[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/biz/ - Business & Finance


View post   

File: 201 KB, 1200x635, C5539C21-1A2D-41C9-A6A9-26AD06F1E7ED.jpg [View same] [iqdb] [saucenao] [google]
58652722 No.58652722 [Reply] [Original]

It’s unbelievable how much tech illiterate retards are overvaluing NVDA

>> No.58652746

>>58652722
explain why
processing cards are elementary to AI which will subsume everything in the entire economy, and competition is years behind

>> No.58652761

>>58652722
I bet you have a GPU in your PC right now, OP

>> No.58652776

>>58652722
They are building the tech that will replace programmers, accountants, front desk people, pretty much any white collar work. And their recent releases are ai factories for training virtual ai so they know how to interact once they are in the real world. Shits just getting started.

I don’t even own any.

>> No.58652777

>>58652761
I had a card in my pc 15 years ago too!

>> No.58652818

>>58652746
AI is only kinda useful for some stuff and you can run most models just fine on Radeon or Apple M2/M3. Over the last year there's been an explosion in new models which don't do significantly better than GPT 3.5 but require massively more compute. It's over.

>> No.58652824

>>58652776
No they aren't. It's all marketing hype and BS. They'll be begging workers to come back when "AI" fucks up everything because it's not nearly as great as it's being promoted to be. The first big company bankrupted by hallucinating, malfunctioning AI will be the tipping point.

>> No.58652832

>>58652722
Let's pretend you understand supply and demand. Right now there is a Tesla hit piece going around saying Elon diverted his chips from Tesla to xai, possibly to the detriment of Tesla. Whether this story is true or not is irrelevant because nearly everybody believes it. What does imply about the supply? More importantly, what does that imply about perceived demand?

>> No.58652842

>>58652746

both google and Amazon and everyone else makes their own tensor chips

>> No.58652975

>>58652824
This is not going to happen. Capitalism is following the same path that it always has, first from introducing computers in the 80's, then to Jeets being imported en masse in the 2000's and now to AI. It's not about finding a machine that executes flawlessly (i.e. The Terminator), it's about finding the lowest cost of labor that is "good enough".

AI will be used to make low cost labor better and cheaper or replace them entirely. Some people will continue to get better jobs while the majority of people will receive increasingly shittier ones as time goes on. AI will keep improving and drive efficiency higher.

We're getting to the other side of the slope where humans increasingly will not be needed in white collar jobs anymore. This probably means they we'll just go back to killing each other for pointless reasons at some point, but hopefully that's far away.

>> No.58653076

>>58652975
At some point this will happen but the current trend of throwing more and more GPUs at it and training larger models isn't it. The current technology is simply insufficient to replace humans for the vast majority of real-world applications. It's severely limited by not being able to react to unexpected situations which aren't present in the training data, and when it encounters those it hallucinates and makes a mess of things. I can't think of any job where you never have to respond to novel situations, which is something that at least jeets are capable of.

The current state of things is the same old bullshit Silicon Valley has been doing for ages. Promising some new approach will revolutionize the world, a bunch of tech startups funded by VCs popping up, shovel sellers having their stocks pumped, and eventually those startups going under because the tech isn't that great and they haven't turned a profit in years. The fact is the economy is in the toilet and the only thing propping it up is a few tech companies selling snake oil as a panacea to investors. Once they realize it's not all it's hyped up to be the rug will be pulled and the entire market will dump. That could happen any day now.

The next iteration of AI that's actually useful will eventually pop up but there's no way of knowing how it will function or if the chips Nvidia is making now will even be useful for it. Hell it might run on quantum computers for all anyone knows, which means all these chips will be trash.

>> No.58653080

>>58652975

the current incarnation of AI is a slop generator. It has not and will not produce anything worthwhile.

It's the equivalent of paying a pajeet 5 dollars to make a logo on fiverr.

the only reason why it's ballooning is because the (((californian turboreddit eurasian technoelite))) thinks that it's the path forward to the advent of the new God, a machine that thinks like a human.

AI will cement itself as a domain specific tool in the coming years, and that's where it will shine the most. Fixing blemishes in photoshop faster? Sure. Generating snippets of code? Of course. Generating shitty logos? Yes. But nothing else.

>> No.58653088

None of that matters they are using NVDA to prevent gme from going to infinite le preddit moass
Meaning NVDA will go to 600k eoy

>> No.58653090

>>58653088

N

>> No.58653102

>>58653090
V

>> No.58653107

D

>> No.58653111

A

>> No.58653254

>>58652746
a lot of research now is moving to running models on literal toasters, see the whole apple presentation and also google is trying to do something similar with android. a lot of research aims at efficiency, in fact. so just for inference, demand should go down except in industrial and b2b applications.
finetuning CAN also be done on other chips, though admittedly pretty inefficiently. but theres only so much demand from that aside from a few huge players, researchers, and hobbyists. almost all AI applications outside that use existing models and dont even bother to finetune, but rather use techniques like RAG (essentially querying context from a database and feeding it to the LLM).

i frankly dont see nvidia being worth more than fucking apple or microsoft, in the latter case especially since theyre still effectively the forerunners on the software side for AI, as well as having a fuckload of other business units that are not AI when nvidia essentially just has whatever needs to render shit, some research, and thats it pretty much

>> No.58653258

>>58652722
The most annoying part is it'll probably go to $200 ($2000) before it comes back down

>> No.58653263

>>58652776
AI will make them more efficient, not 100% replace them. its a very good autocomplete, but everyone who unironically uses any LLM without factchecking tends to be retarded
had lots of coworkers who had it output code, didnt know shit about coding, and then asked me for help to fix their stuff

>> No.58653749

>>58652746
this but for RLC

>> No.58653780

>>58653263
Yes, as of right now any programmer worth their salt should have had a 10-100x efficiency boost from opus or gpt4 if they use it correctly. But it will be able to handle much much bigger context windows with less hallucination eventually. Even 5 years from now what we have will be completely mind blowing if you make anything on a computer. It really is the 4th Industrial Revolution in a lot of ways. I hope it doesn’t replace it, but I think it will within 5-10 years.

>> No.58653794

>>58653076
Thank god there are still some people who are level headed, or at least understand the tech enough to see through the hurricane of bullshit people are spouting. "All jobs will be replaced and you won't have to work anymore" is the perfect message for both the bankers and the generation of hopeless coomers they created. Bad news, fuckers. This shit doesn't fucking work and it never will with this kind of model. You will still have to work, but from now on without a functional internet and an infinite sewer pipe of AI generated garbage permeating all media you consooom. Buckle up.

>> No.58653836

the real power of ai doesn't come from excess compute, but from langchains the ai knows how to use

>> No.58655281
File: 67 KB, 480x480, 159096133563.jpg [View same] [iqdb] [saucenao] [google]
58655281

>>58652722
and undervaluing other tokens, stocks and developments in general, the ginormous amount of potential logic-based ai has and its being overshadowed by "muh 300B super llm" nvidia faggots will cry once the neural network craze comes to a sudden stop only to be replaced by the energy-efficient and vastly superior models like what tau net is developing, i seriously cant wait for formal specification be be cool again i have some frens in college i need to seriously clown on for besmirching my beloved logic models

>> No.58655316

>>58652746
i wonder how much of this will hold true once we start reaching towards quantum computing and hyperspace computing
we've kinda reached a halt in how big our processing cards can be in terms of cores and you can see this in all the way the tech market is failing to deliver a proper new generation of hardware, i could genuinely see this being the point were we stop going higher in the processing power and focus more on far refined, more complex software that rely less on computing and more in efficiency

>> No.58655320

>>58652776
this is so cruel to me, the fucking system forces us to create the programs that will replace us and we obey because we have to it, programmingbros... its over

>> No.58655329

>>58652975
>>58655281
read this back to back anon, realize that what he's saying is right, i know the tau net guys, they're basically on a campaign to destroy programmers and coding in general, i dont care how "revolutionary" their tech is, you're just leaving hundreds of thousands of professionals without jobs just so what? "now everyone can be a programmer" yeah sure... whatever

>> No.58655395

>>58652776

weeks remaining: two

>> No.58655412

>>58652746
AI is massively overhyped right now. language models are not the end of technology. I'm a massive believer in nvidia in general because they are the #1 producer of compute and compute is going to take over the world even more than it has so far, but I think buying right now is risky because the whole thing can dump at any moment now.

>> No.58656001
File: 451 KB, 1011x925, 1641324659613.jpg [View same] [iqdb] [saucenao] [google]
58656001

>>58655281
ur being a silly goose if you think nvidia is only shilling for llm's fuck off

>> No.58656007

>>58652722
I thought it and you said it.

>> No.58656019

>>58655412
IT IS RISKY THO

>> No.58656020

>>58655281
OpenAI and Nvidia are the initial harbingers of AI right now. They'll be deprecated in time for better things. It's just how shit works, people are clearly putting a lot of chips on the table for this technology with no clear intention in mind other than being "the first". It's the dot com craze all over again, history does repeat itself lol

>> No.58656023

>>58655320
do you really think they are going to replace us?
thats bullshit

>> No.58656024
File: 66 KB, 639x571, 1654111845986.png [View same] [iqdb] [saucenao] [google]
58656024

>>58652776
Eight billion people and counting exist on this planet btw. I don't know what the fuck we are doing. There's only so many factory jobs and blue collar shit available, not everyone will be able to get one of those when the time comes.

>> No.58656031
File: 22 KB, 720x662, 8789.jpg [View same] [iqdb] [saucenao] [google]
58656031

>>58652975
So we can already predict it?
just saying..so i can stop doing the stupid thing I'm doing and die at once.

>> No.58656032

>>58653076
>>58656020
>it's the dot com craze all over again
holy fuck you are right

>> No.58656040
File: 144 KB, 1013x1008, 1654012507849.png [View same] [iqdb] [saucenao] [google]
58656040

>>58653080
>(((californian turboreddit eurasian technoelite)))
Lmoa

>> No.58656055

>>58656040

I bow down to the judeohapa master race of the future

>> No.58656080

>>58652722
i have been generating ai porn for nearly a year and i didnt think to buy stock in nvidia. i deserve poverty

>> No.58656089

>>58652761
I use CPU, I would never use GPU

>> No.58656191

>>58652722
Check Theranos, Wework or Nikola trucks.
Nvidia ain't shit compare to that

>> No.58658063

>>58652818
>AI is only kinda useful for some stuff and you can run most models just fine on Radeon or Apple M2/M3
I can't even run the last version of tensorflow or pytorch on an AMD videocard.

>> No.58658664

>>58652722
True.
https://4techmix.com/2024/06/20/irrational-exuberance-in-ai-with-nvidia-at-record-high/