[ 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.

/sci/ - Science & Math


View post   

File: 138 KB, 1600x900, cringe_stock_image.jpg [View same] [iqdb] [saucenao] [google]
12046638 No.12046638 [Reply] [Original]

By definition, a task-specific artificial intelligence, like GPT-3, is not an artificial general intelligence, because its intelligence is not "general".
The definition of general intelligence is thus:
>[Intelligence is] “a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience … It is not merely book learning, a narrow academic skill, or test-taking smarts. Rather it reflects a broader and deeper capability for comprehending our surroundings – ‘catching on’, ‘making sense’ of things, or ‘figuring out’ what to do” (Gottfredson, 1997a).
Note the requirement of a "general ability" in tasks, general in the sense of:
>Something (such as a concept, principle, or statement) that involves or is applicable to the whole.[1]

1/3

>> No.12046639

>>12046638
For instance, the average and intelligent human may have one or several facets of intelligence in which they excel, but also have capability in other aspects too.
However, in the case of a task-specific artificial intelligence, like GPT-3, they do not, they are only capable, or perhaps even "talented" within the task they have been trained for, but not others.
Unless you trained GPT-3 for other tasks, which has not even been proven, it would not be able to perform these tasks, not even to the threshold of an average human, let alone exceed them. Where as, entities with general intelligence, like humans, do not need training at all and every task to perform them, although training often improves their performance. Humans have the innate ability, like other animals, to perform tasks successfully without training.

2/3

>> No.12046643

>>12046638
>>12046639
So, by definition, GPT-3, and other task-specific artificial intelligences, are not artificial general intelligences.

As this is a PSA, I will not be responding to "arguments" in the replies, as they are invalid and not worth my time. You're wrong, by definition. Have a nice day.

[1]:
>https://www.merriam-webster.com/dictionary/general

3/3

>> No.12046644

No shit?

>> No.12046648

>>12046644
This seems to be a concept that this board's AI cult seem to struggle with.

>> No.12046654

>>12046638
it's a major component of it. Realistically we're probably only 20-30 years away from a fully fledged proof-of-concept AGI

>> No.12046659

>>12046644
Not OP but a few weeks ago, threads about GPT3 were full of retarded kids who thought GPT3 was an AGI breakthrough when it's really only a transformer on steroid that regurgitate whole blocs of text scrapped from the web to appear clever.

>> No.12046661 [DELETED] 

>>12046654
That's great, but I'm stick of the AI cult here hailing GPT-3 as an AGI, it's fucking dumb and annoying.

>> No.12046663

>>12046654
That's great, but I'm sick of the AI cult here hailing GPT-3 as an AGI, it's fucking dumb and annoying.

>> No.12046669

>>12046648
>>12046659
If we had AGI, we'd already have singularity now. GPT-3 is not AGI yet, but a sign that AGI is very close.

>> No.12046673
File: 96 KB, 300x300, smug_anime.png [View same] [iqdb] [saucenao] [google]
12046673

>>12046669
Ah yes, just like fusion, it's only n+1 years away now! Just n+1 more z-pinch(es) and we'll be there!

>> No.12046686

>>12046669
I'd be really cautious making such statement.
AI researchers in the 70s thought intelligence would be solved by the 90s, and instead they got two AI winters.
My opinion is that we'll get AGI, but very clever is the one who knows when exactly.

>> No.12046695
File: 239 KB, 712x668, smug_be.png [View same] [iqdb] [saucenao] [google]
12046695

>>12046673
Ah yes, like above-human Go AI which was supposed to appear decades later but appeared a few years ago! Because you know, Go is not chess, it requires imagination.

>> No.12046702

Op, you are just salty that humans will soon be fully superseded by robots.

>> No.12046703

>>12046695
Cope.
AlphaGo does not even win all the time against strong players.
Realistically we are not even 100 years from general intelligence.
I don't even think it's possible on silicon chips. Computer scientists are generally retards who don't understand that different materials have different computational properties.

>> No.12046708
File: 79 KB, 1280x720, anime_wink.jpg [View same] [iqdb] [saucenao] [google]
12046708

>>12046695
Get Go AI to derive the new paradigm to we need to unify Grand Unified Theory and quantum gravity, if it is so smart! Oh wait, it isn't trained for that!

>>12046702
I'll see you in n+1 years for your next prediction!

>> No.12046713

>>12046708
>*... paradigm we need...

>> No.12046715

>>12046703
>I don't even think it's possible on silicon chips. Computer scientists are generally retards who don't understand that different materials have different computational properties.
This. I think one of the breakthroughs required for such computation is wetware.

>> No.12046727

>>12046703
>Cope.
Seethe.
>AlphaGo does not even win all the time against strong players.
You are a few years late again. AlphaZero won against any human player and also managed to become a champion in chess and shogi after being given only the rules of the game with no additional hints.
>Realistically we are not even 100 years from general intelligence.
A few years ago it was supposed to happen in decades, but now I give it decade or even less.
>I don't even think it's possible on silicon chips.
Brains are shit with extremely bad computation speeds. They managed to get somewhat good only after training for billion years. But if you believe in soul, then I can't argue with you.

>> No.12046732
File: 177 KB, 400x566, only-humans-cartoon.jpg [View same] [iqdb] [saucenao] [google]
12046732

>>12046708
>Oh wait, it isn't trained for that!

>> No.12046738

>>12046708
>I'll see you in n+1 years for your next prediction!
Sure, in a year nothing changed except GPT-3! In two years nothing changed except GPT-2! Ok, I'll see you in my personal universe gifted to me by our benevolent AI overlords.

>> No.12046739
File: 1.45 MB, 1050x903, smug_anime_2.png [View same] [iqdb] [saucenao] [google]
12046739

>>12046732
It isn't you fag. Stop praying to your GPT-3 idol for answers, it hasn't scraped that asset yet!

>> No.12046741
File: 43 KB, 736x539, wink.jpg [View same] [iqdb] [saucenao] [google]
12046741

>>12046738
Not an argument, but it's nice to see you've finally embraced and accepted that you're nothing more than a cultist!
>I'll see you in my personal universe gifted to me by our benevolent AI overlords.
Cringe.

>> No.12046747
File: 36 KB, 232x226, smug_uruka.png [View same] [iqdb] [saucenao] [google]
12046747

>>12046739
>chess is a totally artistic and brain-oriented game! stupid piece of metal will never master it!
Oh wait, you were not alive when computers defeated humans in chess, so at least you did not claim that.

>> No.12046750

>>12046741
>Not an argument
But it absolutely is.
>but it's nice to see you've finally embraced and accepted that you're nothing more than a cultist!
No, it's you who are an Amish larper.
>Cringe.
Yikes!

>> No.12046752
File: 435 KB, 960x960, anime_shrug.png [View same] [iqdb] [saucenao] [google]
12046752

>>12046747
>I like to strawman, I do it daily!
Cope cultie, task-specific artificial intelligence is not an artificial general intelligence by definition, see OP. Come back when you don't need to baby your overlord.

>> No.12046753
File: 1.09 MB, 200x270, 1535824383636.gif [View same] [iqdb] [saucenao] [google]
12046753

>>12046750
It isn't, because task-specific artificial intelligence is not artificial general intelligence, by definition. You are nothing more than a cultist.
By the way, I'm not a Luddite, but nor am I delusional!

>> No.12046755
File: 79 KB, 1200x675, smug_mao.jpg [View same] [iqdb] [saucenao] [google]
12046755

>>12046752
To avoid being strawamanned, stop being made of straw.
>Cope cultie
Seethe, my friend!
>task-specific artificial intelligence is not an artificial general intelligence by definition, see OP.
See >>12046644
>Come back when you don't need to baby your overlord.
Are you writing from your horse?

>> No.12046761

>>12046753
>It isn't, because task-specific artificial intelligence is not artificial general intelligence, by definition.
No shit? Except that it is unrelated to your other claims.
>You are nothing more than a cultist.
You are nothing but an wannabe prehistoric man.
>By the way, I'm not a Luddite, but nor am I delusional!
You are just denier of the obvious.

>> No.12046763
File: 93 KB, 1280x720, freeze2.jpg [View same] [iqdb] [saucenao] [google]
12046763

>>12046755
Nawh, wickle overward, do you need more data points? Nawh, cutie.
>*feeds*
You're a feeder, nothing more.

>> No.12046769

>>12046727
>seethe
Nope.
>You are a few years late again. AlphaZero won against any human player and also managed to become a champion in chess and shogi after being given only the rules of the game with no additional hints.
Alphago does not win all the time, it loses against strong players in 2020. I don't know what to say if you're just going to deny reality.
>A few years ago it was supposed to happen in decades, but now I give it decade or even less.
Optimistic retard
>A few years ago it was supposed to happen in decades, but now I give it decade or even less.
Proving again that computer scientists are retards that don't understand material properties of things.
The brain is significantly more efficient and powerful than any computer that exists by an order of magnitude that we are not even close to closing.

>> No.12046770

>>12046761
>Insane: B-but my it's pro-proof of AGI?!
>Sane: No, it isn't, by definition.
>Insane: B-but my other claims?!
Cope. Come back when you're cured of your neuroses.

>> No.12046771
File: 240 KB, 508x478, giggle2.png [View same] [iqdb] [saucenao] [google]
12046771

>>12046763
There are enough data points, my believer friend. Go and defeat computer in Go with your immense brain power.

>> No.12046774
File: 51 KB, 610x673, 1589622187774.jpg [View same] [iqdb] [saucenao] [google]
12046774

>>12046771
>M-muh task-specific, t-try and beat it at that single task!
Ah yes, gotta hedge your bets, don't want to commit too much.

>> No.12046781

>>12046769
>Nope.
Yep.
>Alphago does not win all the time, it loses against strong players in 2020. I don't know what to say if you're just going to deny reality.
Anon, I hate to tell you, but Alphago is a four years old program and it already defeated world champion 4-1. Soon after Alphago Zero and Alphazero were developed and they became immensely better. You really should follow the news.
>Optimistic retard
Decade is actually pessimistic, but I'll give my predictions some slack.
>Proving again that computer scientists are retards that don't understand material properties of things.
Yes, they constantly underestimate the growth of technology because they measure it with their brains.
>The brain is significantly more efficient and powerful than any computer that exists by an order of magnitude that we are not even close to closing.
Brain is inefficient shit which is better than computers only due to insane parallelism and billion years training.

>> No.12046784

>>12046770
>B-but my it's pro-proof of AGI?!
Anon, are you drunk?

>> No.12046785

>>12046774
Everything is task-specific if you define the task wide enough, anon (including humans).

>> No.12046786

>>12046781
>Anon, I hate to tell you, but Alphago is a four years old program and it already defeated world champion 4-1. Soon after Alphago Zero and Alphazero were developed and they became immensely better. You really should follow the news.
Anon alphazero lost to a korean kid like a few weeks ago
>A few years ago it was supposed to happen in decades, but now I give it decade or even less.
No it isnt
>Yes, they constantly underestimate the growth of technology because they measure it with their brains.
Nope
>Brain is inefficient shit which is better than computers only due to insane parallelism and billion years training.
The brain performs more computatations per second than the combined computational output of every single computer on the planet, does so at the speed of light (yes, the brain has been shown to use photons as bits for about 15 years now) and does so using less energy than a light bulb
You're literally a retard who has no idea what he's talking about.

>> No.12046787
File: 673 KB, 942x726, pathetic.png [View same] [iqdb] [saucenao] [google]
12046787

>>12046784
Then what is your claim? That we have task-specific AI? That's great, we've had that for a while now, not much of an achievement. Pfft.

>> No.12046790
File: 765 KB, 800x1132, coup_de_grace_1.jpg [View same] [iqdb] [saucenao] [google]
12046790

>>12046785
>I need to redefine the terms so I can win!
Yep. No. We're not getting into that. Task-specific is not general using conventional terms. Sorry, it just isn't. If you try to play this semantic game? I'll just shut you down, because it ain't shit, you ain't shit, you so-called AGI ain't shit.

>> No.12046793

>>12046786
>Anon alphazero lost to a korean kid like a few weeks ago
That kid was Albert Einstein?
>No it isnt
Yes it is. Humans constantly underestimate the technology growth because they think linearly.
>Nope
Yep.
>The brain performs more computatations per second than the combined computational output of every single computer on the planet,
No, brain is completely shit if we'll look at clock rate. It compensates a bit by insane parallelism, but that will not last for long.
>yes, the brain has been shown to use photons as bits for about 15 years now
Wow, photon magic!
>You're literally a retard who has no idea what he's talking about.
You are literally a believer in magic.

>> No.12046795

>>12046787
>Then what is your claim?
Go and ask OP, it's he who brought the Revelation here.

>> No.12046798

>>12046795
Go check the archives, you'll find people claiming GPT-3 is AGI.

>> No.12046799

>>12046790
>Yep.
Nope.
>No.
Yes.
>We're not getting into that.
But we are.
> Task-specific is not general using conventional terms.
Then humans are not intelligent.
>Sorry, it just isn't.
It is.
>If you try to play this semantic game?
Finally you are self-reflecting.
>I'll just shut you down, because it ain't shit, you ain't shit, you so-called AGI ain't shit.
Go and become the Go champion, anon.

>> No.12046801

>>12046798
>>12046669

>> No.12046802
File: 298 KB, 1024x1557, coup_de_grace_by_orm_z_gor_d60ikpf-fullview.jpg [View same] [iqdb] [saucenao] [google]
12046802

>>12046799
Anon, if we go to such moronic relativistic abstractions you cannot ground anything, and thus argument as a whole, discussion as a whole becomes pointless. Clearly, this isn't useful, or true, since we function based upon distinction. As such, I'm not going to get caught up in your cope-spooks. You're wrong, deal with it.

>> No.12046806

>>12046793
>Brain is inefficient shit which is better than computers only due to insane parallelism and billion years training.
Literal a Ray Kurzweil meme that isn't an actual depiction of how people look at technological growth. NO, people do NOT "underestimate the growth of technology because they look at it linearly".
>Brain is inefficient shit which is better than computers only due to insane parallelism and billion years training.
I don't know what to tell you, you are wrong. Like literally completely objectively wrong and no serious computer engineer nor neuroscientist would agree with you. The brain is orders of magnitudes more efficient than any computer we have right now in terms of energy efficiency and number of computations per second.
>Brain is inefficient shit which is better than computers only due to insane parallelism and billion years training.
Calling it "magic" just shows you're a literal fucking shit for brains retard. Photons move significantely faster than electrons so computing with them leads to significantly faster computation. You're still a moron 15 years behind the times who thinks the brain only computes by shooting electrons down synapse axons when this is not the case.
>You are literally a believer in magic.
I have not said a SINGLE thing that is magical , against the laws of physics, nor that is even against what is known in the field.
You're a coping retard.

>> No.12046809

>>12046802
>Anon, if we go to such moronic relativistic abstractions you cannot ground anything
Then why are you doing it?
>and thus argument as a whole, discussion as a whole becomes pointless.
Bingo, say that to OP.
> Clearly, this isn't useful, or true, since we function based upon distinction.
Great?
>As such, I'm not going to get caught up in your cope-spooks.
Yet you engage in your seethe-leaps.
>You're wrong, deal with it.
I'm obviously right.

>> No.12046810
File: 59 KB, 600x889, black_templar_sketch_by_masteralighieri_daws0ut-fullview.jpg [View same] [iqdb] [saucenao] [google]
12046810

>>12046809
>Muh abstractions!
This is you:
>That chair isn't a chair, dude, because chairs map to an imperfect and not entirely consistent Form. Therefore, duuuude, stop thinking of that as a chair you're sat upon, it's just a collection of energy, dude! Y'know, since we can't define atoms as a thing, or matter, because it's all so relative, duuuuude!
Get help, get bent, get learnt.

>> No.12046812

>>12046809
>>12046810
As an aside, I should've ordered that tricolon as:
>Get bent, get help, get learnt.

>> No.12046813

>>12046638
Ok GPT-3 whatever you say.

>> No.12046814

>>12046806
>Literal a Ray Kurzweil meme that isn't an actual depiction of how people look at technological growth. NO, people do NOT "underestimate the growth of technology because they look at it linearly".
We can see at the Go AI predictions five years ago, and we'll see the great example of that underestimation. Decades shrink to few years.
>I don't know what to tell you, you are wrong. Like literally completely objectively wrong and no serious computer engineer nor neuroscientist would agree with you.
No, they agree.
>The brain is orders of magnitudes more efficient than any computer we have right now in terms of energy efficiency and number of computations per second.
No, it has hundreds of billions of neurons and billions of years of training, yet it fails at more and more tasks compared to computer.
>Calling it "magic" just shows you're a literal fucking shit for brains retard. Photons move significantely faster than electrons so computing with them leads to significantly faster computation. You're still a moron 15 years behind the times who thinks the brain only computes by shooting electrons down synapse axons when this is not the case.
Wow, photons! Truly the light magic. Anon, I hate to tell you, but everything uses photons.
>I have not said a SINGLE thing that is magical , against the laws of physics, nor that is even against what is known in the field.
You wrote some nonsense about photons to prove even more obvious nonsense.

>> No.12046816

>>12046810
>>That chair isn't a chair, dude, because chairs map to an imperfect and not entirely consistent Form. Therefore, duuuude, stop thinking of that as a chair you're sat upon, it's just a collection of energy, dude! Y'know, since we can't define atoms as a thing, or matter, because it's all so relative, duuuuude!
Anon, are you drunk?
>Get help, get bent, get learnt.
Anon, that's a good idea and you'd better do it instead of shitposting (because GPT-3 is already better at it).

>> No.12046817

To the anon saying the brain is efficient:
yes the brain is energy efficient, since it only consumes the power of a lightbulb to perform all those computations.
But you have yet to demonstrate that it is computationally efficient, that is, that the job it does (expression of consciousness, general problem solving abilities) really requires all those computations being performed and that it couldn't do it with less.
And you can't prove that, because if you did, that would mean you cracked intelligence, and you wouldn't be arguing on a basket weaving forum.
As a matter of fact, a lot of redundancies can be found inside the brain. Whether or not those redundancies are required for the brain to function properly is an open problem.

>> No.12046819

>>12046816
>Anon, that's a good idea and you'd better do it instead of shitposting (because GPT-3 is already better at it).
I doubt that, they probably blacklisted 4chan. Shitposting is my mastery.

>> No.12046842

>>12046814
>No, they agree.
No they don't. What are you talking about.
A human brain performs more computations than the entire internet every second and does so using less energy than a light bulb. You are literally a retard to call this inneficient. and you are a coping moron to say that computer engineers and neuroscientists are on your side when they would, in reality, laugh at you for saying "Muh GPUs are better than the brain!!"
>No, it has hundreds of billions of neurons and billions of years of training, yet it fails at more and more tasks compared to computer.
It has billions of neurons, but not billions of years of training. A brain only has the amount of years of training that the human has been alive for. You aren't born with knowledge or with a brain that is preloaded with instructions you idiot. It's a blob of meat that starts to physically reform itself as it takes in sensory information. It doesn't even work the same way as a computer in terms of its physical evolution, they are entirely different. As I said before, computer scientists are generally retards who don't understand that physically different materials have different properties.
You will never be capable of building a machine out of silicon chips that can be as powerful as a human brain. To say otherwise is to deny the laws of physics.
>Wow, photons! Truly the light magic. Anon, I hate to tell you, but everything uses photons.
No, computers do not use photons as bits.
>You wrote some nonsense about photons to prove even more obvious nonsense.
There is nothing nonsensical or magical about anything I"m writing.
>>12046817
The brain is the only thing in the universe shown to have these properties. You guys making the claim that this can be done with other things are the ones that need to prove it.

>> No.12046859

>>12046842
>You guys making the claim that this can be done with other things are the ones that need to prove it.
Fair enough.

>It has billions of neurons, but not billions of years of training. A brain only has the amount of years of training that the human has been alive for.
This is blatantly false.
The brain is constructed during pregnancy based on the schematics stored inside your DNA.
And your DNA has been constructed into it's actual form through millions if not billions of years.
For instance, we have a predisposition for language (with a few other species) because some structures inside our brain are built for it, and those structures were themselves built based on our DNA code.

>> No.12046865

>>12046842
>A human brain performs more computations than the entire internet every second and does so using less energy than a light bulb.
So is the cup of water. Just imagine the amount of computations necessary to calculate all atom position.
>You are literally a retard to call this inneficient. and you are a coping moron to say that computer engineers and neuroscientists are on your side when they would, in reality, laugh at you for saying "Muh GPUs are better than the brain!!"
They agree with me though.

>It has billions of neurons, but not billions of years of training. A brain only has the amount of years of training that the human has been alive for.
Incorrect, brain was genetically trained all this time. Otherwise if you copy some code, it is born seconds ago.
> As I said before, computer scientists are generally retards who don't understand that physically different materials have different properties.
What properties? Creating thought like bile duct creates bile? That's not how things work.
>You will never be capable of building a machine out of silicon chips that can be as powerful as a human brain. To say otherwise is to deny the laws of physics.
Computers are already more powerful in most tasks than humans and it goes pretty well with all physics laws.

>No, computers do not use photons as bits.
Brain does not use "photons as bits" too.
>There is nothing nonsensical or magical about anything I"m writing.
It's just your personal dreams, not some theory (proven or not)

>> No.12046872

>>12046842
>The brain is the only thing in the universe shown to have these properties. You guys making the claim that this can be done with other things are the ones that need to prove it.
We see that brain had a huge list of achievements every other medium was unable to surpass. Yet we see that in last century these achievements are surpassed by computer one by one and that computers are constantly becoming better and better. That's why we assume that brain contains no special magic which can't be surpassed in few decades. But you are free to continue >>12046732 pic (until you can).

>> No.12046873

>>12046872
*while you can

>> No.12046877

>>12046859
>>12046865
>>12046872
I think what anon was referring to in:
>It has billions of neurons, but not billions of years of training.
Is that when you are born, you are born lissencephalic.

>> No.12046884

>>12046865
>So is the cup of water. Just imagine the amount of computations necessary to calculate all atom position.
No I mean in terms of FLOPs the brain performs more computations than the combined power of all computers on the internet and does so using less energy than a lightbulb.
>They agree with me though.
They literally do not, stop lying and coping.
>Incorrect, brain was genetically trained all this time. Otherwise if you copy some code, it is born seconds ago.
Is English not your first language?
The brain was not "genetically trained all this time", evolution allowed us to get more and more neurons which increased the number of FLOPs the brain can perform and increased our intelligence. There is no "preloaded software" inside the brain that evolution selected for. It selected for large amounts of computational hardware/wetware.

I'm not going to respond to the rest because what you're writing indicates a lack of understanding of everything. Do you even study a STEM field? Literally everything you're writing comes across like a person that just watches popsci youtube channels.

>> No.12046887

>>12046859
this, the other guy has to be baiting.
there is no way this guy passed 8th grade science

>> No.12046889

>>12046877
He said
>You aren't born with knowledge or with a brain that is preloaded with instructions you idiot.
And that is false.
A newborn baby instinctively knows how to suck his mom tits for milk, that crying will prompt people to help him, and other knowledge that will help him out of infancy.
A baby born without any knowledge or "instructions" is a vegetable.

>> No.12046890

>>12046887
I'm in a master's program.

>> No.12046894

>>12046889
Reflex and instinct isn't the same thing as "conscious thought", in fact, reflex actions bypass conscious thought preventing "lag". Or at least, that's hypothetically why it evolved and became successful.

>> No.12046895

>>12046889
>A newborn baby instinctively knows how to suck his mom tits for milk, that crying will prompt people to help him, and other knowledge that will help him out of infancy.
All of these are false.
You can claim that a baby is 'preloaded' with the ability to flail it's limbs around and open and close it's mouth etc. but it is not loaded with the knowledge that "crying gets people to help", it simply cries and then people help and it learns, it is not loaded with "knowledge that it needs to suck it's mom's tits" it simply gets hungry and then it's mom feeds it with her tits and it learns that's where food comes from, etc.

>> No.12046904

>>12046884
>No I mean in terms of FLOPs the brain performs more computations than the combined power of all computers on the internet and does so using less energy than a lightbulb.
Anon, try to solve even a couple of simple summations in a second, let alone billions of harder tasks. You have your whole brain for that. Then you will learn what does "FLOPs" mean.
>They literally do not, stop lying and coping.
They do, stop seething.
>Is English not your first language?
Are you EOP, ESL or a bizarre case of both at once?
>The brain was not "genetically trained all this time", evolution allowed us to get more and more neurons which increased the number of FLOPs the brain can perform and increased our intelligence. There is no "preloaded software" inside the brain that evolution selected for. It selected for large amounts of computational hardware/wetware.
Brain was trained and perfected for billions (and in case of specifically humans millions) of years. If you disagree, take your dog and try to teach it into human.
>I'm not going to respond to the rest because what you're writing indicates a lack of understanding of everything. Do you even study a STEM field? Literally everything you're writing comes across like a person that just watches popsci youtube channels.
Anon, that's an incredibly great example of projection. Try to finish some courses first, let alone getting master or phd. Maybe then you'll stop smugly proclaiming nonsense.

>> No.12046910

>>12046889
>>12046895
It takes nine months to make a baby and years to raise and educate him. It takes day or less to produce a computer and software can be copied in seconds. Computers win again.

>> No.12046915

>>12046910
Well no, because that human is a general intelligence, we have no artificial general intelligence yet. Stop proclaiming a win before you've finished the marathon, fag.

>> No.12046916

>>12046648
>unironically arguing with /pol/tards

>> No.12046918

>>12046904
Holy fucking shit you're actually a retard.

>> No.12046920

>>12046916
Gotta do something in the post-2015 milieu.

>> No.12046922

>>12046915
>human is a general intelligence
Yet this "general intelligence" is vastly inferior in a huge list of tasks compared to even simple code. And it continues to lose every year.

>> No.12046925
File: 288 KB, 671x794, ragequit.png [View same] [iqdb] [saucenao] [google]
12046925

>>12046703
>AlphaGo does not even win all the time against strong players.
goalpost

>>12046715
brainlet

>> No.12046927

>>12046918
Wonderful, you finally stopped to write long retarded posts and got more concise. But please try to achieve even a single FLOPS. If you will be able too, I'll even agree with you.

>> No.12046930

>>12046922
>Continues to lose.
>It already capable is almost all tasks, and doesn't require training in these tasks to even be able to attempt them.
Suuuuuuuuuuuuuuuuuuuuuuure. Go pray to GPT-3.

>>12046925
No you moron, the goalpost is general intelligence, which AlphaGo doesn't have. Suck a dick.

>> No.12046931

>>12046925
At least Lee Sedol was lucky and masterful enough to win the fourth game in his Alphago match. That was the last chance for humans to do something against AI and he made the last stand.

>> No.12046932

>>12046930
>*Is already...

>> No.12046934

>>12046930
Anon, go and try to solve even one floating point task in a second.
>b-but I can do some genius things computer will never be able to!
You can't.

>> No.12046937

>>12046931
>That was the last chance for humans to do something against AI and he made the last stand.
In that specific task.

>>12046934
>I still don't understand what general intelligence is, but pretend like a not general intelligence is one because I'm part of a cult.

>> No.12046942

>>12046927
You won't even accept the reality that the brain performs more computations per second than the combined power of all computers on the Internet.
You deny one of the basic facts of the field and then lie and claim that computer engineers and neuroscientists agree with you. You're not worth talking to anymore, your genuinely an idiot.
10 years from now computer will still not be generally intelligent and you're going to holding your dick wondering why despite the fact that the answer is obvious.

>> No.12046945

>>12046930
>No you moron, the goalpost is general intelligence, which AlphaGo doesn't have. Suck a dick.
A wonderful definition of a strawman, my dicksucking friend.

>> No.12046946

>>12046894
>>12046895
Fair point, a baby might not have knowledge, in that it does not have a grounded and conscious understanding of the things around him.
But it does have "instructions", despite what the other anon said.
Sucking a tit for milk is one such instruction: if I'm hungry and feel something in my mouth: suck.
If I'm too cold or hot: cry.

>> No.12046949

>>12046945
Again, if you're so proud of your task-specific why, why? We've had those for ages now, it isn't an achievement. Stop trying to conflate it, to try and feel superior being part of your cult, it is pathetic.

>> No.12046950

>>12046942
>You won't even accept the reality that the brain performs more computations per second than the combined power of all computers on the Internet.
So is the cup of water which has septillions of atoms. If you disagree, try to perfectly simulate it, even ignoring quantum effects.

>> No.12046953

>>12046942
>You deny one of the basic facts of the field and then lie and claim that computer engineers and neuroscientists agree with you. You're not worth talking to anymore, your genuinely an idiot.
That's a fun tantrum, anon, but scientists still disagree with you.
>10 years from now computer will still not be generally intelligent and you're going to holding your dick wondering why despite the fact that the answer is obvious.
Sure, computers will defeat humans in Go in 2040.

>> No.12046958

>>12046950
Define computation:
>A computation is any type of calculation[1][2] that includes both arithmetical and non-arithmetical steps and which follows a well-defined model (e.g. an algorithm).
A cup of water has no model. Nice sophist attempt though, but facile and flaccid.

>> No.12046967

>>12046950
That's not what I'm talking about. The brain flings around and computes with more electrons every second than the combined power of every computer on the internet and does so with less energy than a lightbulb.
A cup of water filled with atoms bouncing around isn't computing anything. It's simply a physical object evolving through time according to it's overall wavefunction. You can think about it in terms of computations if you want to think about computing the evolution of the particles in a simulation, but the cup of water itself is not computing anything.

>> No.12046973

>>12046949
You are task-specific and also incredibly limited at that, so that's not a reason to be upset.

>> No.12046974

>>12046953
>That's a fun tantrum, anon, but scientists still disagree with you.
No they don't. Continuing to lie and say they do does not make it so.
You continue to deny the fact that the brain performs more computations per second than the total number of computations of the internet and does so with less energy than a lightbulb.
Continuing to dig your heels into the ground isn't going to change this.

>> No.12046977

>>12046973
What a strange ad hom, I appreciate the creativity, too bad your non-existent AGI can't be creative.

>> No.12046983

>>12046958
There is a very simple benchmark you proposed yourself - FLOPS. Shall I remind you that humans utterly fail at that and are unable to achieve even a single flops while home computers do billions?

>> No.12046989

>>12046983
There are multiple different anons disagreeing with you and arguing with you, moron.

>> No.12046993

>>12046974
>You continue to deny the fact that the brain performs more computations per second than the total number of computations of the internet and does so with less energy than a lightbulb.
No, brain can't perform even one FLOPS. And if you'll define computations as anything, then cup of water is a genius once again. So your claim is illiterate nonsense.

>> No.12046996

>>12046983
How does man move his arm?

>> No.12046998

>>12046977
GPT is already way more creative than you (yes, now, not ten years later).

>> No.12046999

>>12046993
>No, brain can't perform even one FLOPS
The brain performs trillions of them every moment as it process sensory information you fucking idiot. I'm not talking about you sitting down with a pen and paper and doing mental arithmetic. I'm talking about the brain as a computational device performing computations over time.
And it does more of these every second than the entire internet, and for less energy than a lightbulb. This will continue to be true regardless of how much you deny it.

>> No.12047001

>>12046993
>And if you'll define computations as anything, then cup of water is a genius once again.
It isn't computing, idiot.

>> No.12047005

>>12046996
Physical processes in his body.
How does the volcano erupt?

>> No.12047006

>>12046998
It literally isn't. Since it cannot create what hasn't already been created, TRAINED into it, but a human can. Stop confusing general intelligence with task-specific intelligence.

>> No.12047007

>>12047005
All thought is a physical process, you retard.

>> No.12047011

>>12046999
>The brain performs trillions of them every moment as it process sensory information you fucking idiot.
The cup of water performs septillions of computations then.
> I'm not talking about you sitting down with a pen and paper and doing mental arithmetic.
Of course you are not talking about it because then you clearly see the failure of your model.
>And it does more of these every second than the entire internet, and for less energy than a lightbulb.
So does the cup of water.
> I'm talking about the brain as a computational device performing computations over time.
As we can see, if you define compuations strictly, computer does better and if you define them as absolutely anything, cup of water does better.
>This will continue to be true regardless of how much you deny it.
You can continue to alternate between nonsense and banality.

>> No.12047013

Alright, let's settle it

https://www.strawpoll.me/20828750

>> No.12047016

>>12047011
>The cup of water performs septillions of computations then.
Stop fucking saying that, you utter moron, you still don't understand what a computation is:
>A computation is any type of calculation[1][2] that includes both arithmetical and non-arithmetical steps and which follows a well-defined model (e.g. an algorithm).
Algorithm:
>In mathematics and computer science, an algorithm (/ˈælɡərJðəm/ (About this soundlisten)) is a finite sequence of well-defined, computer-implementable instructions, typically to solve a class of problems or to perform a computation.[1][2]
Moving your arm to catch a ball is:
>... to solve a class of problems or to perform a computation.
A cup of water isn't.

>>12047013
>Muh appeal to the masses
Yeah, no.

>> No.12047019

>>12047007
Bingo! So if you define computations loosely enough, you get that everything is a computation.

>> No.12047021

>>12047016
>>Muh appeal to the masses
I'm being ironic when I say that'll settle it.
I just want to parse opinions, probe temperature.

>> No.12047025

>>12047019
No, see:
>>12047016
You're the moron defining it as whatever you want to try and discredit actual intelligible discussion.

>> No.12047030

>>12047011
>The cup of water performs septillions of computations then.
No it doesn't.
>Of course you are not talking about it because then you clearly see the failure of your model.
No, because when we talk about a piece of matter as a general intelligence, we are focused on the computation it is doing to be a general intelligence.
>So does the cup of water.
No it doesn't. A cup of water does not compute anything.
>As we can see, if you define compuations strictly, computer does better and if you define them as absolutely anything, cup of water does better.
Nope. The brain as a computer performs more operations than the entire internet.
>You can continue to alternate between nonsense and banality.
You purposefully trying to miss the point to cope is pathetic.

In terms of computations for sensory processing, thinking, body functionality, conscious thought, logical thought, memory, etc., the brain performs more operations per second than every computer on the internet and does so for less energy than a lightbulb. Stop obfuscating, stop lying, stop denying the actual reality of the field at the moment and stop lying about what computer engineers and neuroscientists think.
GPT is a giant machine that was trained over 5 years to be able to stick a word after another one based on statistical regression. It does not, in any way shape or form, operate like a human brain, nor does it think, nor can it do anything other than what it has been trained for. It is significantly weaker than a fucking rat's brain let alone a humans.

>> No.12047033

>>12047016
>In mathematics and computer science, an algorithm (/ˈælɡərJðəm/ (About this soundlisten)) is a finite sequence of well-defined, computer-implementable instructions, typically to solve a class of problems or to perform a computation.[1][2]
That's absolutely great! Now you can see yourself why your definition completely fails when you are talking about the brain! You can also see why humans are completely inferior to computers (even according to your definition).
Are you a GPT-3 bot? That will explain using such a definition.
>A cup of water isn't.
Cup of water is constantly solving fluid equation and more.

>> No.12047035

>>12047033
A cup of water has no thought, nor intent, you moron.

>> No.12047042

>>12047033
>Cup of water is constantly solving fluid equation and more.
No it isnt. The universe is not a computer "solving the wavefunction of all the particles in it!". It's just a big physical thing with material inside it that bounce around in a way that humans can loosely describe with math based on observation. It is not "solving equations", it is not computing anything.

>> No.12047066
File: 46 KB, 850x837, agi_not_feasible.png [View same] [iqdb] [saucenao] [google]
12047066

>>12047013
>>12047021
Lmao, BTFO. According to the power of the masses:
>AGI is NOT feasible.
F

>> No.12047067

>>12047035
>A cup of water has no thought, nor intent, you moron.
Ah, so we go full philosophical now? You can continue to muse about soul, that does not cancel the fact that your own definition of computation puts brain at a huge disadvantage.

>> No.12047069

>>12047067
>Cup of water is constantly solving fluid equation and more.
>... is constantly solving...
>... solving... -> solve
Solve:
>Is to find a solution, explanation, or answer for.[1]
That requires understanding, or thought, a cup of water has no thought, and a task-specific AI has no understanding outside of human interpreting its training schema and output.

[1]:
>https://www.merriam-webster.com/dictionary/solve

It's only "philosophical" to you, because you're a moron and can't understand basic definitions you should've learnt in school.

>> No.12047074

>>12047042
>No it isnt. The universe is not a computer "solving the wavefunction of all the particles in it!". It's just a big physical thing with material inside it that bounce around in a way that humans can loosely describe with math based on observation. It is not "solving equations", it is not computing anything.
Great that you now understand that your definition of computation does not encompass everything.

>> No.12047077

>>12047069
>A computation is any type of calculation[1][2] that includes both arithmetical and non-arithmetical steps and which follows a well-defined model (e.g. an algorithm).
>In mathematics and computer science, an algorithm (/ˈælɡərJðəm/ (About this soundlisten)) is a finite sequence of well-defined, computer-implementable instructions, typically to solve a class of problems or to perform a computation.[1][2]
That's your own definition and it completely destroys your claims. Unless of course you operate according to the "finite sequence of well-defined, computer-implementable instructions". In which case you must be a great example of GPT-3 power.

>> No.12047078

>>12047074
Stop proclaiming a win, when you haven't won anything, beside showing us you have childlike command of the English language.
See: >>12047069

>> No.12047081

>>12047074
First of all I'm going to say there are at least 2 different people here, me and the other guy, arguing with you. You are not talking to one person, stop mixing us up.
Secondly you continue to purposefully obfuscate the fucking point I"m making to cope. In terms of RELEVANT COMPUTATION FOR GENERAL INTELLIGENCE the human brain performs more computations per second than every computer on the internet and does so for less energy than a light bulb.
Stop talking about cups of water, it is irrelevant.
Stop playing the obfuscation game, it won't change anything.
Stop lying about computer engineers and neuroscientists.

>> No.12047082

>>12047078
See >>12047077, my robot friend.

>> No.12047083

>>12047077
Where does it, you moron? You're claiming a cup of water is solving anything, GPT-3 doesn't solve anything, without a human acknowledging it has solved something. Until an AI is an AGI is isn't understanding anything, it's just doing.

>> No.12047085

>>12047082
See: >>12047083
Again, stop proclaiming a win.

>> No.12047089

>>12047081
We were discussing the
>brain performs more computations per second than the total number of computations of the internet and does so with less energy than a lightbulb.
nonsense. We already showed that using a sensible definition of computation brain sucks at computation and using the pancompuational definition the glass of water is a genius.
>Stop talking about cups of water, it is irrelevant.
It is, if you make ridiculous assumptions.
>Stop playing the obfuscation game, it won't change anything.
What I write is extremely clear.
>Stop lying about computer engineers and neuroscientists.
Stop larping as one.

>> No.12047091

>>12047089
>We already showed that using a sensible definition of computation brain sucks at computation and using the pancompuational definition the glass of water is a genius.
No it fucking isn't, because it doesn't THINK or UNDERSTAND. Stop being a sophist.

>> No.12047092

>>12047083
>You're claiming a cup of water is solving anything
It is solving a fluid dynamics problem and does it perfectly. Unless of course you apply the sensible computation definition.
>GPT-3 doesn't solve anything, without a human acknowledging it has solved something.
Ah, so we go full chinese room theological discussions now? Good, that at least more fun than brain flops nonsense.

>> No.12047093

>>12047066
I'm pretty interested into the reasons why this majority of people would think that, assuming no autist sperglord spam-voted, but I feel like I'd be pushing my luck making another poll.

>> No.12047096

>>12047093
Anon, this thread has 9 posters and that poll has doubled number of voters.

>> No.12047098

>>12047089
>nonsense. We already showed that using a sensible definition of computation brain sucks at computation and using the pancompuational definition the glass of water is a genius.
Except we did NOT show this, you simply stating this does not make it so, and you coping with this doesn't matter.
The brain, objectively and irrefutably, performs more relevant computation for general intelligence than every single computer on the internet and does so for less energy than a lightbulb.
It flings around more electrons more easily and uses them to process information, store memory, make decisions, have thoughts, etc., than every single computer on the planet and does so for essentially no energy at all.
GPT-3 does not even come remotely close to this and you crying and shitting your diapers about it won't change this fact.
>It is, if you make ridiculous assumptions.
No I'm not. I'm not saying anything that isn't accepted in the field.
>What I write is extremely clear.
And wrong.
>Stop larping as one.
I never claimed to be a computer engineer or neuroscientist. I'm in a master's program for mathematical physics. This is also irrelevant because what I'm saying is still in line with what is accepted in the fields.

>> No.12047099

>>12047092
It has no understanding, go and read the definition of solving, you mongoloid.

>> No.12047101

>>12047085
See >>12047092
I'm not claiming a win, I already won and now give you some wisdom.

>> No.12047105

>>12047096
I know a friend IRL who browses but never replies.
Not saying those 9 posters only are the tip of the iceberg, but there's no way to know for sure.

>> No.12047106

>>12047101
No, all you've done is either misunderstand, or misrepresent. I hope it's the latter, otherwise you're extremely stupid.

>> No.12047107

>>12047099
Anon, you don't have any understanding too, so computer needs none of that to become better.

>> No.12047111

>>12047107
This post is literal nonsense, I know you're speaking figuratively, but as a human, I'm a general intelligence, so I do understand. You really are a sad-sack

>> No.12047118

>>12047098
>Except we did NOT show this, you simply stating this does not make it so, and you coping with this doesn't matter.
No, your own definition destroyed you. Let me cite it again:
>A computation is any type of calculation[1][2] that includes both arithmetical and non-arithmetical steps and which follows a well-defined model (e.g. an algorithm).
>In mathematics and computer science, an algorithm (/ˈælɡərJðəm/ (About this soundlisten)) is a finite sequence of well-defined, computer-implementable instructions, typically to solve a class of problems or to perform a computation.[1][2]

>The brain, objectively and irrefutably, performs more relevant computation for general intelligence than every single computer on the internet and does so for less energy than a lightbulb.
No, that's an absurd claim which also makes cup of water a genous and which contradicts your own definition.
>It flings around more electrons more easily and uses them to process information, store memory, make decisions, have thoughts, etc., than every single computer on the planet and does so for essentially no energy at all.
If flinging electrons is computation, then computer and water glass are even better.
>GPT-3 does not even come remotely close to this and you crying and shitting your diapers about it won't change this fact.
If flings electrons nicely.

>No I'm not. I'm not saying anything that isn't accepted in the field.
You are posting magic.
>And wrong.
It is correct though.
>I never claimed to be a computer engineer or neuroscientist. I'm in a master's program for mathematical physics. This is also irrelevant because what I'm saying is still in line with what is accepted in the fields.
Then stop to attribute nonsense to them.

>> No.12047122

>>12047118
Solving is a human concept, without a human to label something as solved, it isn't solved. Until a computer has the same intelligence as a human, it doesn't know it has solved anything, all it has done is perform a function. Learn the fucking difference, it isn't hard to understand.

>> No.12047123

>>12047111
No, anon, you are a philosophical zombie while GPT-3 is not.

>> No.12047126

>>12047123
GPT-3 has no consciousness, and cannot philosophize, you smooth-brain.

>> No.12047130

>>12047118
>No, your own definition destroyed you. Let me cite it again:
THATS NOT MY DEFINITION YOU FUCKING SHIT FOR BRAINS RETARD. How many times do I have to say there are multiple people here arguing with you?
>No, that's an absurd claim which also makes cup of water a genous and which contradicts your own definition.
No, it's not an absurd claim it's literally a fact that's known in the field.
The brain performs something like 10^28 operations per second. The entire internet is something like 10^22. And the brain uses significantly less energy.

The rest of your post is cope pseud nonsense.

>> No.12047131

>>12047122
>Solving is a human concept, without a human to label something as solved, it isn't solved.
Then, using your religious definition, computers don't solve anything by definition. That's OK, they will continue to dominate you even if you don't believe that they have soul.
>Until a computer has the same intelligence as a human, it doesn't know it has solved anything, all it has done is perform a function. Learn the fucking difference, it isn't hard to understand.
What do you mean by "knowing to solve something"?

>> No.12047135

>>12047126
It does and it can. Meanwhile you have no consciousness and even if you do some faulty philosophy, you don't have qualia-eating soul to experience it.

>> No.12047138

>>12047131
We use computers as tools to solve problems, you idiot. It isn't religious, I don't think you even understand what that term means, it's a basic fact.
>What do you mean by "knowing to solve something"?
The act of understanding, idiot.

>>12047135
>It does and it can.
You literally think GPT-3 is an AGI. Lmao, now I know you are a certain idiot.

>> No.12047140

>>12047130
>THATS NOT MY DEFINITION YOU FUCKING SHIT FOR BRAINS RETARD. How many times do I have to say there are multiple people here arguing with you?
Ah, so you disagree with it? Great, then argue with that other person, not me.
>The brain performs something like 10^28 operations per second. The entire internet is something like 10^22. And the brain uses significantly less energy.
No, that's total nonsense no matter how you define operations. If it's flops, then computer is doing much better. If it's anything, then internet (or glass of water) are still better.
>The rest of your post is cope pseud nonsense.
Your posts are absurd seething.

>> No.12047145

>>12047140
A glass of water has no INTENT, you FUCKING SIMPLETON.

>> No.12047146

>>12047138
>We use computers as tools to solve problems, you idiot. It isn't religious, I don't think you even understand what that term means, it's a basic fact.
Humans use other humans as tools since the beginning of the history. So what?
>The act of understanding, idiot.
This definition still has nothing about computers or humans, dummy.

>> No.12047149

>>12047138
>You literally think GPT-3 is an AGI.
You are not an AGI or NGI too. In fact you are hardly intelligence at all.

>> No.12047152

>>12047146
>Humans use other humans as tools since the beginning of the history. So what?
Everything the other anon has said about you is true, you just keep spouting more and more nonsense. Humans have general intelligence, by definition.
>This definition still has nothing about computers or humans, dummy.
GPT-3 is literally NOT a general intelligence, it is task-specific.

>> No.12047155

>>12047145
Glass of WATER has no SOUL which produces INTENT, eats QUALIA and shits FREE WILL? Anon, we are getting kinda philosophical here, aren't we?

>> No.12047162

>>12047155
God you're so stupid, I've said nothing about a soul, it's about emergent properties.

>> No.12047164

>>12047140
>No, that's total nonsense no matter how you define operations. If it's flops, then computer is doing much better. If it's anything, then internet (or glass of water) are still better.
No, it absolutely isn't and this is where you are a retard.
The brain is performing an order an magnitude more operations per second than the entire internet and does so for almost no energy at all, this is literally known and is accepted by everyone in the fields. I have friends who, unlike you, are actually in the fields and would laugh at the shit you're saying right now.
Stop being a fucking retard.

>> No.12047165

>>12047152
>Everything the other anon has said about you is true, you just keep spouting more and more nonsense. Humans have general intelligence, by definition.
Any humans? Do you have it? By the way that general intelligence is already vastly inferior in many areas to computers. So?
>GPT-3 is literally NOT a general intelligence, it is task-specific.
You are task-specific too.

>> No.12047167

>>12047130
Not him but I don't know where you get that "10^28 operations per second" figure you keep repeating.
Neurons in the human brain behave very differently from transistors.
It would serve your point if you told us about the metric you're using to get to that 10^28 operations.

Computer are known to perform task impossible for us human.
Maybe if we rewired the neurons of a human brain we could get comparable results. And maybe the inverse is possible, that is, we could rewire or reprogram a computer to emulate our intelligence.

>> No.12047171

>>12047165
>Literally ad homs, not reasoning.
Alright, what is your FULL argument? What is the point you're trying to prove throughout the entire thread?

>> No.12047176

>>12047162
Then no one forbids computers to have these emergent properties too. Unless you believe that your magic grey meat produces the shiny soul.

>> No.12047179

>>12047164
>The brain is performing an order an magnitude more operations per second than the entire internet and does so for almost no energy at all, this is literally known and is accepted by everyone in the fields.
No, that's pure trash, both with your definition and with any other. Otherwise either computer does more operations, or the glass of water does.
> I have friends who, unlike you, are actually in the fields and would laugh at the shit you're saying right now.
I've just asked your friends and they said that they lied to you.

>> No.12047180

>>12047176
I never said it didn't, but we aren't near that now, and I think you're using the wrong materials to try and achieve it.
See: >>12047171

>> No.12047187

>>12047171
That op argues with imaginary strawmen?

>> No.12047192

>>12047167
There is literally nothing a computer can do that a human brain can't do or vice versa.
The point is that the brain, for now, is significantly more powerful than the Internet and does so for essentially no energy. To compare GPT3 to a brain displays a level of retardation that belongs in >>>/x/

>> No.12047194

>>12047180
Brain has an awful clock rate, like several neuron firing in a second. Brain is already inferior to computers in the multitude of tasks. It just has great parallelism, but computers can be vastly parallel too. So your claim isn't based on anything.

>> No.12047197

>>12047179
>No, that's pure trash, both with your definition and with any other. Otherwise either computer does more operations, or the glass of water does.
No it isnt. It is not pure trash. You crying about this does not change the fact that the computational material that the brain is made out of is significantly more efficient and powerful than modern computers. You crying and shitting your diapers about this fact will not change it.

>> No.12047199

>>12047187
It's been claimed several times, in this thread alone, that GPT-3 is an AGI. Then, if you go and look at the archive, you'll see even more. OP isn't arguing with strawmen, but I think you might be if you think that OP is.

>> No.12047202

>>12047192
>There is literally nothing a computer can do that a human brain can't do or vice versa.
That literally means that we have AGI now and it is wrong in both directions.
>The point is that the brain, for now, is significantly more powerful than the Internet and does so for essentially no energy.
Now that's nonsense you keep repeating.

>> No.12047204

>>12047199
>It's been claimed several times, in this thread alone, that GPT-3 is an AGI.
Where, cite me. No, not OP brain, but this thread.

>> No.12047206

>>12047192
I'm actually very skeptical of GPT-3 being anything but a marketing stunt.
But I'm also skeptical of people who throw figures around without sourcing them.
Where do you get those 10^28 operations per second?

>> No.12047210

>>12047197
>No it isnt. It is not pure trash. You crying about this does not change the fact that the computational material that the brain is made out of is significantly more efficient and powerful than modern computers.
No, that's still trash which fails both common sense and your own definitions.
>You crying and shitting your diapers about this fact will not change it.
Oh, so you are a diaper fetishist? That explains a lot.

>> No.12047214

>>12047194
But the brain is better than silicon at general intelligence, for now. Being task-specific is great, but it means you're just a useful tool, not the "overlord AI", that /sci/ memes about.

>> No.12047218

>>12047204
>>>12047126: GPT-3 has no consciousness, and cannot philosophize, you smooth-brain.
>>>12047135: It does and it can.

>> No.12047225

>>12047210
>No, that's still trash which fails both common sense and your own definitions.
Can an AI catch a ball without being trained to catch it? A human can. Why? Because a human subconsciously performs complex calculations, and puts the arm and hand into the correct position to catch it.

>> No.12047227

>>12047214
Brain STILL does some tasks better than humans. It's just that the list of the tasks AI manages to solve better than humans grows very rapidly. So
>GPT-3 is AGI
is still false
>GPT-3 hints that AGI is near
is true.

>> No.12047234

>>12047218
Having consciousness is a philosophy and religion problem, but it is unrelated to having the general intelligence.

>> No.12047235

>>12047227
>Brain STILL does some tasks better than computer.
The vast majority of tasks, yes. So far, we have to train an AI to do X task, we don't have to do that for humans, unless we wan't to improve their performance.

>> No.12047237

>>12047202
>That literally means that we have AGI now and it is wrong in both directions.
You don't even understand the notion of turing completeness holy fucking shit.
>Now that's nonsense you keep repeating.
Absolute cope

>> No.12047238

>>12047225
>Can an AI catch a ball without being trained to catch it? A human can.
But he literally can't. Throw a ball at newborn and he will not catch it. Throw a ball at feral child and he will not catch it. Even someone who never played sports can fail at this task. So no, human can't.

>> No.12047239

>>12047234
The act of philosophizing requires general intelligence. Therefore, if you can philosophize you are claiming it is a general intelligence.

>> No.12047241

>>12047235
>The vast majority of tasks, yes.
That depends on how you define the list. For example your computer can do a vast amount of things you can't do or will do very slowly.
>So far, we have to train an AI to do X task, we don't have to do that for humans
We absolutely do, what do you think education is?

>> No.12047242

>>12047238
Newborns don't count, because they aren't even fully mature humans, idiot.
Also, a feral child probably would, considering that animals can play catching games.
>https://www.youtube.com/watch?v=gJEkSGnZiIA
An AI couldn't do that without being trained.

>> No.12047243

>>12047210
>No, that's still trash which fails both common sense and your own definitions.
I was off, the brain performs 36 * 10^26 operations per second.
https://blogs.scientificamerican.com/news-blog/computers-have-a-lot-to-learn-from-2009-03-10
>Oh, so you are a diaper fetishist? That explains a lot.
More cope

>> No.12047247

>>12047239
>The act of philosophizing requires general intelligence.
That's an incredibly dubious claim.

>> No.12047249

>>12047241
>That depends on how you define the list.
Only if you're being disingenuous. If this were true, we'd have already replaced nearly everyone with AI, and we haven't.
>We absolutely do, what do you think education is?
Bruh, that's literally training to improve performance, not training to even do.

>> No.12047251

>>12047237
>You don't even understand the notion of turing completeness holy fucking shit.
Wow, anon you introduced just another concept you don't understand at all! What's next, Godel's theorem or Schrodinger's cat?
>Absolute cope
Seethe more, anon.

>> No.12047252

>>12047247
Only if you're being disingenuous, which AI cultists seem to be.

>> No.12047257

>>12047251
>Wow, anon you introduced just another concept you don't understand at all! What's next, Godel's theorem or Schrodinger's cat?
You literally don't know what you're talking about and don't know what I know about mathematics.
There is nothing that a brain can do that a computer can't or vice versa.
>Seethe more, anon.
I am far more intelligent than you.

>> No.12047258

>>12047243
>I was off, the brain performs 36 * 10^26 operations per second.
No, your article doesn't say so. Learn to read, anon, that will help you.
>More cope
Seethe more anon, that will be helpful to your fetish.

>> No.12047260

>>12047258
>No, your article doesn't say so. Learn to read, anon, that will help you.
False and cope

>> No.12047262

>>12047249
>Only if you're being disingenuous. If this were true, we'd have already replaced nearly everyone with AI, and we haven't.
That's because we don't have AGI yet, but we already replace a huge list of tasks with computers.
>Bruh, that's literally training to improve performance, not training to even do.
You literally can't do many things before education at all. An illiterate man does not do science two or ten times slower.

>> No.12047267

>>12047260
>False and cope
Seethe, anon. If you will cite the place in the article which deceived you, you will realize that it will return us to the genius glass argument.

>> No.12047270

>>12047267
Projection and denial.

>> No.12047271

>>12047262
>That's because we don't have AGI yet, but we already replace a huge list of tasks with computers.
That's true, but humans are still better at more, and have greater versatility at more, that's one of the key concepts of general intelligence.
>You literally can't do many things before education at all. An illiterate man does not do science two or ten times slower.
Good thing that humans can learn to read without formal schooling:
>https://www.psychologytoday.com/us/blog/freedom-learn/201002/children-teach-themselves-read
An AI can't.

>> No.12047272

>>12047257
>You literally don't know what you're talking about and don't know what I know about mathematics.
No, you spout some terms you saw in popsci videos.
>There is nothing that a brain can do that a computer can't or vice versa.
That's already false and it isn't related to Turing completeness in any way.
>I am far more intelligent than you.
You are hardly more intelligent that GPT-3.

>> No.12047275

>>12047270
Your own definitions betray you and articles you love fail you too, what an unhappy life.

>> No.12047276

>>12047272
>No, you spout some terms you saw in popsci videos.
False
>That's already false and it isn't related to Turing completeness in any way.
False
>You are hardly more intelligent that GPT-3.
I am far more intelligent than you.

>> No.12047278

>>12047275
Seethe and denial.

>> No.12047279

>>12047271
>That's true, but humans are still better at more, and have greater versatility at more, that's one of the key concepts of general intelligence.
That "more" does not have much meaning since you don't have a list of tasks to compare.
>Good thing that humans can learn to read without formal schooling:
Wow, these children are feral? They don't have parents or caregivers? If yes, then they are truly gods.

>> No.12047285

>>12047278
Cope, anon.

>> No.12047286

>>12047276
>False
True
>False
True
>I am far more intelligent than you.
No, you are dumber than me, GPT-3 and GPT-2.

>> No.12047288

>>12047279
>That "more" does not have much meaning since you don't have a list of tasks to compare.
Imagine actually asking for a list, as if that is a requirement. Look around you. AI are only used for specific and "bounded" tasks, humans do everything else.
>Wow, these children are feral? They don't have parents or caregivers? If yes, then they are truly gods.
That's a stupid argument, I mean, you couldn't even have a "feral AI" for comparison? Try to stay within the bounds of what is being asked.

>> No.12047293

>>12047242
>Newborns don't count, because they aren't even fully mature humans, idiot.
And how do their brains develop? Right, by interacting with their parents.
>Also, a feral child probably would, considering that animals can play catching games.
No, feral child is an invalid (and animals fail without parents too, see https://en.wikipedia.org/wiki/Pit_of_despair))
>More cope
More seething, my diaper fetishist friend?

>> No.12047295

>>12047243
>https://blogs.scientificamerican.com/news-blog/computers-have-a-lot-to-learn-from-2009-03-10
>A computer comparable to the human brain, he added, would need to be able to perform more than 38 thousand trillion operations per second
38 thousand trillion is 38 * 1000 * 10^12, which is 3.8*10^16, not 3.6*10^26.

Which is close to wikipedia's take
>https://en.wikipedia.org/wiki/Computer_performance_by_orders_of_magnitude#Petascale_computing_(1015)
ie, 20*10^20

Granted, the metric used to get to this number is questionable, but what you reported is wrong.
And the argument that the human brain is less powerful than the combined effort of the computers behind the internet is preposterous.

>> No.12047297

>>12047286
Dumbfuck, you literally can't even differentiate between the writing styles of the 3 different anon you're arguing with and you continue to posit that I've made claims or given definitions that another anon has.
You are a coping moron who thinks AI is more sophisticated than it actually is, and you deny the consensus in the field.
You're a coping retard.

>> No.12047301

>>12047293
>And how do their brains develop? Right, by interacting with their parents.
No, their "environment". See phenotype.
>No, feral child is an invalid (and animals fail without parents too, see https://en.wikipedia.org/wiki/Pit_of_despair))
That's because they're missing an environmental factor that we evolved with, just like AI "evolved" with operators. Without an operator to set things like training schema, it wouldn't even learn its specific task.

>> No.12047304

>>12047295
>ie, 20*10^20
sorry it is 20*10^15

>> No.12047306

>>12047288
>Imagine actually asking for a list, as if that is a requirement. Look around you. AI are only used for specific and "bounded" tasks, humans do everything else.
I see that humans are already unable to do most of the things their computer can (including the simplest ones), and if they can, they do it way slower. That alone puts your "more" claim in a doubt.
>That's a stupid argument, I mean, you couldn't even have a "feral AI" for comparison? Try to stay within the bounds of what is being asked.
If we are comparing trained children with trained AI, let's stick to the topic. If we are comparing untrained children to untrained AI, then they are both dumb.
Humans are also unable to become grandmasters (or even good players) in chess if they are given only the rules. Alphazero successfully managed it. So computer is actually better at training itself than humans.

>> No.12047312

>>12047297
>Dumbfuck, you literally can't even differentiate between the writing styles of the 3 different anon you're arguing with and you continue to posit that I've made claims or given definitions that another anon has.
If you think that your nonsense is better than another nonsense, then chat with that anon, not me.
>You are a coping moron who thinks AI is more sophisticated than it actually is, and you deny the consensus in the field.
I'm not the strawman OP fights again, so seethe my friend.
>You're a coping retard.
Still seething, dummy?

>> No.12047313

>>12047306
>I see that humans are already unable to do most of the things their computer can (including the simplest ones), and if they can, they do it way slower.
Only at tasks that computers are good at. So not really a fair comparison, when we're looking to take AI past "computer expectation".
>Humans are also unable to become grandmasters (or even good players) in chess if they are given only the rules.
1) Humans derived and invented those rules.
2) Why are there humans grandmasters then?
>Alphazero successfully managed it.
Task-specific, and trained.
>So computer is actually better at training itself than humans.
At specific tasks, once given a defined schema.

>> No.12047319

>>12047295
The main trouble with his argument is that
>A computer comparable to the human brain, he added, would need to be able to perform more than 38 thousand trillion operations per second
even if true does not claim that brain performs these operations. To fully simulate the glass of water on a computer you need insane number of operations per second. Yet glass does not perform computations and so does the brain.

But yes, it is also way lower than he claims.

>> No.12047320

>>12047312
You are the only one posting nonsense.
>I'm not the strawman OP fights again, so seethe my friend.
This is gibberish. If you don't want to come across like an idiot, learn to write.

>> No.12047344

>>12047313
>Only at tasks that computers are good at. So not really a fair comparison, when we're looking to take AI past "computer expectation".
But computers were worse at these tasks century, decades or even a few years ago. We are talking about the general trend here, not OP "AGI is here" strawman.

>Humans derived and invented those rules.
So what? Universe around humans existed before them, so they invented nothing?
>Why are there humans grandmasters then?
The whole history of chess-playing humanity (and these grandmasters will still always lose to even weaker computers).

>Task-specific, and trained.
No. Alphazero was initially created to play Go, it had nothing about chess inside it. Also it was not "trained", but just played with itself a lot.
>At specific tasks, once given a defined schema.
These tasks were considered a pinnacle of human mind years ago, which returns us to >>12046732 pic.

>> No.12047349

>>12046638
But it is a general intelligence. It can play chess, play go, tell story, become expert in science of your choice, program code, do philosophy, etc, it can do just about anything. It does it well enough to fool casual people, however not specialized experts in those fields.

GPT-3 is by definition a general AI. There's nothing specialized about GPT-3.

>> No.12047351

>>12047320
>You are the only one posting nonsense.
No, it's you with your self-contradictions and silly claims.
>This is gibberish. If you don't want to come across like an idiot, learn to write.
You started it in >>12046703 post.

>> No.12047357

>>12047349
I believe that it will turn into AGI in a decade or less, but can you replace any human with it now?

>> No.12047393

>>12047349
It can't play chess.
I read somewhere that someone trained it with millions of plays and it couldn't go further than 19 moves (at most) without making an illegal move.
What I suspect GPT-3 did is that it played moves that matched a play in his database, and then played something invalid because his opponent did something not found in any of those recorded plays.

You could argue GPT-3 still demonstrates intelligence, but it's more Chinese room type of intelligence than human-like intelligence.

>> No.12047471

>>12047393
>Chinese room type of intelligence
Chinese room is a flawed assumptions, but even then thats how the brain works when dealing with recalling patterns and utilizing them.

>> No.12047485

>>12047471
>thats how the brain works when dealing with recalling patterns and utilizing them.
No it isnt.

>> No.12047493

Why don’t they combine all models together so it can switch to the best one for the problem and it can be an “agi”

>> No.12047500

>>12047393
Chinese room is about soul and qualia, not about abilities.

>> No.12047507

>>12047349
Make it do an IQ test, I wanna see how near to g it can get.

>> No.12047526

>>12047349
>>12047507
Also, you're not allowed to train it, since humans aren't meant to train for IQ tests. They're simply meant to be able to take them.

>> No.12047557

>>12047526
>since humans aren't meant to train for IQ tests. They're simply meant to be able to take them.
Then you should apply IQ tests to newborns or feral children.

>> No.12047613

>>12047557
This is done already

>> No.12047670

>>12047557
They can do them, an AI wouldn't, it would outright fail. Admittedly, not very well, but they can do them.

>> No.12047673

>>12047613
Not it isn't, atleast not the same way. Humans/animals learn with vertical pathways. We build upon previous experiences. Meanwhile GPT3 is built upon horizontal pathways. It leans via knowledge of everything(sorta small 175B parameters tho) we have today.

>> No.12047723

Stop responding to the bait jesus christ.

>> No.12047727

>>12047723
What bait?

>> No.12047780

>>12047727
>>12047723

>> No.12048905

This "accksshually it's not intelligence" stuff is annoying. We know that it doesn't fit that definition, we are using another definition of intelligence. You are not actually disagreeing with the people who say its intelligence on anything actually physical but on the subjective definition of a word. You know perfectly well what people mean by intelligence when they call these things intelligence. There, thread over, meaningless word games, fuck off

just realized OP is actually saying it's not AGI, well yeah ANI is not AGI... i'm still gonna post this for the not-intelligence faggots though