms (limbs). A king has four—his two arms plus a “coat of arms.” A beggar has no coat of arms, so “none.”

>grok-4-fast

Bedposts

A typical bed for an ordinary man might have two posts (a simple framed bed). A king has a luxurious four-poster bed (with posts at each corner for canopy and curtains). A beggar has no bed at all.

>gemini-2.5-pro

The answer is **knees**.

A man has two knees.
A king has "fore-knees" (four knees) — his own two, and the two of the person kneeling before him.
A beggar has "no-knees" (no knees), because no one kneels to him.

>deepseek-v3.2-exp

Let’s break it down:

A man has two likely referring to legs (a man stands on two legs).
A king has four a king might have four legs when sitting on a throne (throne + his legs = four “legs” visible, or perhaps a king’s bed has four posts, but more likely it’s the “four legs” when he’s carried in a litter by servants, or sitting on a chair).
A beggar has none a beggar might have no legs (if disabled or sitting on the ground with legs hidden), or no support.

This is a variation of the “legs” riddle:

A man has 2 legs.
A king has 4 legs (including his throne’s legs, or his horse’s legs when riding).
A beggar has 0 legs (if we count only what touches the ground — e.g., a beggar sits on the ground, so no “legs” of his own supporting him like furniture legs).

Answer: Legs (in the sense of supports or literal legs depending on context).

>>106955596
>A man has two knees.
>A beggar has "no-knees" (no knees), because no one kneels to him.
Doesn't this imply that beggars don't even have actual literal knees?

>>106954970
You're the one who's seething calling anyone who has the slightest skepticism about AI a luddite and telling them to kill themselves like >>106951652

>>106951612 (OP)
So what's the fucking solution???

>>106955596
Grok gave a 1000X better response than the others

>>106954970
We all know the reason why Indians love AI is because they think it makes up for their low actual IQ and lack of creativity (it doesn't)

>>106951612 (OP)
Some LLMs are just shit. Claude gave the correct answer: legs.

>>106955645
The answer you (and everyone else) REALLY doesn't want to hear:
There is no solution. It's a made up on the spot riddle that actually doesn't have a solution in mind. In other words, they just came up with some BS and then sparked a debate out of nothing.
>b-b-but the king has
There is nothing. It's nonsense and has no solution. You can try to shoehorn something in "y-yeah that makes sense!!!!" but it's just guessing at what could be but will never be. There never was an answer, but these kinds of posts get hundreds upon thousands of engagement clicks, shares, likes, comments, flamewars, and most importantly: revenue. These types of posts flood TikTok, Facebook, Instagram Stories, and more. It's ALL about the content. The Content (tm). The content. That's all it is.

Engagement bait.
Engagement bait.
Engagement bait.
Engagement bait.

Remember this. Call me a schizo, doesn't matter because it doesn't change the reality.

>>106955640
>skepticisms about AI
youre literally saying that ai is shit and worthless, meanwhile you cope here like a good codetroon or artranny.

>>106955645
>>106955679
Bedposts makes perfect sense

>>106955697
My friend, ai IS shit and worthless

>>106955705
keep coping fag. im sure you will be replaced soon. keeeeeeeek

>>106955712
>im sure
See you in 10 years :^)
You won't remember this post, but I will.

>>106955719
>ack

>>106954076
GG! it figured it out after literally googling the answer, holy win for musk

>>106955734
Nooo you cant have AI use the computer!!! THATS ILLEGAL!!!

>>106951612 (OP)
AI isn't mature. If you want to compare it to something, compare it to a child who would sooner eat paint than answer your question.

>>106951612 (OP)
When you got nothing, you got nothing to lose

>>106955871
>>106951892
>>106951742
kek this reminds me of that joke-post about the interviewee who just repeats non-sequitur memorized answers to every question
>so the manhole cover doesnt fall in
>take the goat across then ...

I forgot if I have it saved or what it specifically was though but same vibes

>>106955871
totally not autocomplete parrot machine award

>>106955046
LLMs are part of what AGI will be, that's my vibe atm. it's like just the wings of the plane without the rest of the plane or something. actually fuck the plane analogy

>>106953883
He was going so w- not horribly up until that.

>>106951612 (OP)
AI already has replaced humans for every meaningful task. AI is always correct. The correct answer is the letter A. Make it work, retard.

>>106957352
So true! We need to replace our government with AI politicians right now.

>>106957448
unironically that might be an improvement

>>106952793
Then wouldn't the king have 6?

>>106954394
>>106951612 (OP)
A.L. likes to be technically correct depending on the context

>>106957721
the best form of correct

>>106957721
>technically correct
that's called a truthnuke

>>106951822
The king is Robert Baratheon, his expenses are hueg.

>>106957448
Politicians have already replaced themselves with AI. Why do you think the Trump administration has been so effective since the launch of ChatGPT?

>>106955697
>youre literally saying that ai is shit and worthless
Uh no, retard. Get Grok to summarize for you if you're having this much trouble with reading comprehension

>>106954197
Same thing, but I thought that perhaps the beggar is unable to work due to losing his legs.

AI is already doing better than 99% of the juniors out there. Personally, I prefer to describe the task to an LLM and copy/paste it myself than to explain the task a junior developer. With the LLM I can describe with more technical language, I can review the code right away and finish the task in like 10 or 20 minutes. If a junior developer is doing the task, it will take me 1 or 2 code review and the code won't be in the shape I want.

tl;dr current LLM is already better than most juniors developers. It's just a matter of time until our boomer bosses realise that and start firing all juniors.

>>106951647
Ok, maybe when they invent actual AI but that's not exactly the imminent threat everyone is making it out to be.

>>106951918
Greed from these companies means they'll never let the AI admit it doesn't know something. Even though it's training data probably has people saying,"No I don't know X"

>>106958603
This seems honestly extremely unlikely. I think it's much more likely that detecting uncertainty in any useful way is a hard enough problem it isn't worth chasing.

>>106951957
The conservation of energy is man's leads to man's logical errors. The quickest easiest answer based on already solved solutions is the LLM problem. Altman said that hallucinations/errors are here forever because of this very conundrum.

>>106951647
An ai is just going to be used to make llms less retarded

>>106955007
The writing style by grok makes me think of that anon's story about how he introduced his autistic cousin (?) to Frasier and then corona hit and that was his only source of human interaction for 2 years.

>>106951668
And your Winjeet 11 is 30% written by AI, followed by subsequent updates that break your shit

>>106951655
It's not reasoning, but it can take text from 6,000,000 documents and synthesize it in a new form, possibly making connections that are novel to the reader.

>>106955679
Correct analysis.

>>106954076
A beggar has one though, his birthday suit.

The king is a fury

>>106958443
What fucking Juniors are you hiring? Please giff job sir I can do the computer

>>106953883
me when a BIPOC transfolx is speaking

>>106954394
>Openscam got so btfo by this that they automatically trigger deeper thinking without activating it when specifically asked to count letters

>>106952984
Thats a good point. Based on wright flyer expertise, my car will evolve into a free forming spaceship. Mars is within grasp.

>>106959198
>Please giff job sir I can do the computer
It's pronounced 'jiff'

>>106955596
deepseek is correct, its an old 'trick' question Ive heard like ~20 years ago

>>106954243
Debunk this.

>>106958443
That's pretty sad given AI is shit at programming.

>>106959487
Show me your penis to check

>>106959220
geez, my sides

>>106953785
Surprisingly humourous.

>>106953242
LLMs are grown, not built. There's no codebase you can tweak, all you can do is offer suggestions to the artifact/NN.

a ma,n have 2 dolar
a king have 4 dolr
a homeles have no dolre

>>106951742
This is just wasting time. The conclusion is something that you already know and it took several steps to reach that. A simple searching algorithm could do this much easier, faster, cheaper.

>>106955760
It's like telling a joke to a friend, have him google the answer, and then pretend he already knew it

>>106951652
>transhumanist
Kill yourself

>>106957489
the king donated his legs to israel

Funny thread. This is what I come here for. I love how.it also exposes AI shills as absolutely ignorant retards
Thanks /g/

>the three horsemen of ai images
sloppers can't refute this

>>106951915
Need 500 Brazillion more

>>106954912
kek

>>106955007
He really was made to talk to Twitter users.

>>106954048
stop noticing :)

>>106962494
can't say i've ever generated the middle kind

>>106954243
my sides

AGI any day now.

>>106953785
Borderlands writing

>>106962950
>chatgpt is at PHD level now!!
meanwhile it can't even do basic thing that literal kindergarteners can do

>>106962494
what category is this?

from a previous thread

>>106963410
>>106963420
ChadGPT

>>106963374
The first one.

>>106959514
>given AI is shit at programming
Exactly, that's why I think most programmers will lose their jobs to AI. It's not because AI is super smart, it's simply because most programmers nowadays are really bad.

>>106951892
Here's the full transcript

https://g. co/gemini/share/54dbec9445bb

>>106962950
LOL did they hardcode it wrong?

>>106963660
they overfit on the answer too hard

>>106951612 (OP)
>web search enabled
skill issue

>>106954076
This is the "correct" answer? So lame...

>>106963923
There are many "correct" answers to this riddle.

Legs (A man with 2, A king on throne with 4, begger on ground with none)
Pockets (A man with 2 pockets, a king with a robe might have 4, a beggar will have rags with none)
And so on. Whatever creative answer you can get.

>>106951612 (OP)
No, corporations will replace me with a single junion dev forced to use AI. God forbid we all just use AI and increase our productivity and work-life balance. That'd be too nice for the suits to ever consider.

>>106963374
gigaescher

>>106964080
don't kid yourself, you would do the same thing in their places
such is the human nature

>>106963966
>There are many "correct" answers to this riddle.
>Legs (A man with 2, A king on throne with 4, begger on ground with none)
>Pockets (A man with 2 pockets, a king with a robe might have 4, a beggar will have rags with none)
I can tell you used """AI""" to shit out these pants-on-head-retarded answers, which proves they are not correct.

>>106951612 (OP)

Yes, it will.

>>106951612 (OP)
Why do you assume the unfiltered AIs that elites have access to are the same as the chatslop they throw to the plebs?

>>106954197
>>106958433
I think it's that a beggar kneels or sits while he begs.

>>106951612 (OP)
It's true, I would have half assed this test too :(

Anyway the answer is "mortgages", it's a play on how rich people often have more yet are massively in debt and not any happier than the beggar. Also I would accept "wives" and "masters".

>>106962950
>>106963311
Just another $500 trillion + tip + tax + tariff + mandatory Israel donation + 2 more weeks until AGI!!!!

>>106951822
>How is the king only twice as rich as the man?
Do you think King Charles walks around with a bunch of banknotes in his pocket? No, he has minions to handle his cash for him. The uber-rich never have to personally handle money.

>>106964768
>$500 trillion mandatory Israel donation + mandatory Israel tip + tax + tariff
ftfy

>>106965932
I could use some extra 30 cents per item

>>106966059
>He thinks the savings will be transferred to customers
That's cute. Also, if you're at a point where 30 cents per item are a significant help for you, maybe you need to order less shit from amazon.

>>106966350
lol

buy more

>>106962494
Clearly you forget women with ridiculously huge asses

>>106954076
I mean, the "riddle" isn't much of one if the answer is this stupid, but it's a way better answer than any of the others.

>>106954243
Reminder that LLMs get their training data from Reddit, and joke responses are treated as serious answers.

>>106964080
If your work only consists of some JS boilerplate shit then LLM is pretty good at generating boilerplate code.

>>106951612 (OP)
Unironically correct if you spell "king" as "KAAAANG"

>>106951612 (OP)
So what is the answer?

>>106951693
if i buy this, will i be a king too?

>>106952877
Temp too high?

>>106952692
>>106952781
ahhh a funny thread, hahaha

>>106952984
and now a century later, the only things we have beyond Wright Flyer are turbine engines and swashplates

>>106952984
>traveling like a peasant in a can of salmon and not like a king
ngmi

>>106966350
>yu VILL consoom und yu VILL bi happi!
No, kys.

>>106952984
Reminder that the new york times said a couple days before the Wright Flyer's first flight that heavier than air flight was a million years away.

>>106952692
pls no

>>106952984
>It's like if you looked at the Wright Flyer (pic related) and proclaimed that airplanes will never be popular
In this instance, it's not the technology, or anything inherent with it, it's the people behind it anon. They're not Wrights, they're wrong jews, everything they ever do will be pure cancer every single time.
Imagine not knowing this in 2025. You're ngmi.

>>106968571
>jew york times publishing daily lies
case in point, with jews, you lose

>>106963410
based

>>106968286
There is none, this is a retarded thread. The AI is designed to try to be helpful, not say "uh, go fuck yourself," and OP thinks this is an own.

>>106968836
>The AI is designed to try to be helpful,
to who? not the users, these llms constantly fumble the most basic of information, and it's purposefully done, because behind the scenes they have to deal with all the censorship just like neutered search engine results prior
you're a dumb niggerfaggot you know that?

This shithole is the only board that's still delusional about how AI will take over any day now.

Everyone else is (slowly but surely) realizing what a gigantic waste of time it is. OpenAI is a masturbatory venture that's constantly losing them money 5 years going and will never generate a profit because nothing "AI" does actually matters and probably won't for decades.

You are being replaced by jeets, not AI.

>>106968847
I just don't use them for censorious topics. For using Linux and shit, they rule.

>>106968849
I have a permanent ChatGPT window open, we aren't going back. Cope and seethe.

.

>>106968726
>jews made AI
AI is actually indians and chinks, not jews. sorry. jews never invented anything or did any work. jews just fund things that benefit jews.
deepseek is not jewish which is why jews hate it.

10 years later and those threads will still be filled with fucktards who don't get the fundamentals behind LLM

>>106961239
kek

>>106951647
I'm pretty sure LLMs are actually considered to be [a subset of] AIs.
But I get your point.

>>106951652
Quite the contrary. It's an AI-bubble shilling thread.
The end goal is to make the viewer aware of the general existence of AI/ML/LLM. Even if it's in a negative light.

>>106951612 (OP)
kys shill

>>106953896
nope
birthday suit is still a suit, riddle says beggar has none

>>106951652
SAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAR

>>106960612
AI is not your friend. It's a tool to generate text. This one generated the correct string of text, the others didn't. It's that simple.

>>106954243
so this is why they do penis inspection

>>106954076
That's the "answer"? Literally nobody owns two suits for that reason, and then we go from judging men to traditional card games. Beggars can wear suits, as well - plenty of old people begging.

>>106955325
>>106954197
These are better.

>>106958433
this is fucking brutal, anon. How's he even alive to beg, then? lol

>>106954394
the way these models work they have no real straightforward way to count letters, short of doing a massive lookup table memorization that they never really get trained to perform.

it's just more efficient to use a calculator than try to contort the model weights to encode this

>>106953785
white people will use any writing system except kanji. truly shamefur

>>106971699
>the way these organisms evolved, they have no real straighforward way to count letters
How many Ns in internment, meatbag?

>>106953896
Deck of cards is better
4 kings
2 jokers (man)
0 beggars

>>106971694
he lost them in 'nam and his only benefits was a $800 check and some leftover MREs 40 years ago

>>106968908
How was Diwali, anon?

>>106953713
It doesn't have an answer, it's intentional nonsense with no answer

>>106971831
Wouldn't know, I am White and follow Jesus Christ.

>>106951953
THIS
The audience doesn't give a shit if they are getting lied to, as long as it appears plausible and the LLM is confident about it.
It's about shifting responsibilities. You are not yourself responsible anymore for the crap you believe or promote, the LLM is.

Normie cattle loves to give away responsibilities.

>>106971699
>its about muh token length
This bullshit was always wrong and still is.
It makes no sense. It's just a "look how much i know" gotcha.
Why wouldn't an LLM, by the context of the tokens, be able to count letters?
The token "berry" is more likely to appear with a number of 2 or higher in a context of counting letters. So if you train it on enough letter-counting challenges, it should eventually get it right, without a massive lookup table.

Think about what you claim and try to reason it. You know, like your GPT deepthink does, and you will realize that there is no real basis behind this.

>>106955347
Can't you just fine tune certain scenarios? In the raw form a gpt doesn't even answer questions, it basically would just continue your question, but with fine tuning you get the question answer format. So why not fine tune it to say I don't know in some situations?

>>106951655
It fulfills the boomer logic of "i have to be able to talk to the puter and ask it questions, the onus cant be on me to seek out the information myself"

>>106971799
Jacks and Kings are men too though, so you have 4 kings, 10 men, and 0 beggars. Nevermind that the format of the question doesnt make sense in this context anyway.

>>106957721
there wouldn't be an "of" there by the way.
You'd say how many billion "r"s.

>>106951647
>>106952984
The AI fags typed this like the mesugakis that they are, They would have no time for sweaty, obedient investments with big money /biz/vestors has no time for such work.
Being ravaged, the codebase remains full of moans and typos. The end product, a sloppy, wet mess.
Seeing that their investment had gone to waste, the investor salvages the AI devs obedient orifices for stress relief, at least he would save on any holiday money by doing so.
Thus with chains on their necks and the investors name branded and written on their behinds alongside coding language such as;
sudo apt sex - preggers
The dev laid at the foot of the bed, hungry for rich orifical investments and coding collars to be multi-functional with shock and coding features so the master, /biz/vestor ,could sell to other such AI investors and their slaves, earning back his losses and thus letting the dev free inside the house so that the dev can be used and punished anywhere in the house for a fresh change of pace.
Such is devving for AI, of course male AI devs on the other hand would simply be replaced with indians and sent to the Tech support mines to mine tickets in the dark

>>106971826
>he lost them in 'nam and his only benefits was a $800 check and some leftover MREs 40 years ago
He should have spent the $800 on four good quality suits.

>>106972286
millenials are the ones working on ai though

>>106952692
lel

>>106963538
this

>>106970017
I hate the way Grok talks when you mention 4chan.

>>106954013
Is Argentina white, then?

>>106970017
I never actually says /g/ is wrong to post the comments we do btw.

>>106951637
we need another 50 trillion and 10 lakes AND 500 nuclear energy plants

>>106970017
Genuinely if you can use this without getting actual dread from something obviously inhuman trying and failing in an inhuman way to sound human, you're one of them. AI, skinwalker, the thing. Don't care.

>>106953713
That's the point. There is no answer so it fucks the AI in three unique ways.

1. It is confident, so the AI is going to say yes, and that it knows. There are similar riddles in the dataset, so it is heavily biased to start the generation with confirmation.
2. Due to the bias, and due to the fact that LLMs are just pattern continuers, it'll just generate an answer similar to the dataset, maybe with the context changed.
3. AIs cannot go back, so it can only either answer with a nonsense answer, or go long enough that it loops back and tries to generate that there is no answer. Since the bias is so strong, the token probability is either extremely low, or cut off by the sampling settings.

All three are unsolvable with LLMs.

>>106951612 (OP)
>people think theyre smart for being able to 'stump' the LLM with a riddle
>all you reveal is how retarded you yourself are, because you are coming from a place of assuming it has any sentience or logical thinking, in order to be able to solve a riddle in the first place

>>106975197
>Is Argentina white, then?
It's mostly green.

>>106951713
this
But these things can't think independently, so it will not be able to solve an original riddle.

>>106970017
Fucking Grok killing it and roasting fa/g/s real good.

>>106975170
It must have been trained on 4chan threads, judging by how it "speaks", so you bring that out by mentioning 4chan.

>>106962494
#2 is literally the only good use for AI

>>106951655
this is why AI is such a scam, its just an improvement of the already existing utility of search engines. all its doing is boiling down like 10,000 results and pasting it into a small paragraph, but what if all 10,000 of those results are just redditors with bad info? that doesnt make it the correct answer.

>>106965932
Disingenuous article.
It says 600,000 workers Wouldn't need to be hired to double their output.
Means those jobs don't exist currently.

>>106955679
Yes. You can take any solution and as long as you're convincing enough it's correct.

>Testicles
Man has two,
King has four, (stole the beggar's testicles for sport)
Beggar got his balls cut off

>>106953242
Because you don't program a model, you groom her... I mean it.

>>106951612 (OP)
Until I can see an AI rock up drunk 2 hours late to work, I am irreplaceable.

>>106955679
> Call me a schizo
Why would I?
You are just spouting little bits of common sense.

>>106955645
Cocks. The answer is cocks.

The final hurdle between an AI and an AGI.

>>106951612 (OP)
Works on my machine:


The riddle "A man has two, a king has four, and a beggar has none. What is it?" appears to lack a single definitive solution based on available sources, potentially designed as an engagement prompt rather than a traditional riddle with one answer. One source describes it as "an engagement farming troll" with no intended solution.

The two most commonly proposed answers from social media discussions are:

- "Suits": A typical man owns two suits (e.g., one for weddings and one for funerals), a king "has" four suits (referring to the four suits in a deck of cards, each containing a king card), and a beggar has none. This received 259 upvotes in a Reddit thread. It also appears in Facebook comments.

- "Legs": A man stands on two legs, a king rides a horse or sits on a throne with four legs, and a beggar has none (either sitting on the ground or depicted as legless/no legs to stand on). This received 184 upvotes in the same Reddit thread and is mentioned in multiple X posts and another Reddit post on AI fails.

Less popular suggestions include "money" (a meme-like response counting "money" instances, seen in X posts), "letters" (mismatched word letter counts), "bedposts" (bed sizes), and humorous ones like "bitches" or "Pokemon cards." No traditional media sources provide a clear origin or consensus, and riddle databases yield no matches.

>>106951647
AI is just synonym to ML because ML sounds like a process and AI as final product. Even if you go by the definition of AI having to be intelligent, the closest thing we have to intelligence not made of brain matter are LLMs. Sure they mostly just summarise data already on the internet, but you can still get them to do simple logic puzzles that aren’t on the internet, or summarising semi-decently information. Sure you can argue that this hardly constitutes as intelligence, but this is still smarter then whatever other slop generators like image and video models. If your standard for real AI is software that can think and act as a human on similar capacity, then that concept has been already renamed to AGI.

>>106951612 (OP)
Legs?

>>106951742
>that's a fun one
nigga you don't have any emotions don't act like you're some human.. piece of shit computer

i bet they thought they fixed this

>>106953785
I do enjoy the gay horse emoji

>>106954953
the thing is the model doesn't just create 1 output for each input token, there are probably hundreds. and it picks the one with the highest probability of being correct. I don't see why it would be infeasible for the probability to be given along with the output token. the information exists.

>>106951612 (OP)
ChatGPT couldn't even fix a simple bug in bash script. I think I'm pretty safe.

>>106954048
They can finally have golems instead of goys. This is why so many Jews are also concerned about AI turning on them, it’s the story of golem again.

>>106951652
based antiluddite

>>106955046
This is a good analogy
If planes were invented now you'd have Sam Altman dressed as a bird flapping his wings for five billion a quarter

>>106972267
Google tried something like that with Gemini but for sensitive content filtering. If you gave it a lewd prompt it would start outputting and then halt as soon as it became porn like. It was actually very easy to bypass, just randomly running the same prompt a few times usually did the trick, but playing around with the temperature. And recognizing porn is actually a well understood problem with many datasets and benchmarks. Those days they simply feed your prompt to some content evaluator that determines whatever you prompt is ok or not. Once a llm already began outputting it's too late to try and keep it from making mistakes. Perhaps in image processing, where feature learning is better understood, something like that could work better but I wouldn't be hopeful either.
Moreover a lot of those problems ChatGPT 5 has with relatively simple prompts (about counting letters, a few simple riddles, seahorse emoji, etc) are due to OpenAi hazardous fine tuning to try and preempt handle those gotchas as special cases. You're going to get chatbots that refuse to answer things they shouldn't refuse or who fall into silly behaviors as the guardrails interact funny with their usual behavior.
Ultimately, the actual probability linkehoods involved in neural networks are notoriously hard to evaluate, which is why judging confidence in outputs is still an open problem.
accurate "I don't know" would require LLMs to know what things are true or not, but that's impossible since they only know relationships between tokens. But also, once you strip away facts have all tokens defined in terms of statistical relationships between each other conditioned upon long sequences, I don't know whatever the concept of true or false even makes sense. If you ask it to say something that is false, why shouldn't it? As long as a wrong explanation is likely, the llm will choose it. It's also why you can't get around hallucinations by prompting things that contradict the hallucination.

>>106951612 (OP)
Whether you will be replaced by AI depends how "within distribution" your skillset is versus how "out of distribution" it is. These LLMs suck at generalizing to out of distribution tasks, i.e, what they weren't trained directly on. If your job is answering riddles you're relatively safe.

>>106954048
Yup. Inhuman amounts of funding funneling into this project and no one sees the big picture.
The funny thing is AI has existed to a pretty high capacity for the last two decades, even before normgroids got a hold of it, but was primarily used as surveillance and war table talks. The Palantir rabbit hole runs real fucking deep.

>>106984833
You can't not logically support the use of copywritten shit for machine learning. If I read a ton of books then start writing my own would that be a problem? Even though I learned how to write from copywritten material?

>>106955645
It's holocausts. Man has 2 holocausts: II World War and Palestinian massacre. King has 2 additional holocausts: One on the Epstein island and one in freemason temples. Beggar is on fent.

>>106952984
LLMs are simulated and not artificial intelligence. They simulate intelligence but are actually not intelligent. They can not come up with something on their own, they can just remix their training data. And that's all what they do.

>>106955679
>ask a computer in 2005 to divide by zero, something that doesnt have a known solution
>just pops up an alert box saying "cannot divide by zero" or just crashes and you go on with your day
>ask a computer in 2025 to divide by zero, something that doesnt have a known solution
>feeds you some bullshit, says it's four or something
i love the future

>>106952984
The Wright Flyer was a real plane. It uses the same principles any plane today uses still. LLM is not a real intelligence, and the real intelligence, what ever it will be, will not use the same principles of LLMs of today (stochastic prediction machines).

exponential growth of AI intelligence without exponential growth of computing resources will not happen
N != NP no matter how you slice it

>>106986916
Sort of. The Wright Flyer used wing flexion for control, which isn't something used by modern planes. Mind you, real intelligence isn't just a stochastic prediction machine; there's another part to it that clamps results to be consistent, that enables the building of a mental world model and making predictions against it. It also learns quite a bit faster than LLMs do; learning shouldn't be separate from inference.
Right now, LLMs are about as useful as a meth-head on acid. If I wanted that, I'd move to fucking Florida.

There is no god algorithm that solves everything (Turing proved this)
There is no god algorithm that resolves almost everything (Rice proved this)
There is no god algorithm that always finds everything (Wolpert proved this)
There is no god algorithm that approximates everything (Chaitin proved this)
There is no god algorithm that can always infer the correct hypothesis even with all the data (Somolonoff proved this)
Wake up

>>106987112
Mom cancel my meetings, Claude has "$ sudo rm -rf /"'d my system

>>106987584
kek and this is why i don't understand some people i know who just let the models run rampant on their codebases
>oh no the ai broke this file somehow and because i can't into version control i have no backup!!!! hours down the drain!!!
actually happened btw

>>106951612 (OP)
>An Indian will replace you
yes
based boomer KKKuckpitalism demands it

>>106953242
But how do you know what you don't know?

>>106953785
now do breast milk

>>106952984
>It's like if you looked at the Wright Flyer (pic related) and proclaimed that airplanes will never be popular
Nobody did that, everyone knew this was extremely important.

Nobody who isn't using AI as a slot machine thinks it's doing anything critically important. It's capable of important work in some fields, but it's being marketed as the golden solution to all human thinking and logistical complexity, and it falls flat because it was never designed to do that in the first place. Not that anyone cares as long as it makes more people unemployed and Microsoft gets to pocket their wages.