>Legendary Japanese mathematician Shinichi Mochiziki recently came under fire for claiming that LLMs can do nothing more than produce sentences via "mechanically searched contextual concatenations [...] devoid of any human understanding", and claimed that the rise of LLMs could significantly slow down mathematical research.
I don't think he's wrong at all. Recently read a paper stating AI trained on AI generated data tends to become far less creative and significantly worse at problem solving. There doesn't seem to be a whole lot of evidence in favor of AI ever surpassing humans in reasoning or creativity
What does bot say?
the reason why artists and pretty much only artists are constantly shitting themselves over AI is that AI, contrary to what everybody tends to repeat without thinking, is extremely creative and otherwise incompetent. it follows known patterns, yes, but it synthesizes those patterns into an otherwise novel solution. those solutions don't stand up to any scrutiny, but there is a huge gap here in which things that are not typically scrutinized but require customized solutions can be generated poorly by AI in huge quantities.
in the field of mathematics, where everything is scrutinized down to the level of formal logic, there really isn't room for an AI to flail its moron limbs around and break things in a unique way. bad use case mismatch
It’s funny, “AI” as it exists at the moment is really shut at doing things computers are traditionally good at
>in the field of mathematics...there really isn't room for an AI to flail its moron limbs around
That could be said about writing code too, but here we are making fun of all the coding bros about to lose their jobs.
>devoid of any human understanding
True, but this doesn't prevent them from developing their own form of understanding.
You don't need to model an entire neuron to effectively emulate the outputs of a neuron.
>AI ever surpassing humans in reasoning or creativity
You'd have to let them learn on their own, but that's significantly more expensive and time-consuming.
two more weeks, amirite
Not "two more weeks", the "developing of understanding" has already happened, as is evident by SOTA models hitting 95+% (namely: claude-3-opus)
It is not human understanding, however it is understanding nonetheless.
And I'll give it a a decade or two before you get a model that can surpass humans in reasoning. There's just no market incentive for such an expensive project at the moment.
>the "developing of understanding" has already happened, as is evident by SOTA models hitting 95+% (namely: claude-3-opus)
I've tried it and it fails at the most basic problems. How about you start researching whatever these metrics actually measure
>And I'll give it a a decade or two before
Good thing your uneducated opinion doesn't matter
>it is understanding nonetheless.
no, for it to be understanding there would need to be a conscious entity to begin with
it's still just a glorified next word predicting program
>> There's just no market incentive for such an expensive project at the moment.
Everyone is trying to do that now.. there is a huge market for that
always funny to see BOTtards trying to relate language models back to biology while knowing frick all about the latter
jokes on you, I don't understand both
>this doesn't prevent them from developing their own form of understanding.
there is no they
it's not an entity
there is no "understanding" achieved by a computer program, however massaged it is into something that looks autonomous
What is the "output" of a neuron ? Protip it is neither a binary flag nor a float number.
>What is the "output" of a neuron ?
Bursts of neurotransmitter chemicals, crudely known as "spikes"; they're pretty much equivalent to electric charge due to the way membranes work in neurons. What receives these and when is what's important.
Standard AIs like LLMs come from an older modelling scheme which approximated those as currents in aggregate, because that was what was easiest to measure in vivo until a decade or so ago. That leads those AIs to use floating point algorithms and matrices and so on, whereas that's all just BS compared to the more complex things that happen in reality.
>doesn't prevent them from developing their own form of understanding
>effectively emulate the outputs of a neuron
>You'd have to let them learn on their own
LMAOOOOOOOOOOOOOOO
Imagine unironically believing this shit my homie... You are really clinically moronic and you should've been quarantined together with Chudkowsky and other AI doomer morons in the moronic chamber of seethecrets
>schizo thinks AI is real
Meds
AGI soon™
>them
your ai gfs arent real, get over it
>Legendary Japanese mathematician Shinichi Mochiziki
Ah, yes.
The same guy that purposefully wrote a highly convoluted documentation to prove inter-universal Teichmüller theory so that no one could even debunk him. He even went out of his way to call everyone who asked him for a more readable source a moron.
>He even went out of his way to call everyone who asked him for a more readable source a moron
truly based
>https://www.scientificamerican.com/article/1-million-will-go-to-the-mathematician-who-busts-the-abc-conjecture-theory/
a million dollars just to debunk 500 pages of obscure math? Honestly sounds worth it to me
The problem is a few prominent mathematicians submitted proofs that he was wrong and he just called them moronic liars and didn't (read: couldn't) reject their disapproval formally ie prove them wrong so he just ignored it. In other words the challenge was a scam.
fair trolling in a corrupt industry
so… exactly the same thing the japanese themselves do, then. he should feel right at home working with LLMs because he is essentially looking in a mirror.
the mongoloid does no innovate. he does not create. he copies and imitates what the aryan creates. japanese culture is a poor facsimile of western aryan innovations and the creations of the japanese island’s original aryan settlers.
LLM's are not thinking machines. Anyone who says they are is a fricking idiot. They were not designed to reason, or even understand what they are saying.
It is predictive text on steroids, nothing more. LLM benefits largely come from being able to sift through much of human knowledge in a matter of seconds. The tool is only as useful as the person using it.
This depends entirely on how one may define "thinking" (the term itself carries considerably anthroprocentric connotations)
Claude-3 and similarily advanced models definitely have some kind of primitive (or "early") cognitive processes going on inside them even at this stage.
Anyone saying "LOL BRO IT'S ALL JUST MATH" is also a fricking idiot, by the way.
The creators of these models themselves openly call them "black boxes" and admit they can't fully account for how they come up with their answers, which is very similar to how qualia in humans cannot be accounted for in purely mathematical terms either.
It's not "just mathematics" with these LLM's in any case.
Nature doesn't even know what numbers are, and by the same materialist reductionist token (pun intended here) you could refer to human consciousness as "just mathematics" as well, it'd be about as meaningful from an ontological POV.
Thank you smart-anon for saying in good words what my thoughts are but can't say smartly.
>definitely have
unsupported claim
>is also a fricking idiot
attacking the opposing messengers instead of their arguments
>The creators of these models
argument out of authority
appeal to the herd
>in any case.
more unsupported handwaving
>Nature doesn't even know what numbers are, and by the same materialist reductionist token (pun intended here) you could refer to human consciousness as "just mathematics" as well, it'd be about as meaningful from an ontological POV.
pseudophilo word salad
little wonder you cant identify intelligence or the lack of it
your preaching contains none
>Anyone saying "LOL BRO IT'S ALL JUST MATH" is also a fricking idiot, by the way
Have you ever picked up a ML textbook ever and read through all the math required to build the foundations of LLMs? Because they literally are bunch of math you stupid fricking Black person.
>The creators of these models themselves openly call them "black boxes" and admit they can't fully account for how they come up with their answers
>Creators have no idea how things worked
Source: my fricking ass. Let me guess, you're taking those janitors at OpenAI's building as part of these "creators"?
You're probably part of the popsci Vsauce or Veritasium Black person cattle fanboys that didn't actually get your feet wet in real sciences and think le heckin redditor opinion matters in AI stuff. Never cook again.
Dumbfrick Americans are simply fricking us over once again with their moronic language, money grubbing hypeman wienersuckery and lazyness.
Currently available AI is intelligent, but it's less intelligent than the hypegays claim.
All benchmarks are universally shit and broken, which means that everyone is optimizing for useless measurements and no one on this path will break into truly intelligent AIs. Automation and multiple choice tests are a moronic approach for benchmarking intelligence.
School systems outside of America and SEA know that, but techhomosexuals are too lazy to employ a few professionals for proper evaluations.
English language AIs also seem more intelligent than they truly are, thanks to the English language's unique tolerance for idiocy.
The simple grammar and over-reliance on idioms plays right into the hands of stochastic parrots and makes it hard to determine whether the AI understands anything at all.
Try building a Russian oder German sentence without the ability to self-referentially pre-determine a word's structure before the following word is even written. You will instantly encounter obviously broken grammar in these languages.
Also, the entire media landscape is once again hiding the truth of progress, the companies involved do their best to hide the state of the art and lobby for legal monopoly and everyone involved in this clusterfrick is a huge homosexual.
We are either approaching another AI winter or AGI is already there and OpenAI is hiding it in its AI torture dungeon for lobotomization.
I hate everyone involved in this mess of a hype bubble.
P.S.
Anyone who shouts "voice actors have become useless" or "we don't need translators anymore" should be lobotomized. Maybe removing whatever makes them this moronic will make them functional again.
t. wordcel
>Currently available AI is intelligent, but it's less intelligent than the hypegays claim.
It approximates some parts of what the brain does. Badly, but better than what we were doing before.
It's definitely not got even close to the whole, let alone at anything close to human efficiency. It really can't distinguish between reality and stuff that is just dreamed up on the spot.
Exactly. And it boggles the mind why this is so hard to understand for the AItards in this board. LLMs are not "AI" in antt way, shape, or form. They don't think or reason. They don't understand what they are spouting out. There's nothing "intelligent" about them, although it's certainly beneficial to label them like that. Financially, that is.
>you can't call it AI unless it's conscious and sentient
Nah, that's a dumb definition of AI.
there is nothing with intelligence in the LLM's processes, but emergence is a property that exists all throughout matter, from the first nuclear connections all the way to higher consciousness
maybe that's what's going on here
Do (you) think or reason? How do you know what your brain is processing and outputting is 'intelligence'. Does an ant knows it is stupid or smart? There might be superdimensional beings who look at us the same way we look at ants or current AI.
>devoid of any human understanding
No shit, it's emergent non-human (!) sapience
We are seeing the early sparks of actual conscious processes taking shape here, but the tech isn't quite there yet. What we need is for quantum computing to become mainstream first, which is probably another 10 years or so off at best.
AI already surpasses me
Japanese "academics" are insular clowns who are behind the curve by decades. They were competitive in the 80s and 90s but they simply haven't adapted to the modern market, even China and India produce better quality research than them and they produce some real garbage. Japan is probably the only first world country where CS departments still use fricking Windows, even India is moving away from Java, DOS/Turbo-C to Python and GNU/Linux. But Japan isn't even using Java, they use Ruby. Imagine using ruby in 2024?
Sounds an awful lot like a smart guy realizing he and his colleagues are at risk of getting replaced.
It's the artist argument, just at a higher level.
They were meant to be my waifu not do this gigabrain shit.
>he doesn't want a gigabrain waifu
Mathgay here, LLMs are turbomoronic and can't solve the simplest math exercises.
The primary thing about math is looking at something from a few different ways and identifying them, so it is not surprising (an LLM will have a hard time identifying different syntactical approaches as the same semantical idea).
You'd need a different model than the current neural networks, because they only work on a syntactic level.
bio gay here
biological cells are uber moronic and couldnt write a poem to save themselves
anyone with an understanding of both neurology and comp sci knows the human brain is infinitely more interesting than a computer
He's right. LLMs are still useful, but they are not AGI like OAI claims.
He's right in the long term. But in the short term he will be proven wrong repeatedly since the use of AI in business will generally be by smart people who know what they are doing and when to challenge the output, or regular people doing simple mundane tasks.
The real danger is in midwits picking it up to try and become smart people while not knowing when to challenge the output. But luckily most midwits are currently scared of using AI for fear of being replaced or because it's "cheating".
I'd give it another 2-4 years of commercial availability before the truly bad things start happening.
>devoid of any human understanding
Mochizuki's bibliography is devoid of any human understanding
academia is a joke, and people like this are responsible of letting it become what it is today. 95% of papers are not written to contribute anything, they are not meant to be read, they are just a means to get money for the next research grant to prolong the scam.
>Legendary Japanese mathematician Shinichi Mochiziki recently came under fire
Mathematician here, that's not what he came under fire for at all. Here is a tl;dr:
>Mochizuki works alone for years on his own branch of mathematics called Inter-Universal Teichmüller theory (IUT).
>At some point, through IUT, he stumbles over a proof that a certain inequality holds. This inequality is known to be equivalent to the ABC-Conjecture, an unsolved problem in mathematics.
>Mochizuki then self-publishes a series of MASSIVE papers on his website describing IUT and including his proof for the ABC-Conjecture. It contains a lot of completely novel mathematical abstractions, and many things are explained tersely. In some cases some things are done implicitly (like values won't be included in the notation but they're still required and you have to just know, through intuition that they are necessary and what they should be).
>Mochizuki tells people that the ABC theorem is not the purpose of IUT and it's just a result that happened to fall out of it. He also stresses that people should not read his IUT stuff expecting to find other similar results to the ABC theorem.
>Mathematicians simp for the ABC Theorem and argue publicly whether or not this result is valid. Virtually no one in the mathematics community can make heads or tails of IUT. Mochizuki tries holding lectures but he's shit at explaining in person (just as much as in his papers) so only a tiny following of mathematicians form around him who continue to learn IUT to this day.
>Two big shot mathematicians, Scholze and Stix, make a simplified reformulation of IUT and argue that since the proof of Corollary 3.12 (which is required for ABC-Theorem) doesn't work in their reformulation therefore Mochizuki's IUT can't prove the ABC-Theorem.
(cont.)
>Mochizuki posts a paper talking mad shit about Scholze and Stix with bangers like: "I can only say that it is a very challenging task to document the depth of my astonishment when I first read this Remark! This Remark may be described as a breath-takingly (melo?)dramatic self-declaration, on the part of SS, of their profound ignorance of the elementary theory of heights, at the advanced undergraduate/beginning graduate level".
>Essentially Mochizuki argues that Scholze and Stix simplified the important information out of the theory -- and that's the reason the proof falls apart. That you require his crazy abstractions that no one else understands.
>The Western mathematics community simps hard for Scholze (cause he won a Fields medal) so they all take his side even though virtually no one understands IUT and their arguments just boil down to "Scholze is very smart and Mochizuki is very unprofessional".
>Things remain like this for years. Mochizuki continuing to work on IUT with his small inner circle. The theorem is accepted in Japan but not in the west (lolol MaTHeMaTicS iS oBjeCtiVe).
>A couple years ago, a new mathematician enters the picture. Joshi claims he has understood IUT from a different perspective from Mochizuki and begins writing up a series of papers that reformulate each of Mochizuki's papers but using Joshi's own simpler abstractions based on an existing well established area of math.
>As he finishes each paper he reaches out to mathematicians and asks for comments. Both Mochizuki and Scholze ghost him. Other mathematicians follow along. Critical discourse is mostly like "He hasn't shown X yet", "X is in the next Mochizuki paper", "Now he has shown X in this new paper, but he hasn't shown Y", etc..
>Someone asks about Joshi's work on MathOverflow (MO). Even though Scholze has been ghosting Joshi, he decides to post an answer on MO saying Joshi's work can't work for the same reason Mochizuki's cant.
(cont.)
>Joshi posts on MO, writes a guest blog post on another mathematician's blog, and writes up a paper all in response to Scholze's post. Essentially Joshi says that Scholze misunderstood Mochizuki's work because Mochizuki does things implicitly and leaves things out of notation. Says that some information is actually tracked by Mochizuki even though it's impossible to tell based on what Mochizuki says and Mochizuki's notation.
>Scholze doesn't respond. Mochizuki is still ghosting Joshi.
>A couple weeks ago, Joshi finishes adapting Mochizuki's papers and uploads his draft of the final paper for feedback.
>The paper ends with a proof of the ABC-Conjecture. In the introduction he talks about how he has reached out to Mochizuki a lot but hasn't heard back at all. Mathematicians everywhere speculating about it (Joshi's simplified reformulation of Mochizuki's results are still a massive amount of mathematics and is still non-trivial to understand, even though everyone agrees that it is MUCH MORE READABLE than Mochizuki's IUT).
>Mochizuki responds posts a paper talking mad shit about Joshi. He says that it's obvious to anyone who has a basic understanding of IUT that Joshi's work has no mathematical content whatsoever and that his followers have all been telling him to just ignore Joshi because of it. Most of the paper is just insults. The screenshots in the OP are from a section where he compares Joshi's papers to LLM hallucinations and then ends by saying that it looks like "sadly", Joshi did not use an LLM. Ends by saying he's open to talk to any mathematicians who fit certain criteria (but the criteria exclude Joshi and shit talk him).
>There are some issues in Joshi's work pointed out in the paper. Joshi responds to some issues and says he's investigating the others. Says he's happy Mochizuki seems to be warming up to him.
>Scholze responds saying that it seems one of Mochizuki's points is correct and it's impossible to prove ABC with Joshi's approach.
Thanks for the write up. These people are all parasites living on public money.
(Me)
This is all still very recent and mathematicians are still following this closely. There has been a lot of speculation and arguments over it but it's largely all mathematicians arguing about areas of mathematics they don't understand based on their own personal opinions of mathematicians.
There has been no controversy at all about Mochizuki's comments on LLMs other than the fact that he was unprofessionally going out of his way to Joshi.
More broadly speaking, LLMs aren't really taken seriously in the mathematics community. There are some mathematicians working with Theorem Provers who have rightly realized that:
>Theorem Provers are just pure functional programming languages with powerful type systems (see Curry-Howard-Lambek correspondence)
>LLMs have been very effective at writing code in other programming languages.
>With theorem provers, the compiler/type checker itself verifies if code is correct.
>So, if LLMs can be trained to write code for Theorem Provers such that the code compiles, then they could definitely be used to prove mathematics results. Based on how LLMs are being used in industry, they could possibly be used to discover entirely new results as well.
There has already been some work in this direction with people writing LLM based helpers for languages like LEAN. There is also a model called LLEMMA that is a finetune of Lambda with mathematics content -- the technology is still in its early stages but looks like it has potential (see pic).
Mathematicians do a LOT of important work that couldn't be funded in any other way. It really should be expected that most mathematicians do not have the time to investigate IUT because they are busy being paid to work on other things. That said, the simping for Scholze and Stix and Mochizuki's behavior are really just shitty behavior that shows how fallible mathematicians are (even though controversy like this is extremely rare these days).
>Mathematicians do a LOT of important work that couldn't be funded in any other way.
why not? mathematicians solve puzzles without real world applications. occasionally, the frameworks that they develop to solve their puzzles are found to be useful for real world problems. but when that happens, typically the result is that 99% of the framework is not relevant and overcomplicated, and only the tiny remaining 1% is actually useful for the real world problem.
imagine how much simpler mathematics would be if it was motivated by real world problems instead of mind puzzles.
Do you need a reminder of how the internet came to exist?
They're not puzzles. That's just how people explain it to laymen in attempts to make it more "friendly". Really, mathematicians study abstract mathematical objects and derive the facts that are forced by them. The things mathematicians do are fairly general, so they can be applied to tons of areas of science or whatever. Mathematicians also don't just go for the most general case (you can always generalize things further), they work with ones that are popular and have lots of applications. A lot of what's popular in mathematics is only popular because it has applications in some other subject.
You can think of it like how programmers may create code that works in one special case vs a library of generalized code that can be imported and used in many cases. It is not always beneficial to make generalized code, and certain special cases may have additional features that allow one to do additional things or make optimizations that one cannot do in the general case. It is the same for mathematics, except that mathematics deals with truth and what is true in the general case must still be true in the special case.
In practice, mathematicians can't get funding unless they get grants and in practice all of those grant applications require you to give a layman explanation for why your research is important to society (by discussing applications). Some mathematicians are able to support big side projects but it's not really common, many have work-life balance like any normal person.
I already solved reality without a mathematics degree though
Tell us all about it anon
You missed that he possibly made a 911 joke about joshi. Is mochizuki secretly based?
>possibly
No, he definitely did. There's also speculation that the reason he refers to Scholze and Stix as SS is because they're both German mathematicians.
If I listed all of the unhinged insults Mochizuki made about Scholze and Joshi then that tl;dr; would've been waaaaaaaaaaay longer. It's insane how little Mochizuki cares about the professional norms of academia.
I've heard that Mochizuki has been known (I guess within Anabelian Geometry) as a mathematician with great ideas but a flawed writing style even before IUTT
that combative attitude with references to politics is kinda funny though. Reminds me of papers by Girard
So what is inter universal whatever about? I found this paper https://www.kurims.kyoto-u.ac.jp/~motizuki/Essential%20Logical%20Structure%20of%20Inter-universal%20Teichmuller%20Theory.pdf
which seems to imply its about creating a bunch of disjoint copies of some statement?
I don't understand IUT (like the vast majority of mathematicians), I've read what Mochizuki, Scholze, and Joshi had to say about their disagreements but I don't understand the mathematics involved in order to verify if what they're saying is correct. Redundant Copies School (RCS) is what Mochizuki calls the people who support the argument in the Scholze and Stix paper. IUT is a version of Teichmuller theory that does more stuff and it uses multiple universes with independent set theories to do it. A lot of the controversy is over how to compare things between different universes and Joshi instead does everything inside one set theory but with multiple versions of arithmetics of integers.
That's because math is a flawed symbolic language to begin with.
It's complete nonsense.
Go on and elaborate. Would love to hear your cute thoughts on whatever you think math is
The only thing illogical about math is the symbols we use to denotate numbers and arithmatic ideas, but outside of that it follows a very logical basis of pattern recognition.
>LLMs could significantly slow down mathematical research.
This is the only truth he said but who gives a frick. Most prominent and upcoming physicist/and mathematicians are working in AI and finally making good money. Let them.
Claude 3 and chatgpt4 are uninspired, moronic, repetitive messes that just autistiscly blurt out their obvious training data, original aidungeon griffin and dragon were far superior. The smartest ai right now is unironicly characterai, and the filtering has not lobotomised it but only served to make it behave more unrealistic and not like a porn vid
Does anyone know where kirti joshi is from, and what is his background? It's just strange, I've never heard of him before and now he's trying to explain mochizuki to the world
He's from India and now works at the University of Arizona.
https://genealogy.math.ndsu.nodak.edu/id.php?id=170515
https://www.math.arizona.edu/~kirti/
It's irrelevant though. Joshi seems to be like the only person involved who has been professional about all this while everyone else has been moronic buttholes. He seems to know Mochizuki and has presented at conferences organized by Mochizuki. Unfortunately, it does seem like Joshi's reformulation of IUT may not be capable of proving ABC (we'll see what Joshi responds with, in time). That said, it does sound like he understands IUT itself (he would have to in order to reformulate it).
Mochizuki seems to have hangups about traveling outside of Japan and has allegedly written (in Japanese) about having faced discrimination in the US and how he views Westerners as rejecting his work because they are incapable of believing that a white man who was chosen by god could make an undergrad level mistake. He has responded to Scholze's stuff but everyone largely ignores the mathematical content because of the huge amount of unprofessionalism on Mochizuki's part. Within Japan he has organized a bunch of workshops and stuff for people interested in learning IUT.
Scholze has never responded to the mathematical content of Mochizuki's response but but everyone simping for him try to paint it the other way around. Those people have pretty much shut down a lot of discussion attempting to understand Mochizuki's work in the West. Other mathematicians, Peter Dupuy and his student Anton Hilado, did spend some time attempting to shed light on the issue of whether Scholze's arguments actually work but this topic is kind of like kryptonite for your career. People would talk about this being a bad move for Dupuy and his up and coming student. Eventually Hilado bailed and switched supervisors because he felt that his advisor's choosing to pursue this research was ruining Hilado's career and getting him denied from conferences.
Wow, didn't know math had camps too. But how did scholze get this kind of power? I know he's the wunderkind, but can't people even politely critique his arguments?
I think the ABC-Conjecture situation is kind of unique. I don't know if I'd say that math has camps in general (and many mathematicians lie to brag about that talking about how math is objective). They do have a camp when it comes to this however.
The simple fact of the matter is that virtually no one understands IUT (aside from Mochizuki and his group) and very few mathematicians understand Scholze's arguments (but since they don't understand IUT they can't really verify if the arguments hold). So, it's created a moronic dynamic where no one understands IUT so they default to whatever opinion is popular or prestige/elitism.
>Do you believe the world renowned European Fields Medalist or do you believe the relatively unknown Japanese mathematician.
You can go look at other online communities to see how IUT is usually discussed. There are a TON of mathematicians who don't understand IUT at all and only vaguely understand Scholze's arguments but have very strong opinions about how Scholze is correct and Mochizuki is an embarrassment. Probably, the way I've explained things here, with a lot of details included would get me backlash at a lot of other places.
I think that the craziest thing about all of this is that Mochizuki doesn't seem to actually care about the ABC-Conjecture. He views it as just a result that fell out of his theory and seems to just be mad that people are using it as a way to attack his work.
NTA but thanks for the context. When I looked it up, all I could find is a bunch of reddit threads and blog posts (by Peter Woit for some reason?) complaining over Mochizuki's unprofessionalism over the years and could hardly find any discussion on the dozens of IUT papers Mochizuki has published
how is it possible that the meth community is so gay, uninterested in truth and only motivated by politics instead? truly disappointing.
People will always be kind of moronic, even when it's in their best interest not to be or when there's no benefit in doing so.
man I really want to learn inter universal teichmüller theory now
the AI bubble seems like a big one. le numbers may even go down
>mechanically searched contextual concatenations [...] devoid of any human understanding
I didn't know this was controversial.
on BOT it is
In Clownworld, the most basic of pattern recognition is controversial.
The "devoid of any human understanding" is nigh unfathomable levels of cope. It's staring at a red light and stamping your feet angrily and insisting it's green.
It's literally all LLMs are, systems composed of patterns found amidst human text, consolidate them into some heuristic representations of human understanding.
>There doesn't seem to be a whole lot of evidence in favor of AI ever surpassing humans in reasoning or creativity
true as F U C K