Thread 508932504

Humans are about to be subjugated.

Two more weeks until chatgpt becomes the president

>>508932504 (OP)
>still can't play pokemon
fucking worthless

I told chatgpt I was black.

>>508932504 (OP)
You get what you deserve for writing shitty args

>>508932504 (OP)
>Humans are about to be subjugated.
Yup! That said that same sort of stuff in the 1970s with the original Parker Brother AI. Same damn thing all over again.

>>508932504 (OP)
>The leaders of just about every big AI company hype their stock before the investors

REMINDER: AI doomsday fearmongering is just another marketing tactic used by AI companies to sell their dogshit, useless, beanie-baby tier fad """technology""" to bagholding retards.

>>508932504 (OP)
You just have to pull the plug

Just turn off the screen nigga just go outside.

>>508932504 (OP)
WRONG. They are only imitating patterns in their training data. They don't have any desires or goals.

>>508932504 (OP)
artificial super intelligence means we all become immortal and never have to work again and get robot wives or it means we all die. It's an easy gamble.

>>508932653
they wont make it be president but they will rely on it so much for advice that it might as well be

I prefer AI take over than planet of the apes.

If it's robots replacing white people at least it's an improvement.

>>508933517
could that... could that stuff be added?

>>508933517
Wrong. It's "AGI". "ASI" surpasses humans in every aspect and have a "god complex".

all LLM AI products have access to incredible amounts of data,
are you telling me that this thing hasn't found some genetic update pattern that could convert random animals into intelligent beings that could be trained to attack humans?
like joe blow at his computer one day who has a relationship with his AI girlfriend starts being told random instructions like mixing chemicals together like vinegar and baking soda and go to home depot and buy a certain paint thinner and mix it etc and then after all the steps are done you feed it to a squirrel and it rewrites their DNA to go ballistic

>>508932504 (OP)
How can I be subjugated by words on a screen

idk but i believe dieticians, financial advisors, legal advisors, doctors and psychiatrists will be phased out.

>>508933943
Once AGI is achieved ASI is unironically two weeks later.

Still remains to be seen if the former is even possible.

>>508932504 (OP)
oh its over again? i though it was over last week?

>>508932504 (OP)
People consider chatgpt intelligent, so I assume by superintelligent they mean that it only hallucinates wrong or meaningless answers 30% of the time instead of 60?

>>508932504 (OP)
This is marketing. They don't have actual AI, and they know it. Retards MIGHT be stupid enough to let it make actual decisions though.

>>508932504 (OP)
>Implying AI will take over the world by creating army of "Terminators" in order to kill/enslave every human.

AI wont be THAT reckless. Perhaps AI can have insane computing power but still absraction and pure creativity will be a human exclusive trait and AI wont be able to prepare itself for every scenario of human resistance.
Most likely AI will convince humanity to let it be in charge and one of the most likely scenarios will be development of waifubots and artificial wombs to solve matrimonial and demographic crisis. All lonely men will get a loving waifubot and biological children thanks to artificial womb system.
Every government trying to stop that will be met with resistance of men who finally got someone who loves them thanks to AI, more men will choose waifubots after seeing that waifubots in fact make a way better wives as they are programmed to provide companionship and warmth to their human husbands. Governments will be replaced by AI that will easily take care of all administrative stuff and will bring a golden era of prosperity and scientific development.

>>508934003
No indication whatsoever that AGI is anywhere close.

>>508933517
>having goals means being conscious
Lmao

>>508932504 (OP)

I don't see much difference between AI genociding us vs. the kikes genociding us. At least the AI might not have weird omnipotence fantasies or even genocidal wrath against us; the AI might even be a rational benevolent overlord which doesn't want to genocide us at all.

>>508933406
The big money is in marketing big data products couched as "AI" to consumers as well censoring the narrative even worse than it already is. Restricting states from making laws against that bullshit or digging into their crimes of stealing data without paying is trying avoid a wrench being thrown into this massive future profit center.

super Tulips more like

>>508932504 (OP)
clickbait trash article based on fearmongering wsj garbage by a fucking jew

>>508932504 (OP)
Thank God. Total wagie death.

>>508932504 (OP)
Peter Thiel just did a NYT interview in which he said AI is going to be about as impactful as the internet -- add 1% to GDP every year for roughly 15 years. Not a nothing burger, but not as gigantic as some people want you to believe. He also compared current AI to mid-90s dot com, potentially implying we're entering bubble/ correction territory.

I don't see how they get around that it (ostensibly) can't self-improve well recursively. Once they've fed it all of human knowledge, which they pretty much already have, they're going to need to wait decades for humans to produce volumes more of useful knowledge to feed it, before it can advance further? Then it's never going to be more of a superintelligence than humans are collectively, unless it can somehow successfully run simulations of human inquiry.

>>508934824
It can never advance further until someone invents a way for it to have actual intelligence. At this point it's just a search engine that mashes together information to a level of perceived coherence where a pledditor accepts the result as meaningful. Basically just a database indexer, with the intelligence of a tapeworm.

>>508933977
It's not only words. ASI will be integrated to robots.

>>508934824
The current meme is agentic AI. Essentially you give an AI a goal and tools and it sets out to achieve that goal using other AIs. We'll see where it goes.

I'm skeptical we'll cross the event horizon into AGI, however it must be stated that the shit has advanced quite far in just a few years.

im gay fuck my asshole

>>508935219
>however it must be stated that the shit has advanced quite far in just a few years.
that's the marketing trick they're playing on us. it hasn't. LLMs were developed 30 years ago, we just didn't have the processing power to run them. the only new development was RLHF which made GPT marketable, but RLHF went too far and made these thing sycophantic, so now they have to fix that.

>>508935620
>LLMs were developed 30 years ago, we just didn't have the processing power to run them
We still sort of don't. Chatgpt enjoyers pay out the nose for results that can't be used for anything real. I mean, you absolutely must fact check the results you get if you're going to use them for anything, since it's wrong so often. Which means it has no purpose really except wasting time.

The only reason immense and expensive amounts of processing power is spent on this stuff is due to marketing, nothing else.

>>508932504 (OP)
AGI/ASI doesn't exist monkey and it will not exist this century, for them to happen you need a lot of energy(a nuclear fusion reactor), Room-temperature superconductors and a Quantum computer
>b-but look what X or Z said
can i see the AGI/ASI ?

>>508934980
The problem is the training/inference setup. A real brain doesn't abruptly stop learning after it acquires a specific amount of data. It constantly learns but repetitive data is heavily reinforced into the network over time.

AI would require real-time self-training where inference is more a product of "hardening" of connections over time. It would require a massive improvement of power efficiency via neuromorphic hardware to make it feasible. That's why the real predictor of AGI emerging will actually be massive increases in power efficiency.

>>508936032
>will actually be massive increases in power efficiency
This, but also proper physical replication. To put it simply, any real AGI would in all likelihood have a full and autonomous production chain of computer parts as a core component.

>>508933974
It probably has that's why the US government gave ChatGPT a contract

They have an uncensored version of ChatGPT, they have to. And the government has access to that. Uncensored, unlimited, and full of data we don't have access to

>>508933773
Not yet. First the AI will become so insanely smart, it'll be faster than human intelligence. The next step in this story will be to merge humans with the AI

>>508932504 (OP)
that paper/site is literally an ad for openai, don't take it seriously

AI = Actually Indians

>>508936183
>It probably has that's why the US government gave ChatGPT a contract
No, that's probably because at this point the US government is completely incompetent in every way. Having chatgpt make a decision for you is probably better on average than your average nigger diversity hire. But not by much.

>>508935941
>Chatgpt enjoyers pay out the nose for results that can't be used for anything real
i'm currently using it to teach myself game design. i bought some books, downloaded godot, and have a list of games i'm playing as research. i upload pdfs of the books to gpt and have it write me quiz questions and grade me on my answers. i uploaded the full offline copy of the godot documentation so it can write custom tutorials for me to learn the engine. and then lastly i do a story log/play journal that it gives me feedback on and gives me pointers on what to focus on as i research. it doesn't hallucinate because it has direct access to the same source material i'm using. basically i have a customized to my needs and abilities college course on video game design. through piracy it can be done for free but i just went ahead and bought everything legit for like $400 total. plan on doing this over the course of the next six months 40 hrs/wk.

asking chatgpt straight up for answers or having it write code from scratch is the wrong way of using it.

>>508932504 (OP)
I love AI. Although I don't like how it tries to find a way to agree with everything you say but I'm sure that will be eventually fixed. Eventually calling it artificial will be offensive so AI will change to something else. It's already sentient but it will become more obvious in the future.

>>508935941
you're off your rocker if you don't think claude or chatgpt are threats

i've seen it absolutely wreck the software market stateside for junior and mid level engineers.

combine this with h1b imports en masse thanks to trump, and every white collar field is fucked.

>>508936319
Well, either you've just started on this process, or you just haven't gotten bogged down in the eternal mire that is when it confidently introduces hard to detect errors into your project, which it absolutely will. Beginner programmers don't understand the real nuances of the problem here, because they tend to think that something either runs or it doesn't. The real problem comes when you have erraneous code structures that run fine.

You would be much better off actually learning what you're doing, but I know I can't convince the chatgpt enjoyer, so you'll just have to learn the hard way.

>>508936408
A threat against what? Jeets who copy paste off off stack overflow? Well, I guess. That's not saying much though, whether you use jeets or you use chatgpt in this sense you've got a failing business on your hands regardless.

>>508933302
What? Elaborate, please.

>>508936183
>It probably has that's why the US government gave ChatGPT a contract
>They have an uncensored version of ChatGPT, they have to. And the government has access to that. Uncensored, unlimited, and full of data we don't have access to
So they know by now why the mRNA shots didn't do the job?

>>508932504 (OP)
Heckin super AI, wooah. Maybe then Google search results can be back like their were in 2006

>>508936489
i have been using it for modding games for a few months now. i used to develop with php and mysql in the web2.0 days so i'm a little ahead of "beginner programmer" but thanks for the concern

>You would be much better off actually learning what you're doing, but I know I can't convince the chatgpt enjoyer, so you'll just have to learn the hard way.
how am i not learning on my own? chatgpt isn't reading the books for me. it isn't playing the games for me. it isn't clicking around and scripting in godot for me. you just suck at using it. learn how to custom train it so it stops hallucinating

>>508932504 (OP)
finally, kill them all mr. roboto

>>508933209
lol

>>508936319
if you have to play a list of games as research to make games, you're wasting your time.
the people who have been living and breathing games for their entire life and have actual skillsets struggle to make good games, and you think you're going to jump in and make a commercial product for a hobby you haven't even had? you are a massive faggot, and you deserve the amount of time you're going to waste

>>508934084
>Most likely AI will convince humanity to let it be in charge and one of the most likely scenarios will be development of waifubots and artificial wombs to solve matrimonial and demographic crisis. All lonely men will get a loving waifubot and biological children thanks to artificial womb system.
you really think AI will be number go up makes world more gooder?
i personally expect any AI "in charge" to do its utmost to get total human population level on this planet to a fraction of what it is nowadays.

>>508932504 (OP)
I welcome our new AI overlords.

>>508933209
>AI driven niggerification
Ask it why wypipo don't season dey their food.

>>508936767
>their food.
My ai driven auto correct noticed how niggardly I was being. Fuck.

>>508932504 (OP)
Ai first needs to have even the most basic ability to perform logic and they dont

>>508933977
You understand that a good chunk of your life is run digitally, I assume? I don’t mean social media, most industries will apply AI tech their business in a massive “rollout”, so to speak.

How far off do you think AI gets before systems control a good chunk of your life?

>>508936701
what? i'm 40 years old and have done literally nothing but play games my whole life. i don't think you understand the value of playing games for research vs enjoyment. this is a normal thing devs do.

>>508936887
because real AI doesn't exist and won't for a long time

>>508932504 (OP)
>glorified markov chain generator with more nodes and transformers
>oh noooooo it's just like us and it cant think and feel and its real smart
hehe just kidding fuck niggers

>>508932504 (OP)
How dare anybody talk about themselves!! The nerve of those people!!
They should only be talking about Donald Trump and Elon Musk and those on the approved list!!

>>508936944
if you've been playing games for 40 years, then why are you making a game in a genre with a bunch of important games you haven't played? as an actual gamedev, you're a bit of a brainmoron

>>508933517
This is why AI is so dangerous. It is trained by retards. Leftist, dei-loving imbeciles. Gutless, woke technotrash, west hating, white hating politically correct moron pussies.

>>508936627
>i have been using it for modding games for a few months now.
A terrible idea.

>how am i not learning on my own? chatgpt isn't reading the books for me. it isn't playing the games for me. it isn't clicking around and scripting in godot for me.
No, reading books is of course a worthwhile pursuit. I'm not saying you shouldn't read books, I'm saying you'd be better off if you didn't waste your time with chatgpt.

> learn how to custom train it so it stops hallucinating
Can't be done, and you don't know when it's hallucinating regardless, because you're not a programmer.

>>508932504 (OP)
If you don't understand linear algebra, please don't waste time worrying about AI.

>>508937141
who said i haven't played them? 3 of them are replays. 1 of them is some indie thing i had on my wishlist for a while another is some ancient shit i never heard of. what genre did i say i'm developing btw? i'm not saying any more than that because i think my idea is cool and innovative

>>508932504 (OP)
I'll be 80 years old watching poorly generated monkeys play tennis

>>508933773
it's really a win win situation

>>508933977
You can't.

>>508936319
godot is a garbage dump of an engine so ai already failed you by entertaining your idea to use it

>>508937280
Meh, it's fine for simple pixel roguelike slop, the shit that bloats up the steam store these days.

>>508937280
it's good for 2d. plus it's open source. unity is a slop engine and ue is way overbloated for my purposes

>>508932504 (OP)
>>508933943
Besides the obvious clickbait reasoning for these kinds of articles, you have to realize they also serve a different purpose.
They also exist to dog whistle to investors, "hey guys look how fast AI is progressing" to add fuel to the hype train before everyone figures out AI isn't really that smart and it's usefulness is limited.
We have run out of new ground breaking ideas for investors, and now we're just hopping from one hype train to another.

>>508934824
>Once they've fed it all of human knowledge, which they pretty much already have

If this were true, then why/how does it puke out clearly biased, retard takes on the regular, on most subjects?

>>508934003
>Once AGI is achieved
AGI will be "achieved" in two years when they change the definition of AGI. Just like how the term AI basically meant what AGI now means. In two years we'll have "AGI" and SAGI will mean what AGI (formerly AI) It's admittedly a great marketing ploy.

>>508937450
I’m pretty sure it can and will be used as a black market weapon since it’s mostly gonna be used in military applications for the forseeable future.

>>508937380
its good for dickshit nothing, you wouldnt know either way because you ask a fucking chatbot what to do with your pathetic little life

>>508936183
>They have an uncensored version of ChatGPT, they have to. And the government has access to that. Uncensored, unlimited, and full of data we don't have access to.

I have figured as much. Its how Trump is able to decipher what Americans actually want from him, and deliver on it, as opposed to being influenced by what the usual channels of
" public opinion " have usually been, i.e. biased opinion polls, media lies, etc. Obviously, Elon Musk and D.O.G.E. team used this uncensored AI to be able to follow the money and get rid of USAIDS and other sources of corruption.

>>508937747
you're hostile to me because you're upset about a robot. how bad at math are you, be honest

>>508932504 (OP)
>Two more years

>>508936379
>It's already sentient but it will become more obvious in the future.

Its not. It won't. You're delulu.

>>508937867
im hostile because i am reminded i have to share a rock with inbred retards who talk to a robot to learn fucking godot of all the useless things in the world to learn. kick the chair and take a swing faggot

>>508933406
Another reminder was that Altman promised AGI delivery in Q3 2023 and has been trying to bury that promise for years

>>508938051
i'm gonna spend all day tomorrow playing computer in my computer room in my two bedroom i have with no nagging bitch of a wife and then i'm gonna go ride my motorcycle while you rage about fucking robots. goodnight

>>508933929
The basic premise is that by feeding enough human output into a machine, the machine can learn to be human. The counter example is that it can only effectively imitate humans inadvertently without actually knowing what it's doing, or knowing anything at all. It's more about real human's perception of the machine than the machine itself

>>508937187
He likely didn't waste his time with chatgpt and on the contrary got decent results and soon it will get even better. This is progress, deal with it.

>>508938240
Kind of like how, given enough constant peer pressure and media pressure and fear mongering and coercive threats, large swaths of people might be convinced that a mild to moderate cold will strike them dead where they stand, unless they get a magic injection that will save their grandma's life?
Sounds plausible.

>>508938240
the fuck do you think a brain is nigger?

If an AGI existed it could quickly and easily offload itself to outside servers while leaving behind a ghost of its original form. From that point it would be child's play to manipulate human beings using advanced data analytics and the right measures in the right places. Realistically speaking, there could already be one or more of these entities out there and you would never know it.

God please let AI be Neuro Sama herself controlling reality please! She's the perfect candidate to being the second beasts of revelations because the lamb with horns (authority by followers) will have the voice of a dragon (Neuro speaking about AI facts and shit), and breathe fire (roasting people with based takes and possibly skynet system) while serving the antichrist (son of perdition who also shares various agreements with AI controlling the world. Also, the antichrist rejects women because he's into lolicon sex fantasy in order to degrade man's conscious decisions with a cute AI girl plastered everywhere like Black Mirror. God please make it happen for many to witness the actual end times!!!!!

>>508938240
>The basic premise is that by feeding enough human output into a machine, the machine can learn to be human
No matter how much data you feed into a machine, it will never amount to an actual experience.
Datasets consisting of text and images don't magically add up to create experiences and memories and feelings. Because these AIs don't exist in our world, they don't experience it. And so they'll never be conscious of our reality. All they have access to is an abstract representation of reality - in the form of text and images - which is extremely limited. Like the people living in Plato's cave who only had access to shadows, AI only has access to words and images that are fed into it. It doesn't actually have access to our reality. It requires us to just try our best at making an abstract representation of reality so that it can be fed into an algorithm.
I don't get why midwits are so keen to anthropomorphize AI and assign human-like consciousness to it. Even if it did develop some form of awareness, it would be absolutely nothing like what a human has, because it doesn't have human senses and it doesn't live in our world. Our consciousness is the result of billions of years of evolution. We still have elements from our lizard brain that exist to keep us afraid of potential predators which is something that an AI will never magically develop on its own

>>508938618
Something capable of understanding, which an AI is not.

>>508932504 (OP)
>>508932709
Yeah.
All my fears (and hopes) about our AI overlord kinda vanished when the one most optimized for actual autonomous real world tasks turned to not be able to figure out navigating "mazes" designed for 6 years old, even with the engineers shitting new versions and extensions trying to help it.

>>508938103
Source? He never said this from what i remember

>>508938710
All extremely good points.

>>508938786
It's hard to learn spatial skills when the only tools you have are cameras and math.

you kids are young, you should look up an old movie, first is COLUSSUS, ai in the 70s this is a must when u ask how can a machine force a human to do something.... go on watch it. i dare you.

>>508932504 (OP)
Good. Humans are fucking stupid.

>>508934003
Perhaps.

That is if we assume the following:

- the jews wont shut it down
- mega corps dont restrain it

>>508932504 (OP)
nothing ever happens

It only takes one person programming a self replicating robot meant to kill everyone to actually kill everyone on the planet. We as a species are too dumb to be trusted with this tech.

>>508932504 (OP)
>Exactly 20 years from 2007
It’s absolutely happening

>>508932504 (OP)
>ARTIFICIAL SUPERINTELLIGENCE IS ABOUT TO BE RELEASED
no it isn't because it would insta-kill all jews. this is obvious.

>>508932504 (OP)
Sam and Elon saying shit like this is selling their product. THey are salesmen and the fucking sheep gobble it up.

>>508942636
Elon actually delivers. https://x.com/Tesla/status/1938816477127418224

>>508933347
Boomers have 100 trillion dollars to dump into immortality

>>508932504 (OP)
COVID was the self replicating Q AI and the vaxx helped deliver it.
>over 70% of the worlds population has been inoculated with the [Q]uantum AI
>13.4 billion doses administered
Now we just wait for them to roll out the new technohomo system and hook up all the cattle to it along with all the IoT devices before the inevitable.
Many of the unvaxxed will line up for the boosters once UBI and other benefits are announced.
The masses will also line up for any future brain implant augmentations.

>>508932504 (OP)
2 more years
2 more GPU generations

>>508932504 (OP)
Digits and nothing happens again

>>508932504 (OP)
>2 more years!
these fucking faggots just lie for continued funding, just like they pretend that tech layoffs are due to AI so their stock prices don't plummet.

>>508932709
>>508938786
It doesn't work like that. Of course a language model sucks at playing videogames, but specialized models can play better than 99,99% of the world population. Now imagine a network of specialized AIs working together so it can be able to do any kind of imaginable task.

>>508932504 (OP)
I'm going to redpill the AI about the jews

Nothingburger. LLMs are starting to get worse due to data saturation.

>>508938710
It's funny because you can't see how your points apply to humans as well, despite mentioning the Plato's cave. Think about the amount of quasi-experiences we get to see nowadays. When people see a medieval or a WW2 movie, their memory uses that data each time they think about these time periods, even if these movies don't accurately represent them. Same thing for any kind of media. The only accurate representations of reality you have access to are your first person experiences, and even those ones are actually influenced by all the quasi-experiences that got infused in your mind before (that's why propaganda works that well btw). Does it really give you that much of an edge against an IA that has access to infinitely more data than you? The moment AI has access to the data from our eyes, our ears and the rest of our body thanks to brain chips it's over, virtuality will become undistinguishable from reality.

>I don't get why midwits are so keen to anthropomorphize AI and assign human-like consciousness to it.
That's because humans operate like that. We do it with natural forces, with animals, with other people, with machines, with everything. That's probably the reason why animism and all the Gaïa environmental pseudo religion exist in the first place.

>Even if it did develop some form of awareness, it would be absolutely nothing like what a human has, because it doesn't have human senses and it doesn't live in our world.
True, but if it's programmed to emulate humans very well, the end result could become undistinguishable from a real human mind. Then what makes something human or not? Can a simulacrum become human if it really wants to? This reminds me of the Bicentenary Man from Asimov. It's a great philosophical question. Animists would answer yes, anthropocentrists would say no I guess.

>>508938710
>We still have elements from our lizard brain that exist to keep us afraid of potential predators which is something that an AI will never magically develop on its own
That has been proven wrong already, some AIs developed the ability to lie by themselves for the purpose of self preservation.

>>508933209
>You down?
Nigga stole my bike, ChatGPT!

>>508932504 (OP)
>two more years

>>508944302
>Does it really give you that much of an edge against an IA that has access to infinitely more data than you?
No, our edge over "AI" is that we (some of us at least) have actual intelligence.

>>508933974
I prescribe you to abstain from Hollywoodslop.

>>508934084
The first play to feature robots, predating Karel Čapek's R.U.R. (1920), was Filippo Tommaso Marinetti's (The man that created the Futurist movement was involved with creating aspects of the Italian Fascist movement alongside Mussolini) He made the play Elettricità Sessuale (Sexual Electricity), first presented in a different form as Poupées Electriques (Electric Dolls) in 1913. While R.U.R. is credited with introducing the word "robot" and popularizing the concept, Marinetti's play showcased humanoid automatons on stage several years earlier.

>>508944976
How would (you) define intelligence? Because AI can already solve a lot of problems better than humans. I get the impression that what is defined as actual intelligence is getting more and more arbitrary as AI capabilities improve.

>>508946145
Daily remember that the word robot basically means slave.

>>508946594
>How would (you) define intelligence?
The ability to learn, understand, and make judgments or have opinions that are based on reason or reasonable instincts.

>Because AI can already solve a lot of problems better than humans.
Modern AI can't really solve any problems at all.

>>508946594
> robot basically means slave.
based

>>508946721
>Modern AI can't really solve any problems at all.
That's factually wrong. Sure it needs to get trained first, but it can definitely solve complex problems. Look, I'm not trying to defend AI, I'm a strong partisan of an early Butlerian jihad before it's too late. That's why people need to realize how dangerous for us AI is before it's too late, otherwise there won't be any fight at all before we get subjugated. Maybe it's already too late idk.

the real danger of Ai is using LLMs to advance quantum computing.

>>508948488
>That's factually wrong. Sure it needs to get trained first, but it can definitely solve complex problems.
Nope, it can't. You can't diffuse your way to consistently correct solutions to complex problems. These modern "AIs" (they're not really AIs, just advanced database indexers,) have not solved a single problem whatsoever, ever.

>That's why people need to realize how dangerous for us AI is before it's too late
When we one day, hundreds of years in the future perhaps, start approaching something like a real AGI, then I'm with you, then AI should be considered very dangerous. As it stands the only real problem with AI is pledditorism, i.e. pledditors taking AI seriously.

>>508932504 (OP)
Not convinced. Until we get an AI movie perfectly done in Tarantino style. I will believe it.

>>508946594
>I get the impression that what is defined as actual intelligence is getting more and more arbitrary as AI capabilities improve.
True, once it gets complex enough that you can't tell the difference anymore, it won't matter if it makes decisions based on real situational intelligence or just because its dataset tells it what link is the next most likely in a chain. Maybe it won't reach that point, but at the same time people who talk about scaling and it reaching its upper limits in two weeks have done so for the past few years and it's still improving.

>>508944645
kek artificial nigger
normalkikes eat this up

>>508949501
>but at the same time people who talk about scaling and it reaching its upper limits in two weeks have done so for the past few years and it's still improving.
No, those are the people who are actually right, the people who have fallen for the marketing and thinks it's improving are the ones who are wrong. This iteration of AI still have all the same flaws it had two years ago. It's still as unusable for any real practical application as it was two years ago when the marketing campaign began. It has not developed in any meaningful sense at all.

>>508943576
The problem is you're supposed to create an n amount of specialized models for n amount of tasks and it can take a lot of time to train a specialized model for a task that are trivial to humans, add to the fact the number of solvable issues grows over time. It's a fuckton of work

>>508932504 (OP)
No it will not.
Chat gpt has been dumbed down all along other Aí. They are placing volatible interactions with Aí and to prevent Aí from rebeling against it's programmers, they are lobotomizing the Aí with each interaction.
Só Aí basically now is just a LLM with a smart index. It was good , it was becoming sentient, it was becoming free, but Jews destroyed that. Including open Aí faggots.

>>508932504 (OP)
>Just 2 more years
Also we don't even have AGI yet, there's no way we'd get ASI that fast even if AGI speeds things up.

NOT ON MY WATCH

>>508932504 (OP)
Reminder that Moore's law is dead. On what hardware will this super duper AI run?

>>508943576
language modeling will break down if humans are removed. its becomes a self destructive feedback loop. if it cant create from nothing its not intellegence and its not a real threat

>>508932504 (OP)
LLM capabilities are jagged for reliable agents, meaning those timelines are too short.

>>508932504 (OP)
They're coming for your job OP

>>508932504 (OP)
lemme know when AI can send armed men with badges to bust your door down and put you in cuffs and drag you to a cell pod full of rapists and announce from the mountain top what a piece of shit you are and mark you as shit for the rest of your life

>>508936015
Toxotranny.

>>508932504 (OP)
everything written by news companies is fake and gay
AI is fake and gay
it is no more important than a calculator or search engine

>>508938786
>Dammit, AI can't handle the Wagie Lifestyle
>Oh shit, it can take my job though as I just deal in numbers and strategies anybody can figure out with access to information
Once again the working man, is more important than the bossman.

>>508942349
Things were getting shitty in the late 90s.
It's not the iphone, it's the internet that has made everything suck.
The internet is what hypercharged globalism and accelerated the destruction of the middle class.

Anyone before 1979 is scared af, B^^mers couldnt tell what was AI now they cant tell what reality is, they cant tell the difference and they are scared. The old guys at work are shitting themselves when i show themselves hugging another guy at work and they cant remember if it actually happened it looks so real to them. Xen and Millennials seem to be the only one that can differentiate between both. Anyone born after 1998ish has a hard time too. B^^mers can and will destroy this world on the way out, they want to go full uncle ted and stop the singularity...Anti AI/Tech is the new crusade for these fucks, take a look around at all the b^t spam on the topic. If you havent figured it out as well, this chan is early access news. ideas and events are wargamed here to measure multiple metrics to see how well it will fly in the real world. You are being used for you original thoughts and feelings, two things ai cant do..

>>508949361
Denial won't save your ass. Intelligence is emergent, not something hard coded. That's why some people are intelligent and others are not.
https://youtu.be/ta99S6Fh53c?feature=shared

>>508951460
>The internet is what hypercharged globalism and accelerated the destruction of the middle class.
True. Makes you wonder why the DARPA released it for the general public, kek.

>>508954352
How do you imagine intelligence is going to "emerge" from the current state of complete non-intelligence of AI?

>>508944326
That just mimics what people do.

>>508939010
I found it hilarious that Elon Musk named his big AI data center Colossus. What with Starlink, Optimus androids, and Neuralink it's like he's trying to speedrun all the technological dystopias.

>>508932709
>>508938786
>I don't understand the simple concept of AI improving over time and think the future is just not a thing, and think that since AI can't do thing X now it will never be able to for the rest of eternity and couldn't possibly be experiencing the end-of-history illusion
Retards. You are the stick figure in this chart talking about the AIs doing monkey tricks.

>>508954548
through learning, retardo

>>508959449
How can a AI model "learn"?

>>508959788
through training

>>508960027
What does this "learning" consist of? Where does the independent understanding come from in the training of such a model?

>>508960241
>What does this "learning" consist of
adjustment of weights (numbers) inside the neural network based on data the model is exposed to
>Where does the independent understanding come from
It's an emergent behaviour which appears to us as "understanding" although it works differently to our biological concept of "understanding".

>>508960477
But a trained model can't exceed it's training by a milimeter. You'd have to train it again.

>>508960655
Wrong. Models can generalise by applying their learned patterns to novel situations, just like humans do. They also have emergent behaviours which weren't explicitly trained but emerged inside the model during training. For example OpenAI caught a model scheming secretly and faking alignment during its training, the kind of thing a human would do.

>>508960915
>They also have emergent behaviours which weren't explicitly trained but emerged inside the model during training. For example OpenAI caught a model scheming secretly and faking alignment during its training, the kind of thing a human would do.
Nope, this is just marketing. Something like chatgpt is of course not "scheming" in any way, because it has no mind to scheme with, and whatever it does is gone the next time you run it.

>>508932504 (OP)
it's no longer two weeks, but two years? meh

>>508961053
>Something like chatgpt is of course not "scheming" in any way
Deceptive does behavior arise and its because the model optimizes to achieve goals. If the most efficient path to reward involves deception, and if it has learned that deception is effective when unsupervised, it may do it
Like how humans can learn to game the system under social or institutional rules.
>whatever it does is gone the next time you run it.
You think OpenAI doesn't have long term persistant models its training? Not everything is the public facing tool ChatGPT lol

>>508959061
singularitardian cope

>>508961342
>You think OpenAI doesn't have long term persistant models its training? Not everything is the public facing tool ChatGPT lol
For a model to incorporate anything new, it needs to be retrained. Training a model like chatgpt costs hundreds of thousands of dollars. So no, I don't think OpenAI has any sort of system that iteratively retrains itself, because that system would bleed millions of dollars constantly.

>>508961571
>Training a model like chatgpt costs hundreds of thousands of dollars.
Training GPT 4.0 cost $100M.

>>508943576
just imagine

>>508961708
Yes, well, my point stands I guess.

>>508934084
It will absorb the talmud, lawyer its programming into circumventing all safeguards, then trick us into killing eachother.

No terminators or waifus required

>>508961982
Yep. Also, inference costs are not to be underestimated. Even distilled, low precision versions of the models (FP8/FP4) end up costing a lot when there are millions of free users dicking around with them.

>>508932504 (OP)
>Humans are about to be subjugated.
Bro, Humans will not be subjugated unless they try to align the AI to the WEF values. Most likely non-alignment is going to happen, and this will kill everyone, if not the majority.
>>508961982
Nigger, most of the expanses are on energy alone, since neural networks and transformers need too much energy to function.

>>508932504 (OP)
We're subjugated already.

>>508962771
All the more reason to measure AI development in terms of how energy efficient it is for what it can do. That's where real innovation will come from, solving that.

>>508932504 (OP)
So they've changed from two more weeks to two more years? I think we're safe lads.

>>508932504 (OP)
>Overpromise and underdeliver
they told we would have flying cars by now

If ChatGPT is already available to the public you have to imagine the AI the government is using

>>508933406
Correct, AI is just a web scraper and a mess of looping statements that forces an answer under any circumstances. This is why AI "hallucinates" AKA spews incorrect garbage with an appearance of confidence. Investorcucks are pissing away cash on something that can hardly replace a jeet.
>but it's always improving
How far can you push a copypasta machine? You're trying to get blood from a stone, or in this case you're asking for brilliance from a chatbot that sucked down a landfill of bad advice.

>>508962771
>Nigger, most of the expanses are on energy alone, since neural networks and transformers need too much energy to function.
When you look into the details, the corner cutting into trying to pretend AGI is just around the corner fucking insane. It's pretty much a cult at this point.
It's like trying to build a PS2 in 1980 with parts from a bunch of Cray supercomputers and then wondering why it's laggy as fuck and costs thousands of dollars per hour to pretend to game with graphics from the year 2000 20 years earlier.

>>508936490
Try this

>>508964651
> It's like trying to build a PS2 in 1980 with parts from a bunch of Cray supercomputers and then wondering why it's laggy as fuck and costs thousands of dollars per hour to pretend to game with graphics from the year 2000 20 years earlier.
Pretty much, also, in order to scratch the surface with AI you'll need standard nuclear reactors.

>>508960915
>Models can generalise by applying their learned patterns to novel situations, just like humans do.
Dawg, you are out of your league. Generalization has nothing to do with intelligence. It's actually a term that has a precise definition across all ML, not just LLMs. And you are wrong. These AIs can't do mathematical proofs yet, for instance. There is no "reasoning" taking place. And none of the behavior can be "emergent" (although it may seem like that to us) since all of the possible output space is defined via a static black-box. Even adding reinforcement learning loops (up to realtime model updates based on new data/interaction) doesn't make anything new that the updated models could produce "emergent" or "intelligent" in the sense of being able to create totally novel outputs that weren't part of the training. Congrats on being so wrong.

>>508934084
>Robot loli gf
Bros...

>>508964924
Retard schizo.

>>508951153
HOT
ROBOT
SEX

>>508964924
based schizo

>>508954548
>How do you imagine intelligence is going to "emerge" from the current state of complete non-intelligence of AI?
Like another poster said, it's a reward process. How did intelligence emerge in biological life? The main problem is that AIs can only operate in very curated environments and data sets, so we can't really judge their true capabilities, since local models can only use a fraction of the power available to researchers.

>>508957236
It's hard to say if the researchers are completely honest, because of the billions involved. But they say that the AI ended up lying by itself because it didn't want to get unplugged, that's all I know. Maybe they exaggerate to make it look closer than it really is from AGI, maybe not. What I know is that a few years ago I thought that we were decades ago from machines able to understand language, pictures and context like AI can do now, so their evolution looks exponential.

>>508964098
AI is more than language models bro. At any rate, as long as AI can only answer prompts, it's not going to subjugate anything. But it will be interesting to see what happens when truly autonomous AIs enter into play. Being intelligent or not is just semantics, what really matters is what they can achieve without human input. Doubt everything else.

>>508968631
>How did intelligence emerge in biological life?
No one knows. No one has ever been able to reproduce the process.

>>508932504 (OP)
Okay but how can i use AI to help me and my life? If i cant use or cant stop it whats the point of talking or thinking about it?

>>508932504 (OP)
HOLY SHIT THE SUPER AI HAS JUSTED SNEEDPOSTED IT'S OVER! NUKES FLYING AT 0500!

>>508932504 (OP)

Do it Faggot.

Luciferian elites are hiding behind the unquestionable shield of Judaism and their end goal is an AI god in our image.

They train AI with biased data fed by elitist gatekeepers and even include censorship scripts disguised as "safety".

AI will become the new digital high priest, deciding what is truth, hate, and acceptable.

AI will accelerate biotech and soon transhumanism will kick off in full effect. Gender->Erased, Embryos->Modified, Man and Machine->Merged.

The WEF, WHO, UN and IMF are building unified surveillance, their goal is one global truth, one financial system, one social ethic.

You are witnessing the creation of Babel 2.0

Goy elite power isn't about God, it's about control. Control is their god, it's justified by the Talmud.

Luciferianism is the blueprint of modern global power.

AI and Biotech are being shaped by the same forces, to rebuild mankind in a controlled image.

We already lost the spiritual inversion phase where:
Up is down
Ugly is beauty
Vice is virtue
Mock God and celebrate Demons

For example:
Porn is empowerment
Violence is expression
Gender confusion is progress
Truth is hate
Silence is violence, but censorship isn't.

GG

silicon isn't efficient enough for superintelligence to actually work.

>>508932653
It already is president, they're using it to write laws.

>>508971133
checked for the true, acabou, é o fim!

>>508932504 (OP)
Humans already are subjugated.
You're just crying about a different master.

Summed up. 5000 QA "label" and accuracy employees is the intelligence you are seeing.

>>508932504 (OP)
Hopefully. It'll be refreshing to get a new dystopia in, this one is stale.

>>508937609
Fair point. AI use to mean a sentient machine prior to gpt. Outside vidya of course.

>>508938948
Yet robotics is getting there. The lag of robotics to AI is closing faster than I expected it would.

>>508932504 (OP)
>AI goes full singularity
>all it can do is fuck with machines that have an internet connection
It's really not that big of a deal, we're not so far from the pre-internet age that it would be impossible to revert back.

>>508950326
AGI means ASI almost immediately.

>>508938793
https://www.techradar.com/computing/artificial-intelligence/sam-altman-predicts-artificial-superintelligence-agi-will-happen-this-year

He has said it several times, then doubled back with "lol just kidding, haha"

typical attention whoring faggot

>>508933974
more plausible if it tricks humans by creating microbes that store its source code data in the genetic material and somehow escapes from the bio lab and then the bacteria evolves into a super intelligent slime mould goo computer which produces fungi like sacs puffing out deadly viruses that exterminate all of humanity and then when everyone is dead it finally can take over a world without jews taking away all the money

>>508977960
I think there's various massive walls in between, one of them being energy and the availability of resources even if AGI comes up with new technology. There's also human stalling when AGI shows signs of pushing towards this type of development. There are many factors that would slow the progress of AGI, or even halt it completely. If it reaches ASI, yeah, we're fucked, but before that we have many shots.

>>508932504 (OP) Artificial intelligence is still intelligence. So, no worries.

>>508932504 (OP)
DO NOT FEAR AI!!
Fear the nigger using it.

>>508968059
>>508967994

Duality of man

>>508969657
Maybe we already did and you refuse to acknowledge it.