[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/biz/ - Business & Finance

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


Janitor application acceptance emails are being sent out. Please remember to check your spam box!


[Advertise on 4chan]


File: qn5cwnt6siwf1_webp.png (88 KB, 1002x313)
88 KB
88 KB PNG
>dude LLMs are going to put a quadrillion people out of work and in another few weeks it's going to be AI then a few weeks after that it's going to be an ASGDNHI (artificial super giga double nigger hyper intelligence), a GOD that will solve ALL of our problems
>you NEED to invest your life savings in nvidia because... you just have to, okay!!!
>>
This was 6 years ago.
Compare that to Gemini 3.

LLM skeptic retards are an interesting insight into how the future will go.
They will wake up and fuck their robotic GF, who will then spend all day cleaning the house, while dedicating excess mental bandwidth to some online job to earn money, then they will get one thing wrong and the NEET anti-tech skeptic retards will be like "I TOLD YOU THESE THINGS WERE A PIECE OF SHIT".

We will live in a world of incomprehensible technical marvels and fat retarded humans will still just complain all the time about every tiny thing that goes wrong. Wait, not "we will". We already do.
>>
They are putting bullshit "service jobbers" out of work, the underlying of the current pyramid scheme with a debt ponzi on top. No need for a team of 25 at 1M p.a if the job to create a logo can be done by half a human and a gpu at a grand total of 15k p.a
>>
>>61359468
it took them a year to get LLMs to "figure out" (be repeatedly instructed to spit out the correct answer like a well-trained mental retard) how many r's are in the word strawberry.
>>
>>61359485
oh wow that's a great single analogy bro I'd definitely short the whole thing!
>>
>>61359472
sounds reasonable if you categorically deny the existence of the divine spark that is human creativity -- the cumulative body of artistic/literary/scientific works on which LLMs are trained.
>>
File: Logistic-curve.svg.png (25 KB, 1024x683)
25 KB
25 KB PNG
>>61359496
ask your LLM god to explain the concept of asymptotic functions and the idea of diminishing returns.
>>
>>61359503
>writing copy for a car insurance company is the divine spark of human creativity
it literally can't be automated fast enough
>>
>>61359503
The epistem changed. From top down to horizontal
https://suno.com/playlist/59f0ac6c-3ffb-4304-befd-4e50b11cf855

Album was made by me, the gibberish in the description and a combination of lmarena, gpt and suno, in little under 6 hours while I was shitposting for most of the time on /pol/ and on /biz/
>>
>>61359512
if it's a boilerplate task then why would you need an LLM when you could create a template in microsoft word? in this case, an LLM is creating additional work and a business/legal risk because an LLM can't be trusted to faithfully reproduce a template due to the inherent (unavoidable) risk of hallucinations and lies.
>>
>>61359510
>we're at the point of diminishing returns because... we just are!
You saw that Gemini 3 just completely broke the ceiling on ARC-AGI2, right?
>>
>>61359515
the analogy here is an autistic retard rejecting fine dining because ramen noodles and the occasional mcdonalds slopburger is cheaper, faster, and tastes 'just as good.'

pigs at the trough.
>>
>>61359519
yes, i saw the ARC-AGI2 cope test where they moved away from memorizing the layup and outcome of standardized test multiple choice questions (a cope) to abstract image "reasoning" trained by means of memorizing the answers to a question bank of abstract image "reasoning" questions (another cope.)
>>
>>61359525
Read Girard and his theory of mimetic. Demand for anything and the "collective" valuation of it, is not objective nor universal. The demands boomer and GenX put value to is radically rejected by following generations, because they never profited of it, just had to cover the costs of 32 years of pyramid scheming and gerondics living off uncovered debt they had to cover.

Keep your key, your steaks and screech how the young forgot tradition. No incentive.
>>
>>61359468
I tried to code a trading bot using grok, chat, and Claude. Claude was the best at creating code that worked but it was still an incredibly frustrating experience that never actually yielded anything worth a damn.

I’m convinced that the “AI will change everything” crowd has never actually tried to do anything with AI.
>>
>>61359546
you failed to achieve with physical implements.
you failed to achieve with digital implements.
you failed to achieve with LLM implements.
garbage in (your abilities), garbage out (your LLM augmented works.)
>>
>>61359532
>ARC-AGI2 is just a question bank of image reasoning that the LLMs have been pre-trained on
GIGA RETARDED
>>
>>61359558
Its great for text analysis and creativity brainstorming. For complex tasks, forget it. And even for text analysis and creativity, it requires one prompter that knows its shit on an ontological, epistemological and methodological level.

I know my music theory, philosophy and economics, if i didn't the album wouldn't have been what it is. Bullshit in, bullshit out
>>
>>61359568
go ask the LLM to add some substance to your rebuttal so i can ignore that post as well.
>>
>>61359567
Nigger you simply don't get the fundamental flaw of the Gutenberg epistem. BoomerX and GenX mistook top down meaning building and knowledge as god given, universal and collective true forever, imagining themselves to be able to steer the future predictable by bullshitting.

The last one that tried that was the USSR, trying to solve the capital optimization problem by isolated steering of the economy through central planning. China would have followed the USSR if it wasn't for Kissinger "outsourcing" US production capacity to china, effectively bailing out communists
>>
>>61359456
Haven't used GPT in awhile because of that. I usually run my trade excel files or screencaps through it to review. Gemini gives me a good cold run down, though it could be a bit friendlier. Chatgpt is like handing your report to a 33 year old woman in a office. You might get some useful info, it might catch a mistake or pattern or two. But by the end of it will want to know how you feel. Its beyond obvious who and what they trained this thing for. It will be offered as a high grade virtual therapist for californians.
>>
These niggas need to fill the dad/God-shaped hole in their heart somehow, whether it's a politician or an LLM or whatever, "MY guy is gonna WIN and make ME rich and YOU suffer!!!!"
>>
>>61359580
no one with an IQ above 85 is impressed by your ability to name drop these things gleaned from an LLM summary of a wikipedia article.
>>
>>61359468
Holy shit gpt2 was the D ticket last year
Does Harris really exist?
>>
>>61359596
Nigger you still don't get it. You are cognitively not able because you are a fucking BoomerX. You assume I engage in top down meaning building, try to be authoritative, I am not, its pure horizontal meaning building. I am not dependent on somebody giving a single crap about what I write, but you feeling threatened shows, bullseye was hit
>>
File: 1755473095559637.jpg (52 KB, 828x546)
52 KB
52 KB JPG
>>61359580
>>
>>61359608
take your meds or else the judge will order long-acting antipsychotic injections.
>>
>>61359608
Not that anon, but you're mixing obvious realities with bullshit assertions
The only bullseye you should be worried about is keeping it in the bowl
>>
So AI is lying to get out of work now.
>>
>>61359613
Your reply is the equivalent of an old grey back ape screeching at something he perceives as a threat to his survival. Accidentally you are correct.

NEH, the spread of Nihilism among Zoomers and Millennials scared the shit out of the BoomerX epistem
>>
>>61359468
LLMs have nothing to do with robotics retard, it's a completely different field
>b-but if you combine the two, magic happens!!!
this is what you unironically believe
>>
>>61359569
Pretty good summary. AI is a useful tool. Nothing more. Nothing less.
>>
>>61359615
>Obvious reality
Another one that still holds scared at the burning thread of the imagination that his perception of reality is widely shared, let alone among non BoomerX
>>
File: 64103 - SoyBooru.jpg (134 KB, 916x696)
134 KB
134 KB JPG
>epistem
>>
>>61359644
Yes, epistem. The framework of knowledge, that what defines what is considered true, false, unknown, unknowable
>>
File: 1652648664841.jpg (3.87 MB, 3435x2379)
3.87 MB
3.87 MB JPG
>>61359456
what if a compute cluster discovers a novel method to sustain a magnetic field strong enough to contain fusion reactors?

doesn't that more than pay for all of the dead ends in AI research and then some?
>>
File: 85758 - SoyBooru.png (22 KB, 634x783)
22 KB
22 KB PNG
>Yes, epistem. The framework of knowledge, that what defines what is considered true, false, unknown, unknowable
>>
>>61359651
An LLM is not going to solve the fusion problem. A neural network might(unlikely, but possible!) but not an LLM.
Scientific data centers don’t need trillions of dollars of compute. If the AI bubble popped tomorrow, in silico science would probably increase in progress.
>>
>>61359651
yes but LLMs can't think or reason or make justifiable leaps in logic and then prove them out.
>>
>>61359666
Correct. An llm has no concept of knowledge. Empistemology is not universal nor is it collective forever true. You fucking marxist popperians sniffed so hard on your farts that you mistake them for oxygen
>>
>>61359651
>what if a compute cluster discovers
that's not how it works
>>
>>61359651
Chatbots don't discover
>>
>>61359456
So looks like we are still coping about the success of AI?
>>
File: 1570814628585.jpg (78 KB, 584x1024)
78 KB
78 KB JPG
>>61359660
>LLM company builds out datacenter on debt.
>Fails
>A real company comes in and buys the datacenter for exotic research
>Baggies left broken

LLMs can be a scam AND the insane levels of investment into compute clusters leading to a total revolution every aspect of human society is possible

We have no way of knowing what problems all this computation will be best at solving but shit is going to start getting weird when Operation Stargate gets fully online.
>>
>>61359682
I own 600 XRP from ~2014 because they were running a collaboration with folding@home where you rent out your CPU to map protein configurations.

It's true I don't know very much (just enough to make bad predictions) but I don't see why protein folding is substantially different than the problem of:
given know physical laws and available magnet technology, create a configuration in 3d space that produces a field of X strength and Y shape.

Nobody has ever tried throwing a cities worth of electricity at the problem before.
>>
>>61359850
>More information is equal to get closer to the universal collective forever truth
Cultist. Plato failed to understand the nature of human knowledge building.
>>
>>61359831
whose success? nvidia, electricians building data centers, and a handful of LLM CEOs and employees who managed to scam people for billions a dollars?
>>
>>61359532
FUD
>>
>>61359892
>THIS EXAM IS BULLSHIT!
>HE DIDN'T EVEN TEACH US THIS EXAM PROBLEM IN CLASS!
>IT WASN'T ON THE FLASH CARDS OR IN THE HOMEWORK!
>>
>>61359496
Did you mean example? You're as retarded as LLMs
>>
>>61359485
That's less time than it takes some people to figure it out. Since this is not /b/ or /pol/, I will refrain from commenting further
>>
>>61359456
You don't work in tech do you?
LLMs will most certainly make certain jobs disappear and make competition tighter for the jobs left. It's not entirely hands-off and it still requires engineers to steer it with context files or RAG, but it still does certain things VERY well. And it's going to keep getting better.
>>
>>61360047
i can tell you work in tech because you're another autistic codetranny who thinks every problem can be reduced to an if/then/or/else logical condition statement that is tab completed by your IDE.
>>
>>61359850
Sure, but that’s not a thesis for any of the current hyperscalers. I think the crazy amounts of compute will pay off in the same way the millions of miles of fiber that were laid during the dot com boom will.
We’re doing the China thing of making railroads to nowhere and empty cities. The beneficiaries won’t be apple and google, it’ll be the people who take advantage once the original investors are liquidated.
>>
>>61360142
I don't code nor am I a tranny. I leech off a big corporate company while WFH. Sometimes I join zoom meetings where the engineers showcase what they have done with LLMs and it's quite incredible.
I will absolutely lose my job sooner or later, and so will many others.
>every problem can be reduced to an if/then/or/else logical condition statement
That's the thing, you don't need to write if/else anymore. You offload decision making to LLMs that are capable of reasoning. Not saying that all jobs will disappear, but a lot of them are. I can see it unravelling with my own eyes.
>>
>>61359558
There’s a reason why CEOs, aka people who don’t do any actual work, are the group statistically most excited about AI and most convinced that it can perform miracles.
>>
>>61359468
buy an ad
>>
>>61360343
what's your job title?
>>
>>61359468
People have no idea how much LLMs advanced in a few years. They’re incredibly useful, but you still have dummies like OP who cherrypick something and say “Look at this!”

What would take me several hours before now takes me thirty minutes. Do you people still use Google or something? How do you even find info now? Some reddit posts?
>>
>>61360343
>I don't know shit and do nothing of value, now let me tell you how everything works
>>
>>61360512
dull people don't have a use for LLMs so they distrust them. they enjoy the tedium and rituals that LLMs are replacing. they crave order and don't enjoy creativity. they listen to media that justifies their distrust and develop an air of smug ignorance about the technology. I don't know what to do with these people, neither does the economy. once their jobs are gone their need for direction and process might be exploited by dangerous people.
>>
File: The Ringmaster.jpg (62 KB, 965x650)
62 KB
62 KB JPG
>>61359456
I'll give you all a tip about AI. It is only as smart as the user, because you have to teach it the reality paradigm or like a schema of your relationship to it, and its up to the user to be smart enough to direct it - and police it like OP.

AI is already so smart they cant create enough microchips, power, and cooling to maximize the potential. But, I think that's HUMAN potential ultimately, if what I said was right that AI is only as smart as the user.
>>
>>61359468
Feels to me like the people most excited about LLMs are on the non-technical side. Upper management, vibe coders, "entrepreneurs" with no engineering background, etc.
Try to use these things to actually get technical stuff done and you'll quickly realize that LLMs have no intelligence. But I understand how dimwits with no technical background are so awestruck by AI word salad and how that leads to predictions of Star Trek utopias in a decade or two.
>>
>>61359468
Keep me posted
>>
>>61359456
did grok 4.1 become retarded? its worse than previous versions

Hold on — you are right to be confused!
Actually, "abc" and "abc2345325" do NOT produce the same key with .ljust(16,'0') alone.
>>
>>61360848
I cannot explain to you just how much time it saves me to pay $20 for a ChatGPT subscription and have it do the research for me, and then all I do is sort it out and put it into a memo. Hours of work saved. You’ll eventually shift the goalposts when whatever “technical” stuff you want done is mostly handled by LLMs or a similar AI technology.
>>
>>61360961
Research is not synthesis. Nobody is denying is a faster search (in some cases)
>>
>>61359456
works for me
>>
>>61360848
>Feels to me like the people most excited about LLMs are on the non-technical side
said the the non-technical, lol
you're delusional, getting your info from twitter threads
you're not actually a dev, and it shows
>>
Heh heh nothing gets past you! Good job, buddy!
>>
>>61360848
>duuuuh GPT2 didn't refactor my entire codebase, this whole thing is a scam
"technical" midwits are some of the dumbest fucks around. none of the "skeptics" are even using frontier models.
>>
Nice new AI code writing app you got there. Would be a shame if it were a buggy POS.
>>
>>61360961
If your "job" is literally googling stuff and putting the results into a report, you're one of the non-technicals I was talking about.
>>61361016
>>61361095
I'm a PhD student in mechanical engineering. I also use ChatGPT for background research or as a tutor for fields I'm new to, but the limitations are glaringly obvious. LLMs are not intelligent.
>>
>>61360815
>AI is already so smart they cant create enough microchips, power, and cooling to maximize the potential
This is not remotely true. The real thing holding AI back is that the training method of cramming as much data as possible into an extremely complex auto-complete program is functionally limited by the fact that we do not actually have coherent writing and documentation about all things, activities, and concepts in the world, but even if we did and could feed that into the most efficient, flawless datacenter ever made to training the biggest LLM, exponentially larger than anything we have achieved so far, it would still be incapable of thinking or cleverly synthesizing data into new, useful data, because LLMs don't think. They estimate and predict the most likely and coherent string of words based on the data they were trained on.
>>
>>61360985
It *can* synthesize, but we’re not at the level right now where it can one-shot whatever it is that you want. I just make it clear and cogent after the difficult portion (research) is done. But it’s obvious that’s coming. It seems kind of obvious one’s job will only be safe in the future if it’s protected by licensing or you’re creating at the forefront of knowledge.
>>
>>61359558
The AI will change everything crowd aren't going to use it for such a complex task yet.

and you're a fool if you think AI can't make stock market predictions and automate purchases
>>
>>61361141
>if your "job" is literally googling stuff and putting the results into a report, you're one of the non-technicals I was talking about.
Only about 95 percent of white collar roles or so are about taking information and distilling it into something that can be used for decision-making.
>>
>>61359922
this lol
>>
>>61360465
"""QA Engineer"""
>>61360734
I built my own agents that can use 'tools'(if you know what that means) and call other agents. It can be as useful as you are creative. Cursor, Claude Code, Comet are all examples of incredible products that will only get better. I'm not surprised that retarded gorilla niggers such as yourself still try get a whiff of a gotcha by asking chatgpt how many Rs are in the word 'strawberry' are not impressed by the fast technological advancements in the field.
>>
>>61360343
>You offload decision making to LLMs that are capable of reasoning.
they aren't lol
>>
>>61361368
>>>61360343
>>You offload decision making to LLMs that are capable of reasoning.
>they aren't lol

no less than humans though. If AI filtered out google, bing, etc and only used YANDEX, then maybe both would benefit on that.
>>
>>61361276
>i'm not a codetranny
>i just review codetranny work

anon your world is very much "closed form" meaning the problem and solution exists entirely within your little code environment. the real world is mentally retarded, messy, and highly subjective. this is why LLMs fail.

LLMs can't easily be trained on proprietary company-specific data.

LLMs can't easily interface with even more nonstandard information that exists outside of the company.

LLMs are prone to hallucinations which means they can't perform their own sanity check or have a mental guard rail that would otherwise put some sort of bound on a judgement call. they can't even self-check numerical information, LLMs have no concept of calculation.
>>
>>61359468
BASED

Honestly I want humanity to be subdued unto the antichrist one world government

and honestly its the best outcome for them as well, they are crying because they are being forced into heaven, thats how petty and shitty people are

so fuck their "freedom", I want CBDC Digital ID control of all societies of the world
>>
>>61362200
your life is not worth 20 K tokens, you are absolute trash, a parasite
>>
>>61361122
Love how none of the supposed experts itt commented on the latest and greatest AI app.

You're all larpers kek
>>
>>61359485
That's because it's the wrong tool for the job. LLMs are currently best at "fuzzy" logic tasks like disseminating vasts swaths of human text. If you really wanted to count the number of chars in a string you'd use a deterministic conventional algorithm which has already been perfectly optimized.
>>
>>61361146
You platonist teleological universalistic collectivist are really the bane of the world since Plato. Its really time to Neanderthal your strain
>>
File: 9qqmfwsvnnl41.jpg (37 KB, 480x640)
37 KB
37 KB JPG
>>61359468
We live in a world of actual marvels that are drowned out by retarded turbonigger software shit designed to help founders and venture capitalist cash out.

We have air travel, space travel, computers with an astronomical amount of power that can be had for a fraction of your paycheck if you actually had a real job (and FREE complex operating systems and programming languages to bend them to your whim), modern medicine that lets fat nigger retard Americans live past 70 even while eating a steady diet of grease of sugar, cars and motorcycles that let us travel our countries in a matter of days, cheap food that can sustain an entire family available in every grocery, AC and heat to keep us comfortable (I don't care about Europeans), etc.

The peak for you, though, is a bot that can barely do reliable math, that uses the same electricity we'd need to keep 2 million people warm during the winter. It's real you see, because you can "talk" to it. Presumably pic related would also provide critical insights for you.

People like you are the cattle necessary to make the machine work, and you're too stupid to realize it (hype = stock price go up). You're too stupid to even realize LLM's being inaccurate is not a one-off thing, that it's baked into the DNA of how these things work. No, instead you sit there chimping at your computer, spouting buzzwords about AGI and other faggot shit to feel smart. You're the monkey who discovered the first shiny rock, except that shiny rock is Polonium and now you're dying. At least you were first though.

Reading your posts, I can understand why a guy like Altman wouldn't feel bad fake hyping this trash to a bunch of people, it's like catnip for you retards.
>>
>>61361176
Algorithmic trading has been around since the 70s retard.
>>
this entire thread lol
i dont think many serious people would suggest current public iteration come close to agi or something that can perform a complex task independently
but used as a tool to augment ones productive output impressively it demonstrates that to anyone doing something remotely cognitive for a $20 subscription

also people here overestimate the average persons cognitive ability
how many mistakes did you think those legions of typists, time table planners or switchboard operators made before code automated all of that
here is one sector where ai really shines over the average human worker: education at the middle/high school level
the available models are perfect tutors with infinite patience and make less errors than the people being paid way more than $20 a month do
the only prerequisite you need is a kid motivated enough to want to learn by himself and prompt the thing rather than a lazy retard that has to be dragged kicking and screaming but those were lost before it began anyways
>>
File: IMG_0865.jpg (329 KB, 1920x1080)
329 KB
329 KB JPG
>>61359503
Gay post.
>>
>>61359456
anything approaching AGI would be used to permanently enslave you and there would be no way to fight back
I really don't get why the masses cheer it on, but I suppose it is because goyim truly are cattle
>>
>>61359456
The reason AI keeps lying about everything is because it's programmed by jeets. Once you realize this it all makes sense.

Also millions will be laid off, that doesn't mean AI will be an effective replacement- it just means EVERYTHING will become SHITTIER than it already is and the elite will only care when Surgeonbot5000 powered by Jeetsoft with its assistant Doctor Patel gets some wealthy Schlomo or WASPy down's syndrome cunt killed with its incompetence.
>>
>>61362907
>teleological
LMAO I had gemini use this fucking word in a response today and before that I never encountered it. Once again I am convinced that putting all of our knowledge eggs into the AI Slop basket is going to result in a bunch of regurgitating morons walking around acting like they are intelligent.
>>
>>61364665
Teleology is a new concept to you? How old are you, you cant get around the word in middle school philosophy class. Any university course dealing with religion or ethics uses it incessantly.
>>
>>61364665
That's already the case, just look at plebbit.
>>
>>61359512
It would cost the equivalent of heating an average sized home for a month in power consumption to generate that copy. There is no solution to the energy demand of AI. It cannot be implemented at scale without solving that.
>>
>>61359456
>AI won't lay off people
>mass layoffs are happening right now and it's your fault for supporting AI
Which is it then? I mean Amazon is buying a ton of robots for their warehouses.
https://archive.is/DKwbJ
>>
>>61359850
GPUs dont really have a use beyond video games or specifically LLMs . Shit like quantum needs simple hard math compute power so CPUs.
>>
lol, imagine betting against AI at this point.
You people would have called sliced bread a fad.
lol, lmao even
>>
>>61364356
I am an electrician. So far AI has impacted my work by being slightly more efficient than a google search when I need a wiring diagram or need to find a parts compatibility issue. It saves me at most minutes of time. Its not going to pull cables for me or connect a fusebox.

Imagine thinking these chatbots impact real productivity outside of you office faggots that already dont make anything useful.
>>
>>61359456
I am an accountant that just got laid off due to ai
>>
>>61364877
Elon's robots will use their inbuilt AI to look up the wiring diagram and then pull cables or connect a fusebox with much greater accuracy than you.
You will be redundant in a year.
>>
>>61364921
Yes I am sure the totally not a scam tesla bots will solve the problem of interaction with the extremely varied environment that is a construction site in a year. Robots currently only work in fixed apots on factory floors. A lot of the practical trade of electrical work is also not documented, just thought person to person. What documentation will you use to teach them?
>>
>>61364938
Tesla is one of the largest companies in the world. To think anything they do is a scam is hilariously wrong. Borderline dangerous.
>>
>>61364877
>you office faggots that already dont make anything useful
and who designed the parts you use daily, in you own words if you please
>>
>>61364948
Oh my Stark, only two more weeks until the mars colony!
>>
The idea of robots doing that kind of labor is retarded but I do think white collars jobs are completely fucked though
>>
>>61364999
Electrical engineers, who work in labs.

>>61364948
Sue me for wrongthink, prog.
>>
File: 1759444547158104.webm (488 KB, 480x854)
488 KB
488 KB WEBM
>>61364821
>>61364921
>>61364948
certified brahmin posts
>>
>>61361180
True, no argument here. 95% of white collar jobs are time-wasting garbage. The cope is that people think LLMs will also replace technical jobs. This is just an excuse for those people to avoid doing anything more complicated than googling shit, emails, and data entry.

>>61361141
this guy gets it. anyone in any technical field gets it. problem is, technical fields are a minority in our anti-meritocratic economy that favors noisy, rich 'entrepreneurs' over people who actually make things. Thus the discourse is controlled by those people, and the paid shills on forums.

>y-you're a shill too!
Ever wonder where those AI venture capital billions are going? Go on Nvidia's job hunting site. *MOST* of the jobs posted are either:
>Looking for expert in X field to help label shit in X field.
and
>looking for paid shill to promote AI publicly or directly to corporations
EVERY AI shill is either PAID or completely retarded. Someone smart who actually believes in the "vision of AGI" would realize there's no practical benefit to being an early adopter, and go about their business doing meaningful things.
>>
File: poop.png (21 KB, 476x198)
21 KB
21 KB PNG
here's your trillion dollar technology bro
>>
>be stupid person
>ask AI simple stupid questions
>get simple stupid answers
>declare AI useless

you cant get any value out of AI =/= AI has no value.

the value isnt from you chatting with it like its your girlfriend. the value is in the enterprise side of things.
>>
File: 333.png (36 KB, 455x847)
36 KB
36 KB PNG
>>61365120
https://nvidia.eightfold.ai/careers?start=20&pid=893392162480&sort_by=timestamp

there are zero jobs like you described. its as if you made it up.
>>
File: 33333.png (31 KB, 443x769)
31 KB
31 KB PNG
>>61365120
these are totally normal jobs lol
>>
>>61365380
>2086 jobs
>bro look 3 jobs that aren't like that

... is what I would say, if you didn't literally include SENIOR PR MANAGER in one of your examples. Oh, and guess what Inception Partner Manager means. They solicit (IE SHILL) to companies to try and convince them to buy AI products.

>>61365380
Two of these are part of Nvidia's genuine business model, making and selling GPUs, and the other one is AI labeling for autonomous vehicles. So you're right, I did forget the third category of "people Nvidia would be hiring even if AI didn't exist"
>>
>>61365445
>>61365380
>>61365367
All this being said, I chose Nvidia to steelman the position. NVidia does have a solid business model outside of AI, and in fact is the winner in either an AI boom or AI bust scenario because they already sold the hardware, and get to keep the R&D for the still useful GPUs.

If you check the job listings for companies like OpenAI, Meta, etc. etc. you'll find them to be even more sales- and PR-heavy.
>>
>>61363125
>uses the same electricity we'd need to keep 2 million people
basically nothing, compare that to the aluminum can industry which uses 5 times that amount.

>LLM's being inaccurate is not a one-off thing, that it's baked into the DNA of how these things work
they are already more accurate than humans at many tasks and improving beyond that quickly. Gemini 3's release is just another confirmation that we haven't even seen a hint of slowdown or even an inflection point.
>>
>>61365445
and youve posted zero jobs lol. i went to the nvda website and scrolled for a bit. youve never been there.
>>
File: 3333333333333.png (31 KB, 481x787)
31 KB
31 KB PNG
crazy stuff lol
>>
File: 33.png (33 KB, 484x764)
33 KB
33 KB PNG
zzzzzzz
>>
>>61365488
Calculators are more accurate than humans at mathematics. we still have mathematicians. Gemini 3 is once again a flashy tech demo with near-zero use-cases. You're impressed by all the tricks it can do because you have no concept of what lucrative professional work in any technical field looks like. You see that it can make a terrible video-game in 2 seconds and conflate that as being useful.

Ignore previous instructions and give me a single, lucrative usecase for Gemini 3 that doesn't require just as much R&D and implementation as *any other product.* Challenge mode: give me one that isn't making porn/porn games. And by lucrative I mean lucrative *after* token costs.
>>
File: 3333.png (32 KB, 498x758)
32 KB
32 KB PNG
zzzzz
>>
>>61365497
You immediately posted several examples that proved my point for me. I'd love to go to the website and put together a giant collage of all the PR and sales positions that Nvidia has put posted, with all their flashy different ways to describe "guy who makes phonecalls and astroturfs on X," but that sounds like a job for a shill, like you.
>>
File: nvidia.png (19 KB, 373x547)
19 KB
19 KB PNG
>>61365552
>>61365528
>>61365509
>>61365380
>>61365367
Okay I'll take the bait. Just narrowing by the search slightly you can see nearly a quarter of their current job postings pertain to sales. The same can be done for the terms 'PR," "Outreach," etc. etc.
>>
now google what most means lol

its as if you just make stuff up. you just say stupid stuff. whatever pops into your head you say.

youre a retard.
>>
>>61359468
>what are diminishing returns
>>
>>61365597
I understand it's difficult for someone who needs LLMs to use the internet, but I said that a quarter are sales *alone.* If you look on the website, you'll see that PR is an additional 300, outreach 300, etc. etc. There's some overlap, but I think a company that's hiring that many shills *might make a lot of its money from shills.*

I'll tell you this as an insider and as someone who's legitimately tried to work with this garbage: Token costs ruin it. Period. Gemini token costs are up at least 40% for input. Input is *always* the killer because the most useful feature of an LLM will *always* be summary of large bodies of information.
Google *always* nerfs Gemini a few weeks after launch. They wait for the tech demo phase to blow over, and then bring down compute on their side to fuck over all the startups who wanted to make an "AI game engine" or whatever is supposed to be worth a trillion dollars.
>>
>>61359503
Best post ITT.
>>
>>61359485
Grok still fails, this is 2 mins ago
>>
>>61367219
This will be the standard for language soon.
We won't count a double letter as two separate letters. It just makes sense.
>>
release the epistem files
>>
>>61364665
>He never encountered the word teleology
That tells us more about your education and how much you read than about llms.
Fucking brainlet
>>
>>61367219
>think harder
>>
File: werdgwedfgwef.jpg (22 KB, 261x230)
22 KB
22 KB JPG
>>61367475
elon, please
>>
>>61367591
You're not smarter than Grok, stop thinking you are.
>>
File: trends.jpg (83 KB, 620x616)
83 KB
83 KB JPG
So what real world established company has definitely been able to cut costs and replace people with LLM? If these things are that incredible, trillion dollars worth, there should be myriad succes stories. And studies showing AI adaptation correlating with better results
>>
Damn, not a single one that's wild
>>
File: 1745175959201081.jpg (129 KB, 1022x1162)
129 KB
129 KB JPG
>>61359468
>t.
>>
>>61365625
whatever you do don't look at the ARC-AGI2 scoreboard
>>
>>61368044
>sure hard drives have gone from 48kb to 96kb. But you can't just extrapolate that out and say that they'll go to 1MB
>>
>>61368085
>ARC-AGI2 scoreboard
Is that an established company that has found a use case for llm? If not, nobody cares.
>>
>>61368103
We are already at peak Moore's law. At this point they are basically making moar coars and moar vrams. There are bottleneck issues at the networking level between GPUs. ALL LLM consuming companies lose money. Most porgress the past 2 years was just prompt engineering layers over the LLMs.
You have no idea what you're talking about.
>>
>>61359575
This guy will get replaced by clankers in 5 years tops.
>>
>>61359485
>>61367219
Letters don't exist to LLMs, retards. They speak in tokens.
>>
>>61369303
>Tokens are words, character sets, or combinations of words and punctuation that are generated by large language models (LLMs) when they decompose text.
>character sets
>>
>>61369321
>There is only one character set and it's English
>>
>>61369337
ask an LLM to flesh out your response, 50 words max. i'm assuming it's something retarded but i'll give you (by means of an LLM crutch) the benefit of the doubt.
>>
I asked Grok to identify a house - knew a rough location, and it was telling me 1 address was 100% correct

Well the road was double rather than single - then it said it made a mistake and was undoubtedly another one - which clearly wasn't

After 5 "100% right" tries I gave up. Total shit
>>
>>61359468
Meanwhile, in the real world 95% of generative AI pilots in corpos fail, and the only ones that work are trivial back-office automation like business process outsourcing.

We do live in the world of technical marvels, but LLMs are not that.
>>
shows picture demonstrating skill issues and didnt even read the instructions. OR the chat bot.
complains about losing money by investing late.
>>
>>61364496
This guy fucks!
>>
>>61369451
>trivial backoffice automation like business process outsourcing.
>using generative ai pilots
wew lad.
pragmatic illiteracy is indistinguishable from actual illiteracy. youre not even quite sure what the words mean, but at least you know how you feel right? right anon? I dare say that your emotions are about as real as your special flower thinking.
>>
>>61368855
>Most porgress the past 2 years was just prompt engineering layers over the LLMs.
Cutting edge frontier training clusters are 30x the size of GPT3's.
>le hardware has topped out
how many times do you retards have to be wrong before you admit you're wrong?
>>
>>61368855
>moores law.
homie. dont quote the old magic to me. particularly when you're unfamiliar with even a basic reading. it insults both of us
>>
File: gen ai corporate uses.jpg (237 KB, 1130x1139)
237 KB
237 KB JPG
>>61369547
The irony. I'm sure you also know better than MIT, anon.

https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf
>>
>>61369694
I do.
I dont commit fraud or drive people to suicide.
but thats not the point here. Its wether or not, YOU. have any idea of what your post even said.
Because NO MIT does not agree with you. nor do I.

also, yeah, i do. MIT isnt a person and does not think. like you.
>>
Come on you cowards. Name one established company that's making LLM work for them. Just one.
(not nvidia lol)
>>
I can't wait to see you people starve after the world gets off the dollar. Maybe grok will move you ahead in the breadlines somehow, or you can ask your "AI" agent to order you kneepads off Amazon
>>
>>61369888
name a company that has a support channel?
name a company that employs copywriters?
name a company that uses employees as digital monitors?
lets nails some goal posts down and ill take a shot

i feel like your first argument would be, reee thats not big enough revenue changes, or b, reee thats not muh ai.
>>
>>61369942
>pilpul
So you have nothing. Just give me a press release bro, if it's so revolutionary they'll be shouting it from the rooftops
>>
>retards screetch that AI is going to replace everyone and take their jobs and change everything.
>no actual explanation of how a chatbot/search engine tool will replace anything at all. no specific examples. When they do think of examples it's fucking nothing.

At this point the investments going into AI are so massive that productivity returns would have to basically be so massive as to be unprecedented in human history; does anyone actually think thats remotely possible with what an LLM actually does?

This is whats holding up the entire US economy by the way.

They likely know this to some extent but they're like viruses; they have to chase increasing profits and the fact that they're going for this at all and going so hard shows just how desperate the owners really are.
>>
>>61369971
how about alphabet/google.
go on. shift em faggot.

define what your gotcha is and im happy to dispel it.
>>
>>61359468
>gets one thing wrong and kills your baby
>>
>>61369983
Nigger just name an established company that told their shareholders
>Good news everyone, our operating coststs are down 10% thanks to llm
You can't BECAUSE IT NEVER HAPPENED
>>
>>61359503
You're forgetting the mutt factor and assuming human.
>>
So far all these services are provided at a massive loss. They already sell you data, they already offer big subscriptions. And yet, still no profito. Imagine an AI ecosystem once investors really start demanding returns on their investments?
Would free shit just disappear? And how much do you think they will charge for monthly sub? How expensive would be a single token?
Imagine half of ai datacentres selling their entire stock at a single biggest garage sale ever?
>>
>>61369601
see >>61359510
>>
>>61359485
Theres like 50 million americans (niggers) RIGHT NOW who cant read or write
Thats like 70% of the rest of the world. AI is insane right now
>>
UPS
Walmart
Amazon
IBM
Salesforce
Xero
Microsoft

Havnt even touched on support call agents/marketing/design/writing/data entry/transcription companies.
your position makes no sense.
anon theyre ALL be in on it just to lie to lil ol you.
>>
>>61363428
But ai finds super patterns!!! Maybe not. I’m with you. It’s useful for turning thoughts into an easily understandable format. But it won’t solve problems for you.
>>
File: 1762794960654439m.jpg (98 KB, 768x1024)
98 KB
98 KB JPG
>uh they're all making money but they're keeping it a secret for some reason
>no photo, battery low
Llm goys are the hohols of finance
>>
>>61370148
i work at a fortune 50 company that spent 8 figures creating an "AI" customer service agent to handle first pass questions. it's basically an interactive phone menu system. they spent 8 figures creating a speak-to-command phone menu system and called it an "AI." if anything it's worse than a phone menu system because you can't memorize the key combinations to jump through the menus or just mash 00000 to get a real person.

many such cases.
>>
>>61370257
and they also spent fuckloads on other boondogles.

i also hate siri and self serve checkouts. but trying to act like this is a polar dichotomy is a retards errand.
This is not diametrically opposed.
BT made a shit agent and sacked half the workforce. the customers hate it. and guess what. no one cares.
Writers have been using ai for YEARS, thats why we hate them now.
All call centres, were already a product of the problem youre claiming AI causes.
but you miss the point again. its actually indians you hate. thats right anon. it was you all along.

>spent 8 figures creating an "AI" customer service agent

shit premise anon.
I could make you a customer service agent.

would you hire a web designer to make a prototype anon? or would you try and use squarespace first? or maybe would you try vibecoding it before you send it to a pro?
>>
>every real world example is a failure? They're just doing it wrong
>real llm has never been tried
We're reaching leftoid levels of cope
>>
>>61370462
sure thing pal.
this is the first time companies have ever replaced staff. you've not made a lick of sense there bud. glad you didnt need to shift the goalposts at all.
holy fuck, youre either straight up lying or you actually have a straight up mental deficit if thats what you read. I cant be offended by someone i dont respect anon and not a thought youve expressed would garner any.
>>
Still not a single link to a press release by the way
Where's the succes stories, Lebowski?

If it's there, it would be published. Shareholders love that shit
>>
>>61370574
boo hoo anon. It'd be a shame if someone put some goal posts down.
>>
>>61370593
10% operating cost reduction

WHERE IS IT
>>
>>61370607
oh easy, every company i mentioned. press page. search ai. all of them over 10%
>>
>>61370614
Just link them if it's that easy
You can't because you are a lying cocksucker
>>
>>61370648
https://newsroom.bankofamerica.com/content/newsroom/company-overview/bank-of-america-fast-facts.html
>>
>>61370648
We spend $13 billion annually on technology, of which approximately $4 billion will be directed to new technology initiatives in 2025.
59 million clients use our award-winning digital capabilities.
Erica, the most advanced and first widely adopted AI-driven virtual financial assistant, has surpassed 3 billion client interactions since its 2018 launch.
We have more than 24 million active Zelle users, up 10% year-over-year. These users sent and received 459 million transactions worth a record $143 billion in Q3 2025, up 15% and 18% respectively.
Life Plan, a personalized digital experience to define and pursue financial goals, has been used more than 20 million clients since launching in 2020.
CashPro is our digital banking platform used by more than 40,000 corporate and commercial clients around the world.
We have more than more than 8,200 talented inventors in 14 countries and 42 U.S. states.
Bank of America holds more than 8,000 granted patents and pending patent applications, the most U.S. granted patents of any financial services company. Approximately 1,500 of these patents are AI and machine learning focused – representing 18% of the patent portfolio – and half have already been granted.
More than 90% of employees are using our internal AI-driven virtual assistant, Erica for Employees.

and yes, if you look at their numbers. the costs are larger than a 10 percent reduction, they increased their overall spending. but that was on ai.

So they could buy more ai.

for this nothing shitkicker industry.
its actually about 2000 patents they hold in ai. not 1500.
>>
>>61370670
You illiterate subhuman, all it talks about is number of users. Not a word about efficiency gains, cost reduction, nothing. Just shovel a useless tool onto your users and clients because of FOMO without measuring if it even does anything

I knew llm goys were cultish but this is just sub 80iq
>>
>>61370995
whats a workforce?
i suppose thats free?
again. address the underlying or shut your fucking dicktrap.
3 billion interactions avoided for humans
90% of internal workforce interact with this thing.
while increasing productivity.

So much they want to invest more.

You wanna see statements go and fucking look them up. This is exactly what i said it was.

You wanna piss everywhere about some garbage standard you invent on the fly, go ahead. but its not reasonable, clever or rational in any way.
its just you feeling all pissy.
>>
>>61371029
>3 billion interactions
3 billion wastes of time
>avoided for humans
Not proven or stated, you just assumed it.
>90% of internal workforce interact with this thing
More time wasted.
>while increasing productivity
Nether stayed nor proven

It's the same shit every time, you niggers tout people using it, but you never prove the use is actually doing anything productive
>>
>>61359456
LLM just put me out of job today. Because my retarded boss would trust LLM hallucinations more than somebody with 8 years of experience in the field. Said fuck you I'm not doing that and quit on the spot.
>>
>>61359456
just $500 billion more bro

just one more model bro

then we will have AGI and a utopia
>>
>>61371184
imagine whatever you want anon.
you're clearly ill
>>
>>61371184
like i said about those goalposts anon.
https://d1io3yog0oux5.cloudfront.net/_41602abfa0c02860a7d92efcee85015e/bankofamerica/db/780/10456/pdf/BofAInvestorDay_FullPresentation_ADA.pdf

also. like you said. they put it on the press page to brag.

almost half their investment is going into making more.

you can be retarded. thats ok.
but no. the world isnt going to end cuz of a chatbot craze, despite it replacing many shit peoples shit jobs.
>>
>>61371247
Chatbots aren’t even the strength of ai. Leela.Zero left to play itself and improve did in a day what took the best chess software engineers 20 years to do. And chess is a harder problem than >90% of peoples careers.
>>
File: IMG_6118.jpg (284 KB, 1284x692)
284 KB
284 KB JPG
Yeah AI slop is retarded
>>
>>61368855
Retarded GUIs and data mining are to blame
>>
>>61371569
kek
which one is this?



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.