If LLMs "can't reason", then why do they keep doing PhD-level mathematical thesis proofs in hours?
>>108809913>If LLMs "can't reason", then why do they keep doing PhD-level mathematical thesis proofs in hours?Because academia is insanely repetitive and predictable.
>lust provoking image>irrelevant time wasting question
>>108809913Can you show us any of these PhD-level mathematical thesis proofs?LLMs literally do not reason. They do linear regression in a chain. Please read something. I promise it's good for you.
>>108810017>LLMs literally do not reason. They do linear regression in a chain.How do you know you can't reason by "doing linear regression in a chain" (which isn't even true)?
>>108809913Because these products are already 95% there from a bunch of other researchers, and AI can tie them nicely together.
>>108809913Remember when a few weeks ago the "AI is useless and can't do anything" crowd had their panties in a twist after some guy found a cure for his specific cancer using some of the FAGMAN LLMs and the pharma industry just said yeah nah not profitable enough.The people outraged about that and the people outraged about AI are a Venn-diagram with only one circle.
>>108809949shes literally my daughter
>>108810190>some guy found a cure for his specific cancer using some of the FAGMAN LLMs and the pharma industry just said yeah nah not profitable enough.Yes, that absolutely happened, just like all the other schizophrenic delusions drip-fed to you by your handlers.
>>108810017https://gowers.wordpress.com/2026/05/08/a-recent-experience-with-chatgpt-5-5-pro/
>>108810207Case in point for >>108809922
>>108810204https://theconversation.com/a-man-used-ai-to-help-make-a-cancer-vaccine-for-his-dog-an-oncologist-urges-caution-278735Looking forward to your next projection kiddo. Don't worry, take all the time you need to tell your handler about this, have him beat you for your piss poor performance and then come back to us with what he told you to say as your pathetic cope attempt.
>>108810242>theconversation.com
>>108810236So basically all white collar work can be automated by LLMs, good to know
>>108810248>give me a source>NO NOT THAT SOURCE!>NO I WILL ALSO NOT READ WHAT WAS LINKED OR THE SOURCES WITHIN!I knew your cope would be pathetic but calling this cope is an insult to actual cope. Btw did your handler copy paste that image while he was beating you?
>>108810252>So basically all white collar work can be automated by LLMsTake your antipsychotics and try to read that post again, because you're responding to something unrelated.
>>108810268What white collar work isn't insanely repetitive and predictable?
>>108810260Maybe try something more reputable, like buzzfeed or the Telegraph.
>>108810275Even if all of it was, your statement is still delusional and unrelated to my point.
>>1088102881. LLMs can do academic work because it is insanely repetitive and predictable2. All white collar work is repetitive and predictable3. Therefore LLMs can do all white collar work
the poster above lives in lesotho btw
>>108810323I'm sure even a mindless LLM can explain to you why this braindamaged "argument" fails.
>>108810330>no argumentI accept your concession
>>108810329>>108810330Luddite cope
>>108810342I accept your profound mental retardation. Good thing you went out of your way to fail at a simple syllogism.
This is it. This is the worst /g/ thread ever made.I'm sure at least 100 more replies will be coming and we'll have this thing on the catalog for at least 2 more days.Baiting on this board is so easy.
>>108810360>no argumentI accept your concession
>>1088103841. LLMs can do academic work because it is insanely repetitive and predictable2. Wiping my ass is repetitive and predictable3. Therefore LLMs can wipe my assNice argument, jeetoid.
>>108810390NTA but the first proposition is false
>PhD-level mathematical thesis proofsThat's a lot of words to say "bullshit".Mathematics was solved centuries ago, everything after is just dudes sitting around jerking themselves off.
>>108810404LLMs can't actually do academic work. Or any work.
>>108810415True. The argument is valid but not sound.
>>108810421>Mentally ill cretin is STILL getting filtered.
>>108810436ru ok
>>108809913DEI is everywhere
>80 IQ browns can't construct a simple syllogism>don't understand the problem even after having it directly demonstratedNice board.
>>108810454You will never be white
>>108810342Conclusion of the disproof:The syllogism fails because premise 2 is empirically false (not all white collar work is repetitive/predictable), and even if it were true, the leap to “LLMs can do all of it” doesn’t follow without additional premises about LLMs’ capabilities covering the full range of that repetition (including its novel edge cases).If you want to defend the argument, you’d need to:Define “repetitive and predictable” so broadly it becomes trivial (e.g., “everything follows rules eventually” — which is metaphysics, not labor economics).Show that LLMs handle not just routine repetition but also the kind where the pattern changes unpredictably and requires genuine causal reasoning.Address the obvious counterexample: Can an LLM replace a trial lawyer in a novel, high-stakes case with no precedent? Likely not yet — and that’s white collar work.
@108810457>brown insect seethingImagine repeatedly failing at babby's first syllogism.
>>108810465Premise 1 and 2 are both wrong
>>108810465That's retarded. I know LLMs are shit but even I expected it to do better than that.
>>108810471I think it said something along those lines earlier in the response but i'm not pasting allat
>>108810492So the guy who's sperging out is wrong
>>108810471>>108810492>>108810500
>>108809913>if databases can't reason, then why can i get my account details if i query for them?
>>108810613Because queries don't require reasoning
>>108810613People underestimate how unoriginal their prompts are the same way they underestimate how many tor market databases their account details are on. It's a big internet. :^)
>>108810637Good job, retard. You actually grasped the obvious implication he was going for.
>>108810642Yeah but doing math proofs does
>>108810646Prompting doesn't make token guessers "do math proofs", thoughbeit. The blog post you linked to doesn't describe such an occurrence, even if its technically illiterate author thinks so.
>>108810674How is the Field's medalist wrong?
>>108810774By assuming that if his mind doesn't contain something that gives away the solution, 500 TB of data scraped from the internet doesn't contain it, either.
>>108810191I'm gonna impregnate your daughter
>>108811070bro wtf stop
>>108810637There is no technical argument for databases being unable to reason.
>>108810242hahhahaa this nigger believes random articles he reads online
>>108809913llms definitely have a logic machine in there. I just used an llm to unfuck my car's a/c. It's like an a/c wizard. I used gemini Pro version. \o/ idk if they all are this good at it.
>>108810323white collar != academiaI'll bet you don't even know the difference between black-tie and white-tie
>>108809913Because the proof is already written? It's like asking why a guy with an answer sheet can do a leet code faster than a guy without.
>>108810207>Yeah I didn't know anything about this, but the AI used this other dudes research and that's how it solved the problemWOW! impressive! Just like how I can write a compiler by calling cc.
>>108810450It's a pun you retarddivision, equality, set membership.
>>108809922They hated him, for he told them the truth.
>>108809913What you mean by thesis proof exactly?The AI wrote the thesis or it only calculated the formulas they fed to it?
Do you have esoteric industrial knowledge to keep your job safe from LLMs?
>>108809913>If LLMs "can't reason", then why do they keep doing PhD-level mathematical thesis proofs in hours?Usecase of reasoning for doing mathematical thesis proofs?
>>108813088>Do you have esoteric industrial knowledge to keep your job safe from LLMs?By AItard logic, I must have, because any real-world problem I throw at any piece of this garbage, it always fails.
>>108809913Because it's a pattern recognition machine and it recognizes patterns from other mathematical papers it was trained on.You can see it does not reason because it cannot count the Rs in "strawberry" without being trained on that first.
>>108813163>You can see it does not reason because it cannot count the Rs in "strawberry" without being trained on that first.That kind of failure is even more basic than reasoning. It can list all the letters and it can count, but it doesn't know how to combine the two things because that particular combination isn't represented enough in the training data, showing that it lacks a general grasp of at least one of the trivial concepts involved.
>>108809922why aren't you doing it yourself then>>108810017>humans dont reason. They do ion-based potential thresholds in a chain.
>>108814046>why aren't you doing it yourself thenBecause I don't care and it isn't useful. Also reading the rest of your post I realized you're either a low-tier LLM or a brown subhuman, based on your broken regurgitation of a low IQ corporate talking point.
>>108814056>Yeah man, Im not solving 2 Erdos problems a day because I just dont feel like it, tho I 100% could revolutionize mathematics if I felt like itlolmao
>>108809913>>108810158>>108810207>>108810323
>>108814064>incoherent bot drivel
>>108809913if they could reason you would not need to make these cope threads everyday and instead just cure cancer and make everyone immortal with a proomt
>>108814197>asking cancer to cure cancer using cancer
>>108813067>>108813072How was the proof already written? It came up with it itself. You're just coping
>>108813088No because you can just have it reverse engineer things and document it and read them into context
>>108813110The same type of reasoning applies to basically all white collar work
>>108815037>It came up with it itself.How did you establish this?
>>108815807The genius mathematician said so in the blog
>>108815796>The same type of reasoning applies to basically all white collar workNo, because some of it requires autonomy and common sense, involves responsibility and accountability and can't be encapsulated by a prompt. The only people who will be getting replaced by this trash are mentally ill NEETs like you getting paid 5 pennies per post.
>>108815811How did he establish it? Be sure to provide a quote, since I'm not reading technically illiterate blog posts. If you can't find one, consider taking your antipsychotic medication.
>>108815836>it requires autonomy and common sense, involves responsibility and accountability So basically an Indian getting paid $5 a day can do it
>>108815912So basically I accept your concession.
>>108809913Only Luddites and trans think that ai is useless. We vibeGODS are replacing em
>>108809913Math runs entirely on logic, the only time things become illogical is when people start imposing retarded rules like "Round 5 and above to 10" instead of something consistent like "Round to the nearest 5".
>>108809913Were not there yet but we’ll get there and the real horrors begin.Also nice anime girl I need a fit tomboy in my life. Even if she can’t fix me anymore.
>>108810017>linear regression in a chain.(You)r brain do same
>>108814046>Why aren't you doing it yourself thenI would just like to point out how fucking retarded of a response this is. He just said it is repetitive and predictable. Do you think that academia pays well or something? Why isn't he doing something boring and repetitive for very little money? Gee who knows?
