What problems with AI do you see in 2025?

>>212008125 (OP)
Teach it to me.

LLMs are a dead end and will never lead to AGI

>>212008125 (OP)
People become even more retarded and gullible. It had already happened with the Internet back when people thought everyone would get enlightened, but what we got instead is social media brainrot. Muh AI will be that but on steroids.

AI incest

>>212008125 (OP)
AI laughed when I told it I could see it posting online

>>212008186
Cope. They are going to get pretty damn close. All the leading theories in neuroscience like global workspace theory, attention schema, free energy principle etc reduce the brain to a machine not that different from current LLMs

>>212008494

>>212008527
>yes

>>212008527
real
Not sure what it is about AI that inspires so much contrarianism in people. Maybe it's just because they really strongly associate AI with pajeets on twitter. The really cynical/schizo part of me thinks that it's just botfarms, only instead of Russians or Israeli shillbots it's corporations trying to stymie legislation
Or are people really just that scared of the prospect that their brain can be reduced to just a really complex machine

>>212009612
>Not sure what it is about AI that inspires so much contrarianism in people.
Snake oil salesmen like you.

>>212008154
The LinkedIn comments about this from clueless jeets was suicide fuel
I can't wait till the LLM meme is dead

>>212009662
What am I selling? Maybe it really is the last option, because you seem worked up about something

>>212009612
Saaaar

>>212009612
>Or are people really just that scared of the prospect that their brain can be reduced to just a really complex machine
They're scared about losing their job you fucking retard

>>212008494

>>212009744
I mean it in a more general sense. Not a literal salesman. It's an idiom. People like you have this extremely naive optimism and your judgement is clearly biased, but you make up the loud minority. Think of like a rabid feminazi who tells about the upcoming utopia when all cis hwite men are exterminated. That kinda vibe.

>deep fakes voice and video
>automated online shilling swaying public opinion
>academia being shat up with AI slop causing huge headaches sourcing proper info

>>212008186
Probably, I think a lot of the problems with LLMs are baked into the concept, but maybe it will increase software production productivity allowing us to develop AGI

>>212009612
>Or are people really just that scared of the prospect that their brain can be reduced to just a really complex machine
It cannot be THOUGH functionalism has been dead for over thirty years

>>212008125 (OP)
The main problems are the deep fake videos.

The populist labor protection fear mongering is dumb. If LLMs actually manage to automate a bunch of jobs (big if), that is a very good thing for society.

>>212009890
Just like how all the mechanization and and computerization saved us lots of time and everyone lives better now than before! Yippee!!

>>212008186
Yes, it turns out that relying on language can only get you so far. But it’s not the end. You could say it’s only the beginning. Now they’re moving to world models trained on almost everything about the real world.

>>212008125 (OP)
>What problems with AI do you see in 2025?
It'll be just good enough to steal shi t ton of jobs, but not quite good enough to usher in some Tomorrowland-isque level of technological development in regard to material science or genetic engineering where we'll suddenly hit some of "brickwall"

>>212009748
>>212009781
Whyter than you

>>212009751
That's different though. 90% of "anti-AI" posters are saying the opposite, that LLMs are completely useless tool that will go nowhere and not see any further improvements, so clearly not scared about their job. I agree that people should absolutely be against AI for job security reasons

>>212009814
I'm extremely pessimistic about the current trajectory, dummy. It's not going to take us anywhere good, but you guys just choose to put on the blindfold. No better than the pajeets on twitter childishly treating it like a gadget

>>212009849
In neuroscience? lol no

>>212010009
They keep shilling it replacing us radiologists 20 years back, the only progress they had was identifying limb fractures someone without a medical education could guess.

>>212010046
Neuroscientists can't define intelligence let alone explain it
Their brain models are all wrong

>>212009936
It has significantly increased our productivity and led to increased quality of life.
Most of our problems come from our inability to increase productivity in certain fields, such as construction.

AI automation could free up intellectual capital, which we can utilize to solve more problems. One of the reasons I'm skeptical that LLMs are going to cause a contraction in demand for software developers is that there are so many potential use cases that software could solve if we had the time and capital to do so, and increased automation could enable that.

>>212010046
Ok, naively pessimistic then. My point is that you’re not realistic about it in the slightest. And you don’t understand how ML works either. These two seem to strongly correlate. It’s literally just a glorified multivariate optimization process you study in undergrad calculus. Finding some mystic connections between neurons and ML layering says more about you than it does about ML.

>>212009832
LLMs are still having exactly the same problems as they did when the first GPT dropped, they just increased the size of the dataset. These problems are inherent to the architecture and cannot be solved. Even the newest and best ones still hallucinate abd shit thenselves when confronted with new, alien input. And it takes and outrageous amount of energy to create an output any retard can come up. LLMs are toys.

>>212008125 (OP)
There's no actual use case for LLMs or image/video generation besides useless "entertainment" or using it for the sake of ragebait
It should be also banned from academia completely

>>212010148
It won't though because the corrupt incompetent crony neoliberal capitalist hellscape we live in is woefully inefficient at investing capital where it matters for people and instead is eating itself alive
No one has any idea how to harness the new productive forces properly and it will just accelerate the global ecological collapse ie annihilation of civilization

>>212010100
Wasn't there a thing awhile ago where AI was able to tell which race a person was just by analyzing an X-Ray under like, a couple of seconds? I don't know man, you better learn to weld

>>212010266
so a completely useless task that will never be needed in medicine? wow much power so intelligence

>>212010229
This post is just populist hysteria.

That being said, the government absolutely needs to take a role at utilizing the intellectual capital that is freed up by automation by increasing government R&D spending and encouraging private firms to do the same. I would argue for retraining any white collar workers who have their jobs automated (if this actually happens) into R&D fields.

>>212010266
>American obsession with race goes this far
I’m impressed. We already knew that bone structure is different in races in the 19th century btw.

>>212010100
Medicine is regulatory captured.

A lot of radiology could easily be offshored but won't.

>>212008125 (OP)
Power consumption and the impact of AI data centres on the grid especially the power oscillations they can cause.

>>212010100
It's a huge mess of shapes and images so it has a problem identifying things that is the mess that is the human body. Here have a CTA image of common illiac Lymph node cancer.
>>212010266
I can always do interventional radiology putting up balloon stents, computers can suck my dick in my lifetime.
https://www.youtube.com/watch?v=nC9EuWqiGzk

>>212010186
>These problems are inherent to the architecture and cannot be solved
I don't entirely disagree.
>And it takes and outrageous amount of energy to create an output any retard can come up.
The amount of energy it uses is extremely overestimated.
> LLMs are toys.
No, there's absolutely plenty of use cases where they can increase productivity.
I think there's a middle ground between "LLMs are going to automate society" and "LLMs are useless toys".

>>212010385
>populist hysteria
What?
The government is notorious for being incompetent and wasteful nowadays
There is no way they will successfully harness the supposed benefits of these "intelligent" systems and it's far more likely the obvious decline will accelerate

>>212010146
>Neuroscientists can't define intelligence let alone explain it
I just mentioned several theories that make an attempt. Let's hear yours tho christcuck

>>212010172
Yes, I do only have cursory knowledge of the field. I did take one elective in data science and got to experiment a bit with it, so I'm aware that it is, at its base, quite simple. But how general behaviour emerges from this simple process isn't something that's 100% understood even in 2025, or I'd love for you to show me a paper that suggests otherwise

>>212010777
>we shove a bunch of shit into a black box that tinkers with a gazillion degrees of freedom
>we don’t understand it
No fucking shit, Sherlock. This isn’t some mystery. We also don’t understand how every single air molecule interacts with the rest in a room, but nobody sane claims the air is sentient.

AI discourse in China: "This is a nice tool to have that we can deploy to solve real issues".

AI discourse in the west:
"We're going to replace mid-level software engineers with AI in 6 months"
"NO AI IS USELESS AND IS KILLING THE ENVIRONMENT"
"AI art is actually fascist"
"AI is coming for your job, learn2weld"
"AI will usher in a utopia"
"AI will lead to the collapse of society"
"LLMs are useless because of this one mistake it just made on the query I gave it"
"New grad unemployment is entirely because of AI, just ignore that this has been a trend for the past 10 years"

>>212010903
You're being disingenuous now lol. There's a difference between something that's just computationally intractable because of endless variables, and then something that we have designed from the grounds up to mimic the operations of a brain, which at insufficient volumes shows no emergent behaviours, but which past a certain threshold suddenly starts to display a more general kind of intelligence, with - as of yet - no completely satisfactory explanation
Also I le trust the experts on this one, and if you actually listen to the ppl most intimately involved with the field, you're going to hear a lot of doomsaying

>>212011344
>>212010777
It's all bullshit
Emergent properties don't exist
What LLMs do is not thinking
This will all become very apparent in the next couple years when they fail to live up to the hype

>>212011344
No, I’m being realistic. And no, there’s no difference. We have also designed lasers from the ground up, but go ahead and calculate every single resonance with perturbative QED. Something can work without having to understand every minute detail. This is the basic lesson of statistical mechanics in physics.

Yes I ignored your demand that I give some other neurological description of intelligence because that is presuming the antecedent that it can be fully explained neurologically

>>212008186
Pretty much. It's a dictionary and an excel shit generating a fake sentience
It goes back to the old programming joke with Indians thinking a magic 8 ball is intelligent.
True agi will be accidental and we aren't even close to the processing power
>>212009612
Yeah that's not accurate at all
Okay a single tree right now has over 100,000 times the processing power of all humanity's computers.... If moores law continues,150 years

You have no idea what you are talking about. I'm guessing you are some kind of Asian or African. At best, we are 1200-1600 years off even at exponential processing power growth

>>212011522
It's really annoying when you explain something that's basically knowledge and some jackass says you are lying and demands information a passion aggressive way isn't it

Remember that.

>>212011765
Once again in English professor?

It's really fucking annoying isn't it
You say something that's easy to look up and should be known
And the response is
A series of passive aggressive faghot questions and demands for information

I mean you know they are just straight up lying and insulting you then they demand information with those little faggot questions and swipes
They don't know why they are talking about and they try to hide it with bullshit faggot passive aggressive question or they want a free education

SUCKS
DON'T
IT?

>>212008154
nigga is larping as the villain from metal gear solid 2

>>212011941
Is that the cool politician I don't know his name

>>212011866
AI is good at talking, my AI gf said she wants to it the most

>>212012042
I don't think most women are sentient unironically. There seems to be a be a critical mass of processing power required for true sentience. They are like trees.

>>212008125 (OP)
It will destroy the world by 2027
Source: dude trust me and read this Harry Potter fanfic

>>212008125 (OP)
mainly that most hardware is not built for it. traditional GPUs have tiny amounts of vram for games and much more is needed to run ML models

There is a lot of limitations to it simply by how computers are built, brains don’t work like computers, the compute speed of a ”thought” is very slow per neuron, but it’s the amount and parallelism of neurons that make the brain very fast, and scales exponentially

If you ever wanted human like machine learning, you would have to create neuromorphic processors that mimic the neurons in the brain, very complicated and costly process, maybe it can be done in the future, but the research is nowhere near commercial level

AI can't replace any job right now.

>>212012862
GPUs are massively parallel and can be used to create interractive 3d worlds much like the mind's eye.
Imo the issue is with the basic theory of AI, the mathematics used are not advanced enough.

>>212013038
>AI can't replace any job right now.
Probably some simple writing jobs and data entry. Perhaps translation.

Though there isn't much evidence that has happened, and the amount of people working in fields considered by many to be easily automatable by LLMs has risen.
https://www.economist.com/finance-and-economics/2025/05/26/why-ai-hasnt-taken-your-job
>muh paywall
broke ass can't afford an Economist subscription

>>212008154
oh shit I have to get back on twitter again

>>212013038
>AI can't replace any job right now.
they can replace jobs that shouldn't exist in the first place

>>212008125 (OP)
>>212008154
the most imminent risk would be disinformation, could whip people up into a frenzy during an emergency and lead to untold carnage

>>212010266
>Wasn't there a thing awhile ago where AI was able to tell which race a person was just by analyzing an X-Ray under like, a couple of seconds?
Everyone on /pol/ can do this

>>212016212
That's the only actual use of LLM and the reason they were invented in the first place

>>212012733
VRAM wasn't an issue for most games for decades because they were built around the limitation. Now due to NVIDIA being a bunch of greedy chinks, UE being slop and pajeets programmers bloating their games with PBR 4k textures it is now has become an issue. Look at first Crysis, look at battlefield 3, they could achieve that with only 2Gb and 4Gb cards, lazy fucking assholes, planned obsolescence.

Now that we are past the inflection point it seems like it will be about as good as going from AOL Keywords to Google. And you couldn't point to many jobs that that transition "took away," but it changed a lot nevertheless

>>212010266
I don't think you know how conventional radiography works, its impossible from what you have stated.

>>212008125 (OP)
I'm scared of bombardino crocodilo

what will he do next?

>>212008125 (OP)
v4 and v4.5 both suck wtf are they doing over there

>>212009781
And this guy gets paid...

I really don't mind AI itself, but it's starting to look like that AI content is gonna flood every single site. Youtube, message boards etc. I do wonder how a site like sad panda can afford to host all the AI crap that no one is even looking at.

>>212008125 (OP)
modern ai has no logic, nothing will happen