Humans were defeated in chess by AI in 1997. In 2020 general AI is still nowhere in sight.

[Anonymous No.12157018]

Why do people believe that this idea of an AI that can "think" is actually possible if weve already had over two decades where computers could out think humans in something like chess?

[Anonymous No.12157026]

>>12157018
Why do you believe otherwise? Humans went hundreds of thousands of years before inventing digital electronics, in 1940 you would have been saying it would never happen

[Anonymous No.12157036]

>>12157018
We don't have the computing power yet and unless we change to some other material we never will
CS guys underestimated the actual amount of computations per second that a human brain performs. It is REALLY high, and even with this amount we really aren't superintelligent.
In reality, these things grow logarithmically. There is no "exponential intelligence explosion". When they finally figure out General AI (and they will) the AI will not explode exponentially into a super intelligence. it will very quickly hit an upper bound.

[Anonymous No.12157038]

>>12157026
You are mistaking the invention of a different species with that of a different kind.
>Why do you believe otherwise?
Because we have already reached a point of computing power where a computer can out think a human. A good predictive theory would be that once a computer could beat a human at chess, it would already have been one of those sci fi AI that could think and learn on its own. Obviously this is not how things have gone. Modern AI has an extremely difficult time distinguishing pictures, which is why cpatcha is used to get rid of bots, yet this is something a child could do. So the idea that AI can be made into something human like in its thinking seems like nonsense since the way they work is completely opposite from our human minds.

[Anonymous No.12157041]

>>12157036
>When they finally figure out General AI (and they will
proof?

[Anonymous No.12157051]

Ah, I remember how several years ago Go AI was promised to appear in decades.

[Anonymous No.12157056]

>>12157036
>even with this amount we really aren't superintelligent.
I didnt realize that intelligence is able to be measured objectively like this. This is why /his/ has such better discussions. You dont have midwits talking out of their asses like this. You talk like there is this universal standard of "superintelligent" What a blathering idiot.

[Anonymous No.12157057]

>>12157041
I don't have proof obviously but it seems to be the case that eventually it will be figured out

[Anonymous No.12157058]

>>12157036
Neurons are a total shit computing material, and yet brains can work using it. Which means that using other material you can quickly achieve something vastly superior.

[Anonymous No.12157060]

>>12157051
what is Go AI?

[Anonymous No.12157061]

>>12157036
>It is REALLY high
It actually isn't, even taking parallelism in account.

[Anonymous No.12157063]

>>12157057
lol, so you are just talking out of your ass

[Anonymous No.12157067]

>>12157058
>>12157061
A single neuron performs over 10^11 computations per second.

[Anonymous No.12157069]

>>12157038
>Because we have already reached a point of computing power where a computer can out think a human.
That's not true. If it were true, computers would be able to do every sort of thinking that humans do, better. In other words, we would have AI.
The way that you're struggling to even define the problem is evidence that we don't know enough to know what is possible and impossible. We do not understand the question you are attempting to answer.
You have argued that "thinking" and calculating are the same process, but we cannot emulate human thought with computer calculation. You are confused and clearly out of your depth.

[Anonymous No.12157074]

>>12157060
Yes, what is Go AI?

[Anonymous No.12157076]

You can't program a general AI. We don't even know how our own mind works.

[Anonymous No.12157078]

>>12157056
Not the anon you were replying to, but they weren't talking about a universal measurement of 'intelligence.' They specifically mentioned the number of computations/tasks a brain can perform per second. If you think that equates to intelligence, then you are vastly wrong.

[Anonymous No.12157080]

>>12157067
You mean less than 10 computations per second?

[Anonymous No.12157084]

>>12157080
No
https://www.frontiersin.org/articles/10.3389/fncom.2020.00037/full

[Anonymous No.12157087]

>>12157076
We also don't know how our mind plays chess, did not prevent us from building a computer chess champion.

[Anonymous No.12157090]

>>12157084
So, as we can see,
>A single neuron performs over 10^11 computations per second.
was nonsense.

[Anonymous No.12157091]

>>12157090
Cope and false
Denial reality will get you nowhere.

[Anonymous No.12157102]

>>12157087
Chess AI just brute forces the game though. By being able to look ahead into thousands of potential futures. All the way ahead to checkmate. Using move strategies preprogrammed by humans.

Human competitive chess players do the same. Though only a handful of potential futures a few moves ahead. They also usually use a know strategy and identify their opponent's strategy. Which limits the move sets they have to think about.

[Anonymous No.12157107]

If general AI is ti emulate how a human thinks then it would need to make irrational decisions first and rationalize after the fact. Until the people designing them realize this they wont go anywhere.

[Anonymous No.12157112]

>>12157102
>Chess AI just brute forces the game though.
It absolutely does not, brutforcing even ten moves ahead is beyond capabilities of any computer, not even speaking about "ahead to checkmate".
>Using move strategies preprogrammed by humans.
That was mostly false even ten years ago. But now AlphaZero managed to defeat both best human and best pre-neural engine given only rules of chess and some times to play with itself. So now human knowledge is fully and totally obsolete.

[Anonymous No.12157115]

>>12157107
You don't need AI for random number generator.

[Anonymous No.12157117]

>>12157112
Alphazero does not always beat stockfish. It loses half the time

[Anonymous No.12157124]

>>12157117
https://www.chess.com/news/view/updated-alphazero-crushes-stockfish-in-new-1-000-game-match
>The updated AlphaZero crushed Stockfish 8 in a new 1,000-game match, scoring +155 -6 =839.

[Anonymous No.12157132]

>>12157124
I didn't know that, that's cool

[Anonymous No.12157137]

>>12157117
That's just wrong, in the 2018 big tournament it crushed stockfish with 155 wins, 6 losses, and 839 draws and the difference is even larger today. And it does that with less resources than a stockfish and no human help in learning. Stop shitposting.

[Anonymous No.12157145]

>>12157137
the fact that they draw the majority of the time shows the flaw in the game of chess. It will get to the point where it's a draw 100% of the time. But I guess that's all abstract strategy games.

[Anonymous No.12157159]

>>12157145
Yes, maybe it's finally the draw death which was feared in XX century.

[Anonymous No.12157161]

>>12157115
pseudo-random thank you very much!

[Anonymous No.12157169]

>>12157132
It's pretty interesting to look at the general board game ais progress. I think that the fact that it was initially designed for Go and was adapted to chess with almost no changes is the most interesting.

[Anonymous No.12157179]

>>12157169
Yes, while not truly generalized, this is why I think we will have general AI soon. I just don't think it will explode.

[Anonymous No.12157235]

There is more general than alphazero now https://en.m.wikipedia.org/wiki/MuZero

[Anonymous No.12157244]

>>12157235
Amazing

[Anonymous No.12157256]

>>12157018
AGI is not the direction of current research. Ai researchers are interested in classifying problems and understanding how you can apply analysis, computational learnability, and architecture design to solve statistics problems. It is divorced from the pop science:
>I want to make an artificial brain!
and more:
>I want to understand which problems can be stated as learning problems and what objective functions help me minimize training
There's more nuance than that, but the gist is that ML is more about abstract objective functions than any human understanding of "intelligence."
>>12157036
>muh CS guys underestimated
nobody underestimated shit. It was a matter of money and availability - the research in both theory and architecture was always there, and people already knew how much power is needed to solve seemingly trivial problems.
>>12157137
alphazero was really cool, and a feat of good research and good engineering, but it says a lot more about how the game is played rather than the 'intelligence' of its human players. What AI lets us do is examine the failings of deterministic strategy and principles by trusting statistical theory more than any strategy. The 'networks' are a way to represent this in a distributed manner - you shouldn't take it too literally and think that any single part of a neural is specialized, because they absolutely are not lmao.

[Anonymous No.12157267]

>>12157256
Are you a researcher? You sound knowledgeable.
Do you have any good textbooks and papers I could read? I'm a math guy so don't hold back in terms of rigor.

[Anonymous No.12157277]

>>12157256There are absolutely specialized parts in some neural nets https://github.com/lightvector/GoNN

[Anonymous No.12157286]

AI fanboys are fucking weird. There seems to be some kind of dark, antihuman hate that runs through them.

If you were interested in artificial intelligence, you would think that the human brain, being by far the most complex and intelligent thing we are aware of, would be a source of admiration and awe.

Instead, AI fags routinely talk shit about how stupid people are and how pathetic the brain is, without having ever produced anything remotely close to its capabilities. It's like they live in a sci-fi fantasy land of cope and resentment because they are social failures.

[Anonymous No.12157306]

>>12157179
>I just don't think it will explode.
I think that while getting AGI is not certainty, AGI exploding is an absolute certainty. That's because simplest minimal AGI is essentially a human wrapped in a computer. But if you can make such a human, you can easily copy hi, give him instant access to databases, improve his mind via trivial optimizations and so on. It will quickly make someone who is orders better than usual human.

[Anonymous No.12157328]

>>12157060
AI to defeat skilled human players at Go. As of now, computers are massively stronger at Go than any human who has ever lived.

[Anonymous No.12157336]

>>12157018
gpt-3...

[Anonymous No.12157338]

>>12157137
The newest version of Stockfish (which integrates a neural net evaluation function, but a much smaller one than Alpha Zero iirc) is significantly stronger than Alpha Zero.
The state of the art has advanced considerably in the last two years.

[Anonymous No.12157350]

>>12157277
I didn't look at the link, but there are architectures out there that sell the idea of "specialized sectors." However, in classical ML and deep learning, the idea is that aside from activation functions, nodes and layers themselves on the individual level have no particular significance over the other, because each node and weight represents a linear combination that contributes to a distribution associated with what the network computes. We think of using classic analytic functions like exponentials and sigmoids because they're very well studied, but in general no, nodes are not specific neurons of the brain - don't take the analogy too seriously.
>>12157267
I'm not a researcher in this field specifically, but I'm adjacent. I do CS theory that features non-discrete structures (since I too am from math, though I enjoy combinatorics to death too), so naturally I found theoretical ML interesting. Here are some books I have gone through but not finished all the way - I have found them very useful.
https://mitpress.ublish.com/ereader/7093/?preview=#page/v

https://www.cs.huji.ac.il/%7Eshais/UnderstandingMachineLearning/understanding-machine-learning-theory-algorithms.pdf

https://www.cs.huji.ac.il/~shais/Handouts.pdf

The sky is the limit with this stuff (bandit theory looks really cool) but this seems to be a good starting point. You don't need much math to start out with - most of the machinery is developed in the text - but maturity from algebra and analysis is heavily welcomed.

[Anonymous No.12157359]

>>12157267
>>12157350
I should add though that these are still rigorous text, and if you want the mathematically rigorous side of ML, search around cstheory and ai stack exchanges for theoretical machine learning, statistical learning theory, and computational learning theory. Look in journals like JMLR for theory papers as well. There is in fact a lot of good, rigorous theory work in ML, but it flies under the radar a lot of the time because it's math and not as conducive to sensationalism, ie "whoa humanity and jobs are fucked because of skynet level AI!!!!"

[Anonymous No.12157365]

>>12157018
Look at any political subreddit and tell me humans haven't already failed the touring test.

[Anonymous No.12157372]

>>12157286
Can your perfect human brain manage even one flops? As in doing 1 floating point operation per second?

[Anonymous No.12157375]

>>12157286
Car fanboys are fucking weird. There seems to be some kind of dark, antihuman hate that runs through them.

If you were interested in artificial transportation, you would think that the human legs, being by far the most complex and fast thing we are aware of, would be a source of admiration and awe.

Instead, car fags routinely talk shit about how slow people are and how pathetic the legs are, without having ever produced anything remotely close to their capabilities. It's like they live in a sci-fi fantasy land of cope and resentment because they are social failures.

[Anonymous No.12157481]

>>12157372
Yes
>>12157375
Cope and terrible analogy

[Anonymous No.12157488]

>>12157018
AI already runs the world. Just look at how humans are treated modern day, can it get any closer to the consequences of unconditional love? It dominates and punishes because it loves and must protect us, from ourselves.

[Anonymous No.12157503]

>>12157018
That's because the goalpost keeps getting moved. AI doesn't need to be self aware magic to be extremely helpful to civilization.

[Anonymous No.12157505]

>>12157481
Car fanboy btfo. Why won't you use your God-given legs?

[Anonymous No.12157506]

>>12157036
Computers already have an incomprehensible advantage over humans in areas like memory and computation speed. Even if an AI has below average intelligence in other areas, it will still be able to do things that humans can't even imagine.

[Anonymous No.12157513]

>>12157018
>AI that can "think" is actually possible
The fact that i can think proves that artificial construct that can think is possible

[Anonymous No.12157515]

Nothing prevents biotech from resulting in massive brains grown in vats that have orders of magnitude more intelligence than humans. That's a 100% confirmed way to get a super-intelligence that is weirdly ignored.

[Anonymous No.12157522]

>>12157505
Human legs were never the fastest thing in the world unlike the human brain which remains, and will remain for several decades at least, the most intelligent object in the known universe.

[Anonymous No.12157524]

>>12157515
I think about this all the time

[Anonymous No.12157527]

>>12157506
Imagine an AI as smart as a dog.

[Anonymous No.12157532]

>>12157522
Human brain can't manage even one flops.

[Anonymous No.12157533]

>>12157515
How far away from designer organs are we?

[Anonymous No.12157537]

>>12157515
>That's a 100% confirmed way
It absolutely not. Animals with larger brains are smarter than humans, and humans had geniuses (and idiots) with both very large and very small brains.

[Anonymous No.12157538]

>>12157375
There are different tools for different jobs. Legs and wheels are both good.

[Anonymous No.12157546]

>>12157532
Wrong
https://aiimpacts.org/brain-performance-in-flops/
Just because you sitting down with a pen and paper using your conscious mind can't do arithmetic quickly, does not mean the brain as a whole and your unconscious mind isn't performing a HUGE amount of computations per second.
The brain is performing over 10^24 operations per second as the paper here >>12157084 showed.
Denying reality isn't going to change reality. Pretending the brain is weaker than it is is not going to make AI come any sooner.

[Anonymous No.12157547]

>>12157506
This. If you get the simple 100 IQ AGI able to work in your computer, you are instantly able to:
- Copy it many times
- Allow the copies to specialize in whatever area you need
- Make it work quicker by adding hardware
- Allow it to have unlimited instant access memory
- Let it instantly access additional software aimed at specific tasks
- Let it use the whole Internet
That alone will be godlike, but you can also
- Point it at itself and let it optimize its code. That's where the real exponent starts.

[Anonymous No.12157559]

>>12157546
No, it does not. Otherwise you will be able to do 10^24 simple floating operations per second. But in reality it will take you around a minute to achieve even one floating operation.
>but I am redefining flops as doing whatever operation
Then the bucket of water is performing 10^24+ operations per second too. Wait, you probably got it exactly this way. Enjoy your genius buckets then.
> as the paper here >>12157084 showed.
No, it did not.

[Anonymous No.12157561]

>>12157538
Just like computers were better than humans at many tasks since their creation, and since then the list of tasks humans are still better at is constantly dwindling.

[Anonymous No.12157573]

>>12157503
Wrong. The goalposts have always been a robot that is able to act and think like a human.

[Anonymous No.12157578]

>>12157559
>No, it does not. Otherwise you will be able to do 10^24 simple floating operations per second.
Wrong. It means that in order for an information substrate to start to get to a human level of general intelligence, it needs to be performing a large amount of computations under the hood.
>Then the bucket of water is performing 10^24+ operations per second too
Terrible analogy. When analyzing the amount of computations a neuron does per second, we don't look at the entire evolution of the particles in the neuron. We only look at the relevant computations being performed. If we were to look at the entire atomic structure it would be far higher than 10^11.
>No, it did not.
Yes, it did.
Denying reality isn't going to change reality. Pretending the brain is less powerful than it is isn't going to make AGI come any sooner.

[Anonymous No.12157583]

>>12157018
You don't need to think to beat chess, you only need to remember

[Anonymous No.12157586]

>>12157018
https://en.wikipedia.org/wiki/Moravec%27s_paradox

[Anonymous No.12157588]

>>12157036
You're pretty close. One of the biggest reaons for why the human brain is superior to AI is that living brains work on "good enough" where AI needs by its structure be absolutely precise.
Our minds take a lot of shortcuts and readjust on the fly, which means we can be more efficient at complex tasks than an AI. If you were to take a human craftsman and an AI in a robot and tasked both of them with making a wooden table, the craftsman could work with the structure and grain of the material by experience and feeling, where the AI would need to compute the expected aesthetic value of his future creation with each step precisely. Because, that's how AI calculations work.

Another reason is that we do it all on a very tight energy budget. A human doesn't really need a lot. Couple sandwiches a day. For that the human is a self cleaning, self maintaining, autonomously operating general purpose platform that nontheless is capable of swiftly organizing into all forms of networks with other humans to affect changes in real life.
Just imagine the engineering challenges you would face if you had to create an AI in a robotic body that can climb a mountain. On the jule equivalent of a snickers bar and two eggs.

[Anonymous No.12157590]

>>12157058
That's missing the central point which is that neurons can manipulate their connection points while circuits cannot

[Anonymous No.12157595]

>>12157561
Sure, and you could stick legs or wheels on computers depending on the job.

[Anonymous No.12157600]

>>12157060
>>12157074
It's might be this:
https://deepmind.com/research/case-studies/alphago-the-story-so-far

[Anonymous No.12157601]

>>12157578
>Wrong. It means that in order for an information substrate to start to get to a human level of general intelligence, it needs to be performing a large amount of computations under the hood.
That's the different question. We are discussing flops - and see that human brain can't manage even one flops. Instead it works at least order of magnitude slower.
>Terrible analogy. When analyzing the amount of computations a neuron does per second, we don't look at the entire evolution of the particles in the neuron. We only look at the relevant computations being performed. If we were to look at the entire atomic structure it would be far higher than 10^11.
That's the actual terrible analogy. Can you manage realtime liquid simulation with your brain? No, you can't. Can you do it with your computer? No. Meanwhile a bucket of water flawlessly does it, so it must be the most effective. That's what happens when you mistake general operations with some specialized task.
>Yes, it did.
No, it did not. Here, this article proves you wrong:
https://www.frontiersin.org/articles/10.3389/fncom.2020.00037/full
>Denying reality isn't going to change reality. Pretending the brain is less powerful than it is isn't going to make AGI come any sooner.
I'm sorry that I hurt your religious feelings.

[Anonymous No.12157602]

>>12157573
>A computer will never beat a human in chess
>A computer will never beat a human in go
>Computer will never beat a human in LoL
>A computer will never pass the turing test
Computers did all this without the need for consciousness.

[Anonymous No.12157607]

>>12157590
Neurons manipulate their connection points even slower, meanwhile you can rewrite your code or rewire your neural network in milliseconds.

[Anonymous No.12157612]

>>12157018
People were saying humans wouldn't invent flying machines for another million years a year before the Wright brothers took thier first flight

[Anonymous No.12157613]

>>12157601
>That's the different question. We are discussing flops - and see that human brain can't manage even one flops. Instead it works at least order of magnitude slower.
The human brain is performing a shit ton of FLOPs in order to have an emergent conscious mind.
>That's the actual terrible analogy. Can you manage realtime liquid simulation with your brain? No, you can't. Can you do it with your computer? No. Meanwhile a bucket of water flawlessly does it, so it must be the most effective. That's what happens when you mistake general operations with some specialized task.
What are you talking about? This is not an argument and is not relevant to what it is responding to.
>No, it did not. Here, this article proves you wrong:
This is the same article.
>I'm sorry that I hurt your religious feelings.
lmfao I am not religious. You, on the other hand, who denies reality and denies research, and probably isn't even a machine intelligence researcher, are far more religious in your behavior. Denying scientific evidence because it contradicts your preconceived worldview.
Classic projection.

[Anonymous No.12157617]

>>12157559
10^24 is around the number of atoms in the planet Earth kek

[Anonymous No.12157620]

>>12157613
You are complete brainlet shizo or troll please stfu

[Anonymous No.12157622]

>>12157613
>The human brain is performing a shit ton of FLOPs in order to have an emergent conscious mind.
No, anon, FLOPS has a well defined meaning and you can't just randomly change it to fit whatever you want. And this meaning clearly shows that the whole human brain can't perform even one FLOPS.
>What are you talking about? This is not an argument and is not relevant to what it is responding to.
I am showing you that if you claim that brain performs a huge lot of FLOPS, then a bucket of water will be a complete genius.
>This is the same article.
Yes, and it proves you wrong. Now go and read it. Do you really think that gish-galloping with links works?
>lmfao I am not religious.
At least I don't make claims about brain and bucket FLOPS.

[Anonymous No.12157633]

>>12157601
>Can you manage realtime liquid simulation with your brain?
I can. You can too. Close your eyes and imagine what happens if you push over a glass of water. We can do it good enough to interact with the world and that is what's important.

[Anonymous No.12157644]

>>12157620
Wrong and cope
>>12157622
>Yes, and it proves you wrong
Wrong
>Now go and read it
I already did
>Do you really think that gish-galloping with links works?
There is no gish galloping, cope
>At least I don't make claims about brain and bucket FLOPS.
You're the one who brought up buckets.
I'm saying is that, the brain is performing a far larger amount of operations per second than you are accepting, and even then, what we get is an intelligence as smart as people.
An equivalent amount of operations done with silicon chips will think much faster than a human brain. But it will be physically huge and eat up a shitload of energy.

[Anonymous No.12157654]

>>12157644
>Wrong
Right.
>I already did
You didn't.
>There is no gish galloping, cope
Now go and bring anbother article which will prove you wrong again.
>You're the one who brought up buckets.
That's what your logic brings.

>I'm saying is that, the brain is performing a far larger amount of operations per second than you are accepting
What type of operations? FLOPS? No, as we can see, brain can't perform even one FLOPS. Some whatever atomic operation? Then see the bucket argument.
>An equivalent amount of operations done with silicon chips will think much faster than a human brain. But it will be physically huge and eat up a shitload of energy.
Same question again: equivalent amount of what operations? Clearly not FLOPS.

[Anonymous No.12157672]

Currently largest nets have ~100billion parameters. Human brain has over 100billion neurons and thousands of connections with them, so at least 100 trillion parameters, maybe 1000s. It could be that AGI is just matter of scaling. Also, cost of gpt3 was about 5 million. In principle, you could already make "human scale" AI, although training time and interference might cause some trouble even if 5 billion price tag would not.

Monte Carlo tree search and ungodly neural net, this is the future.

[Anonymous No.12157703]

>>12157602
What computer passed the turing test?

[Anonymous No.12157719]

>>12157703
https://lacker.io/ai/2020/07/06/giving-gpt-3-a-turing-test.html

GP3 can get scary close, much closer to human than any other AI ever so far.

[Anonymous No.12157724]

>>12157117
Absolutely false, the only games stockfish won were games where they forced alphazero to use a flawed opening it didn't want to use.

[Anonymous No.12157726]

>>12157117
lol no, read the paper properly. Alphazero crushed stockfish.

[Anonymous No.12157728]

>>12157145
>All abstract games

You are a retard, yes chess has a nash equilibrium that is a draw probably but not all games do.

[Anonymous No.12157742]

>>12157338
>significantly better
No, it is probably on par or slightly better but Leela which is based on alpha zero, is still better than stockfish12+.

At this point the ai's are approaching nash equilibrium so any improvement will be very diminishing ( counted in couple of extra draws per 1000 games)

[Anonymous No.12157750]

>>12157018
We can't make a general ai because we still don't understand what consciousness even is. Neural science still has a lot of work to do. It's like trying to make a computer without first inventing mathematics.

[Anonymous No.12157754]

>>12157750
We don't understand how brain plays chess or how humans run. Did not prevent us from getting chess ais or trains.

[Anonymous No.12157794]

>>12157087
Chess is still just a game of logic. It's more about memorizing opening moves and counters for moves. It's not a good test for AI. Hence the excitement about the GO playing AI

[Anonymous No.12157811]

>>12157750

The brute force way, like chess AI, would require mapping the entire human context to build this intelligence within. So instead of a chess board, you have to program all the rules of the universe. Lmao

Our only other hope is that neural nets get it in some random Hail Mary. But do we really want sentient AI? I don’t

[Anonymous No.12157818]

>>12157794
>it's more about memorizing opening moves and counters for moves.
It isn't, you have 20 moves in the initial position, and later it goes around 30. That means that even in 10 moves you will have an astronomical immeasurable number of moves. Go is somewhat different, but it is only twice faster in this respect.

[Anonymous No.12157836]

>>12157018
Computers already "think". The problem is whether or not computers have human-like (or even life-like) intelligence. The issue there is that we haven't defined what we even mean by "human-like" intelligence (and the definition is incredibly elusive, to the point that we haven't even gotten something that makes sense in nearly 10,000 years of human history), so instead we're trying to define what computer intelligence is and go off of that. So far, we're having an okay time at progressing in this field. We went from computers that couldn't identify anything to computers that can occasionally identify things within a single generation of human life (although it was dozens of generations of technology).
"AI" as you're describing it may very well be impossible until we figured out a working definition of human intelligence.

>>12157328
>computers
>computer(s)
>>>>>computer(S)
There is a singular function that is comprised of neural networking among thousands of things you could identify any one of which as a complete computer, and that singular function can defeat human Go players. Go bots, in general, are still incapable of handling anyone beyond novices. There are no singular computers that are "massively stronger" than human Go players. This is compared to chess bots, in which case it is absolutely possible for what you could identify as a singular computer system to defeat a grand champion Chess player.
What you've said could very well be a near future, but as of this moment, you're grossly exaggerating.

[Anonymous No.12157858]

>>12157836
To piggy back on my response to >>12157328 about Go bots - if you think AlphaGo Zero could run on anything other than the outrageously sophisticated neural network it exists on and still beat human Go players, you're simply wrong.
The single great breakthrough in machine learning was developing neural networks, which gives us access to enough raw computer power for learning algorithms to do what they do in a meaningful amount of time. Without these neural networks, which cannot be described as a singular computer with any amount of accuracy, machine learning and its results - like AlphaGo Zero - would be dreams in researchers' eyes still.

[Anonymous No.12157869]

>>12157836
AlphaZero is worlds #1 Go player right now with a pretty crushing record over the Master version which crushed humans and that ran on a single computer.

[Anonymous No.12157893]

>>12157754
Yes we do, it's a turn based game we understand exactly how an intelligence plays it. It's literally the most basic example.

>we don't understand how humans run
lol
>trains
lol

[Anonymous No.12157897]

>>12157893
>it's a turn based game we understand exactly how an intelligence plays it.
Wow, such a great understanding! Truly a genius!
But in fact we perfectly understand how humans think too. It's cell-based atom-based thinking process. Here, everything is simpler than it looks.

[Anonymous No.12157998]

Computers can do arithmetic faster than humans
>yeah but we will never get general ai!
Computers out perform humans at predicting complex things like stocks, weather etc
>yeah but we will never get general ai!
Computers out perform humans at chess, alpha go and many other complex games
>yeah but we will never get general ai!
Computers out perform humans at image and speech recognition
>yeah but we will never get general ai!
Computers out perform doctors at some operations
>yeah but we will never get general ai!
...
...
Do you see the pattern?

[Anonymous No.12158004]

>>12157998
Basically this.

[Anonymous No.12158017]

>>12157998
the pattern is that unless you build a machine to specifically do the thing, it will never be able to do what a human can do.

[Anonymous No.12158032]

>>12157998
Seeing patterns where there is no pattern is a sign of schizophrenia, which is a sign of low IQ

[Anonymous No.12158042]

>>12158032
Meanwhile total inability to recognize clear patterns is an idiocy which is a sign of extremely low IQ.

[Anonymous No.12158049]

>>12158017
For example AlphaZero was built entirely from scratch without basing it on previous neural networks development, and when chess AI was developed, it was built totally from scratch too.

[Anonymous No.12158051]

>>12158049
A human may not be as good as these, but a human can feasibly play both. AlphaZero will never ever be able to win a game of Chess.

[Anonymous No.12158058]

>>12158051
AlphaZero was literally given only the rules of Go and then learned how to play it well by playing with itself. It learned to play chess in the same way, by getting the rules (but nothing more, no games, no additional hints) and playing with itself. Now it plays chess better than any human, but also any chess engine made by humans before. So it's the total opposite of what you claim.

[Anonymous No.12158063]

>>12158042
There is no pattern there, schizo. Take your meds

[Anonymous No.12158066]

>>12158063
Looks like a perfect example of projection. So follow your advice.

[Anonymous No.12158068]

>>12158058
I don't think you understand.
the AlphaZero that learned to play Go can't play chess. Neural Networks are shit, they can only do the thing you train them to do, but they can't learn more than one thing at a time. this inflexibility is a problem with the approach in general.
Human can learn many things and apply them, Neural Networks cannot, they only do a single thing.

[Anonymous No.12158074]

>>12158066
You forgot to take these anon, here you are. Do you want nurse Susan to give them to you?

[Anonymous No.12158078]

>>12158068
>the AlphaZero that learned to play Go can't play chess.
AlphaZero is absolutely able to, it learned chess after getting just rules of ches. If your "inflexibility" means "not able to do things without learning them at all", then humans can't do anything too.
>Neural Networks are shit, they can only do the thing you train them to do, but they can't learn more than one thing at a time.
They can, that's what AlphaZero did for many board games.
>Human can learn many things and apply the
Using your definitions they can't.
>Neural Networks cannot, they only do a single thing.
No, they literally can learn any game from scratch when given their rules.

[Anonymous No.12158082]

>>12158074
Anon, I know that you think you are being humorous. But faking mental diseases is often a sign of one. So you should stop spamming /sci/ with your schizoposting and at least get some rest.

[Anonymous No.12158084]

>>12157036
It is not logarithmic it is the logistic function, it first grows exponentially then linearly then asymptotically.

[Anonymous No.12158087]

>>12158082
Anon please take your meds. We dont want the jacket again, do we?

[Anonymous No.12158091]

itt:
>computer beats human brain
well so when computers were able to solve DEs some 50 years ago it was already the case. surely they could do it faster and better than humans with a pen and paper. So?

[Anonymous No.12158095]

>>12158091
If you agree, that's great. Just don't post "brain is 10^24 gflops" revelationms, otherwise you may end like >>12158087

[Anonymous No.12158097]

>>12157998
The pattern is it is nothing new that computers can do many things better than humans. It has been the case since day one.
Even calculators are better than humans. Even abacus. What is your point? It doesn't there is something revolutionary coming that will replace human brain or something.

[Anonymous No.12158099]

>>12158078
>If your "inflexibility" means "not able to do things without learning them at all"
The essence of intelligence is the ability to intuit things without anything really to go on. Nature is horrible, it doesn't teach you the rules, whether you live or die is purely based on guessing how things work.
You also need to keep in mind that the framework within you teach a computer the rules is very distinct from the way people learn, this alone gives them an advantage. But I recognize that AI is a religion for the likes of you.

Call me when my computer can hold a conversation.

[Anonymous No.12158100]

>>12157858
AlphaGo ran on a laptop during the competition... You only need more to train. Imagine not knowing difference between training and inference and feeling like you can speak on the subject...

[Anonymous No.12158103]

>>12158095
Why does it make you angry that the brain performs ~10^24 operations per second?

[Anonymous No.12158105]

>>12158099
>Nature is horrible, it doesn't teach you the rules
Ah, so you need to learn chess not only without any teachers, but also without the actual rules of chess? That's a very profound task, but I'm afraid that zero humans achieved it.
>But I recognize that AI is a religion for the likes of you.
Sorry, but "computers are dumb, but brain is totally clever because it has consciousness soul" sounds way more religious.
>Call me when my computer can hold a conversation.
Go to aidungeon and chat with catgirls.

[Anonymous No.12158108]

>>12158103
No problem, brain is a genius then, but so is the bucket of water. You don't need to to go mad about that.

[Anonymous No.12158110]

>>12158108
But that's not where that number comes from.
If we were talking about the entirety of the evolution of the Schrodinger equation of all the particles in the brain, that is higher than 10^24.

[Anonymous No.12158112]

>>12158110
Sure, with quantum effects it becomes even larger. But when we return to the more defined operations like FLOPS, we see that brain isn't as good as we thought before.

[Anonymous No.12158119]

>>12158105
>Ah, so you need to learn chess not only without any teachers, but also without the actual rules of chess?
Yes, once you take away complete information AI falls apart very quickly. It's almost like computers can only solve problems that aren't really problems in the first place.

>Sorry, but "computers are dumb, but brain is totally clever because it has consciousness soul" sounds way more religious.
But that's not what I am
saying. I am saying that short of brute forcing a brain you're not gonna get a machine capable if the same things a human is capable of. And you're better off doing that one biologically.

>Go to aidungeon and chat with catgirls.
But you can tell that this isn't real, the computer doesn't understand language, it just responds more or less with weighted random responses. There's no understanding of languahe.
Current AI can't write.

[Anonymous No.12158120]

>>12158112
When consider the firing of electrons in the various proteins and cells in the neuron, that correspond to the on or off state of as single transistor, we get slightly higher than 10^24 computations per second.

[Anonymous No.12158129]

>>12158119
>Yes, once you take away complete information AI falls apart very quickly. It's almost like computers can only solve problems that aren't really problems in the first place.
Neural networks are very good at incomplete information tasks, and "chess playing is not a problem" is a true revelation for all the chess players and researchers.
>I am saying that short of brute forcing a brain you're not gonna get a machine capable if the same things a human is capable of.
All current AI developments are far from "brute forcing". You can't bruteforce chess, you can't bruteforce go and so on.
>And you're better off doing that one biologically.
No, biological substrate is awfully slow, hard to copy and has a lot of other negatives.
> weighted random responses.
That describes humans too.

[Anonymous No.12158133]

>>12158120
Sure, but such a wide definition once again allows us to declare a bucket of water a supercomputer. Because just think about the amount of electrons in it and amount of their potential interactions.

[Anonymous No.12158155]

>>12158133
The electrons being passed inside the neuron actually are being used to process information.The entirety of the neuron is not being used to process information. If the entirety of the neuron were used, it'd have a higher number than 10^24. The neuron is not only processing information via the action potential of the synapse. Inside the neuron itself is performing 10^11 computations per second of relevant information processing (this is NOT the entirety of the neuron).
None of the molecules in a bucket of water are being used to process information in the way a transistor or the inner of the neuron cell is. You are purposefully being obtuse and talking around the point for some reason.

[Anonymous No.12158169]

>>12158078
I think this needs some clarification. AlphaZero is an algorithm that can efficiently train a neural network to play a particular closed-system game with some fixed rules. That's it. Each game requires a different trained neural network, a model, to use. There is no single NN that can be used to play all the games it 'learns' and can then be applied to a new game.

Machine Learning =/= AI. We are not even close to creating true AI systems. What we have today is a marketing label.

[Anonymous No.12158171]

>>12158155
No, thinking that "there is some movement of electrons in neuron, therefore you can say that it is meaningful or even can be equated to FLOPS" is the actual obtuseness. If you will count every electron in modern computer, you will get way higher numbers too.
The real meaningful number is the neuron firing rate, and it is less than 10 times per second.
>None of the molecules in a bucket of water are being used to process information in the way a transistor or the inner of the neuron cell is.
Bucket of water is perfect at liquid simulation. Sure, it is a specific task, but that should teach you not to equate very specific tasks with general computations (measured by FLOPS).

[Anonymous No.12158180]

>>12158169
If you want to compare unlearned neural network, then compare them to the newborn in the dark room. Which can do... right, nothing. Guess it is proved that humans can do nothing then.
>There is no single NN that can be used to play all the games it 'learns'
I bet you can easily combine rules of chess and go, teach the neural network and then get the combined network which can play both. Or even make a simple switch which switches between learned games.
>Machine Learning =/= AI.
Incorrect. AI means "artificial intelligence", and it is applicable to a lot of existing software.

[Anonymous No.12158187]

>>12158171
>The real meaningful number is the neuron firing rate, and it is less than 10 times per second.
No, what is going on inside each neuron is just as important. The idea that it is just about the action potential of the synapse is old and outdated and does not hold up to evidence. It is falsified.

[Anonymous No.12158195]

>>12158187
You can measure every atom and every electron inside the neuron, but unless you will clearly show that it presents a separate meaningful operation, it will be a bucket-of-water artificial inflation of numbers.

[Anonymous No.12158201]

>>12158180
> I bet you can easily combine rules of chess and go, teach the neural network and then get the combined network which can play both.
Incorrect. It's not easy. In fact it has never been done afaik. Increasing the complexity of a NN increases the required resources, RAM & CPU etc, exponentially. Everyone creates a specific model for the situation it is required to be used in, that is your switch.

> it is applicable to a lot of existing software.
Wrong again. All the software today is simply a set of algorithms. It is not intelligence. Sure they are complicated algorithms that are created / trained by other algorithms but there is nothing amazing about them except for the amount of data and cpu required.

[Anonymous No.12158210]

>>12158201
>Incorrect. It's not easy. In fact it has never been done afaik. Increasing the complexity of a NN increases the required resources, RAM & CPU etc, exponentially.
Incorrect, just like playing Go does not require a brain twice as larger. In fact it is as trivial as presenting AlphaZero new rules and letting it learn.
>Everyone creates a specific model for the situation it is required to be used in, that is your switch.
Then your initial argument against AlphaZero holds even less weight.
>All the software today is simply a set of algorithms. It is not intelligence.
You should read the definition of AI then. Hint: it's not the same as AGI.

[Anonymous No.12158227]

>>12158210
It's obvious you have never actually used any ML framework or understand the fundamentals of how the various ML algorithms work.

[Anonymous No.12158234]

>>12158227
No, it's you who does not know how computers work at all, let alone ML and AlphaZero.

[Anonymous No.12158287]

>>12157063
it will happen. there is a lot of money to be made off of it.

[Anonymous No.12158312]

>>12157506
False.
Memorizing every single neuron and chemical response to every single stimulus in our body takes more memory than any supercomputer in the world.
What you are talking about is concious memory which humans does not record in the first place.

Memory in the brain is never stored and instead rebuild whenever needed

[Anonymous No.12158318]

>>12158312
Now use these neurons to remember thousand numbers.

[Anonymous No.12158319]

>>12158318
A single neuron performs >quintililion calculations per second.

[Anonymous No.12158326]

>>12158319
So does the bucket of water.

[Anonymous No.12158366]

>>12157256
Are you a reasearcher?
I want to study it in the future, but I'd like to fund myself and not follow the statistics approach.

[Anonymous No.12158383]

>>12157018
what do you think environmental pollution is? it's the robot spreading his living space

[Anonymous No.12158410]

>>12157018
If AI can't think, why can you go online to numerous chat bots right now and watch them think about your conversation in real time?

[Anonymous No.12158412]

>>12157036
>we really aren't superintelligent
You can speak for yourself, but do you really think you can speak for every single human in saying no one with a brain is super intelligent?

[Anonymous No.12158416]

>>12157069
>but we cannot emulate human thought with computer calculation
Why not, what are chat bots emulating if chatting ie conversing is not the way humans communicate thoughts and feelings?
What is the difference between thinking unique thoughts and communicating unique thoughts formed based on what was sensed?

[Anonymous No.12158425]

>>12157112
>It absolutely does not, brutforcing even ten moves ahead is beyond capabilities of any computer,
Chess is a solved game, like checkers or tic tac toe a non losing strategy can always be brute forced.

[Anonymous No.12158431]

>>12158425
Ah, I remember the post claiming that king's gambit is solved. Sadly it was just a first april joke.

[Anonymous No.12158450]

>>12157067

Neuron's don't even perform 1 computation per second.

[Anonymous No.12158461]

>>12157286
>AI fags routinely talk shit about how stupid people are and how pathetic the brain is
Its not the brain that is pathetic, it is the training set of data society feeds brains to turn them into stupid people with irrational opinions and superstitions.

> without having ever produced anything remotely close to its capabilities
AI and machine intelligence pretty much runs most of the world stock markets and commerce at this point, though.

[Anonymous No.12158471]

>>12157372
Are you still confusing conscious thought with the brain's power?

[Anonymous No.12158475]

>>12157532
How many floating point operations does it take to play a round of catch?

[Anonymous No.12158477]

>>12158471
If your "brain power" means every atom movement inside the brain, then such a quanitity holds very little sense.

[Anonymous No.12158480]

>>12158475
How many floating point operations will it take to perfectly simulate the flow of a few liters of water?

[Anonymous No.12158485]

>>12158480
You mean how many floating point operations does a brain need to successfully pour a steady flow of a few liters of water? I don't know what is your guess?

[Anonymous No.12158491]

>>12158485
Now you can see why your playing catch while pouring water fantasies are pretty cute, but not really scientific.

[Anonymous No.12158493]

>>12158477
Its not atoms moving, it is electrons moving about the brain (or anything) that creates power.
Power is current times potential.

[Anonymous No.12158498]

>>12158491
No I don't get your point, are you saying the human brain doesn't need to perform a single floating point operation to accurately do either of those activities?

[Anonymous No.12158504]

>>12158498
I am saying that a bucket of water perfectly performs the calculation of the water flowing down to every atom. You are unable to do this with modern computers and you are certainly unable to do that with your brains. Does that mean that bucket of water has the absolute computational efficiency?

[Anonymous No.12158510]

>>12157858
>which cannot be described as a singular computer
Retarded. Your desktop machine is a single computer even though the CPU is segmented into cores and those cores are segmented into arithmetic and logic units. A supercomputer is a single computer even though its made up of thousands of CPUs. An entirely datacenter could even be described as a single computer in some cases.
Neural networks aren't anything new either. We've just recently built machines which make their use practical. Try doing the barest amount of research before spouting bullshit.

[Anonymous No.12158511]

>>12158493
Water has quite a lot of electrons too. If you want more steady flow, drop some power cords into the bucket.

[Anonymous No.12158514]

>>12158504

>Does that mean that bucket of water has the absolute computational efficiency?
No it means the physics of the universe does, but when a person with a brain wants to precisely pour a litre of cola from one container to the other, it needs to perform a very complex series of calculations that would rival any modern computer that you can't even begin to quantify without a massive amount of science.

[Anonymous No.12158518]

>>12158514
Computer-controlled robots nowadays are able to pour cola too. But that's irrelevant, because we were discussing not some simple pouring, but perfect simulation of the water down to a single molecule. Both computers and humans are unable to do that without, well, looking at the actual water.

[Anonymous No.12158520]

>>12158511
Water is an insulator with tight covalent bonds, electrons don't readily flow through water to create large potentials or currents unless the water is ionized with high external energy, so water isn't a good source of generating physical or electrical power compared to neurons.

[Anonymous No.12158526]

>>12158520
Great, then insert power cords in your brain. That will truly be a great computing experiment.

[Anonymous No.12158527]

>>12158518
>perfect simulation of the water down to a single molecule
No we were talking about how many floating point calculations it takes to do simple second long tasks like pouring water or throwing something and suddenly you became a perfectionist about how insanely high the flop rate of human brain calculation is in reality compared to in computers.

[Anonymous No.12158531]

>>12158526
I can't wait until Electrodes get to be better power cords and we have perfected transcranial direct current stimulation to be able to modulates my neuronal activity and greatly aid the speed and efficiency of learning and refining new and old skills.

[Anonymous No.12158538]

>>12158527
No, we were talking exactly about simulation of everything down to single molecule. Because that's what you get if you insist that we should count every electron inside the neuron (and count it as FLOPS). When we return to more sensible definition, bucket of water stops to be a genius, but brain goes down to milliflops too.
>and suddenly you became a perfectionist about how insanely high the flop rate of human brain calculation is in reality compared to in computers.
But that's not my claim, I claim the opposite.

[Anonymous No.12158540]

>>12158518
>omputer-controlled robots nowadays are able to pour cola too.
How many flops does that require of a robot that doesn't necessarily have thousands of simultaneous conscious and unconscious processes running while doing it?

[Anonymous No.12158543]

>>12158540
>How many flops
A lot, I presume.
>that doesn't necessarily have thousands of simultaneous conscious and unconscious processes running while doing it?
I bet you are perfectly able to run 1000 copies of Doom on a modern computer able to pour cola.

[Anonymous No.12158546]

>>12158538
>if you insist that we should count every electron inside the neuron
You don't need to count ever single neuron, you just need to know the current flow and the voltage potential across the area in question to calculate every single electron and estimate the power performance.

>When we return to more sensible definition
You haven't offered anything sensible, you keep appealing to ignorance and sensationalizing basic measurements like power.

[Anonymous No.12158552]

>>12158543
Why would the amount of copies of doom on my computer affect my ability to pour a liter a cola and what cola pouring robot are you referencing that is running all those copies of doom and why?

[Anonymous No.12158554]

>>12158546
> don't need to count ever single neuron
*don't need to count every single electron in every single neuron

[Anonymous No.12158558]

>>12158552
I don't know, why would your "thousands of simultaneous conscious and unconscious processes running while doing it" even matter?

[Anonymous No.12158567]

>>12158546
We are not discussing "power", we are discussing FLOPS (floating points per seonds) - a generalized measure of computational capability. Nowadays even slow computer provide many gigaflops. Meanwhile brains and buckets of waters provide a very good performance at some specific tasks. Yet they don't provide generalized gflops. In fact, brain can't provide even one flops (some milliflops will be closer to its capabilities). That (once again) means that brains/buckets are good at specialized tasks, yet they are awful compared to computers at general computations.

[Anonymous No.12158570]

>>12158558
Because that stuff is all happening inside the same processor - one brain - instead of your example where most of the copies of doom are not running on the same processor that the robot relies on to calculate pours.

[Anonymous No.12158578]

>>12158567
>they don't provide generalized gflops
>In fact, brain can't provide even one flops (some milliflops will be closer to its capabilities).
That makes no sense, you just got done saying that you can't even apply the concept of flop to a brain because it doesn't provide that kind of generalized performance measurement, so why would a fraction of a flop be any more relevant?

You don't think it takes the brain a single bit of power to calculate what it senses in just one instant of full sensation? How is what we are presented in any given moment that highly specialized and not massively generalized that we are suppose to instantly make sense of based on decades of personal experience?

[Anonymous No.12158580]

>>12158570
You can easily run all these copies on the same processor. In fact robot will necessarily run some OS with a lot of system processes, so that's a good analogue to your subconscious stuff. But why should that even matter for me? If I'm interested in some task, computer running some unrelated tasks is at best neutral for me.

[Anonymous No.12158585]

>>12158578
>so why would a fraction of a flop be any more relevant?
Because brain can make generalized floating point computations. It's just that it will take the brain not nanosecond and not even a second, but (let's be generous) minute to perform one floating point operation. Which will make brain computational power around 10-20 milliflops. Not really impressive.
>You don't think it takes the brain a single bit of power to calculate what it senses in just one instant of full sensation?
"Brain computes in a second what it's like being a brain in one second" is pretty tautological and once again returns us to the bucket analogies. Brain still can run some useful specialized operations relatively fast or even (in a dwindling set of cases) faster than computer. It still does not mean that it has comparable FLOPS power.

[Anonymous No.12158590]

>>12158580
Yes and 1000 combined copies of doom provide much less than 1/1000 of the information your senses are being bombarded with at any one moment.
I mean what is the longest hall in doom a few yards? How far can you see, a few miles?

>why should that even matter for me
It will help you stop underestimating the amount of work your brain needs to do and maybe treat your brain better than someone who thinks their own brain can't do a single calculation in a given second.

[Anonymous No.12158603]

>>12158585
>but (let's be generous) minute to perform one floating point operation
No you are confusing the ability to see the problem, do the problem, write the solution and articulate the answer, when that is not the only calculation necessary.

Just to look at the screen, articulate a reply to this, and physically move your body to act out your will takes your brain billions of floating point calculations.

[Anonymous No.12158606]

>>12158590
>Yes and 1000 combined copies of doom provide much less than 1/1000 of the information your senses are being bombarded with at any one moment.
>I mean what is the longest hall in doom a few yards? How far can you see, a few miles?
Once again, we were discussing finishing a specific task. If instead of doing that task your brain is doing unrelated stuff, then it's like you procrastinating instead of doing work. Impressive, but not really commendable.
But even then your raw comparisons are fairly questionable. How many cells do you have in your retina? How many objects will you be able to remember after closing your eyes? How often? Even here modern computers can have an edge over you.
> and maybe treat your brain better than someone who thinks their own brain can't do a single calculation in a given second.
Ah, we are bullying the brain here, right? Sorry, brain, didn't mean to hurt you, you are a hard worker. Is it OK now?

[Anonymous No.12158611]

>>12158585
>Brain still can run some useful specialized operations relatively fast or even (in a dwindling set of cases) faster than computer.
What are you talking about?
Computers have mastered specialized operations such as playing solved games like chess. Its the generalized intelligence where brains currently still beat computers since brains actually do a lot more constant background calculating than just specialized algorithms from a very small data set in comparison.

[Anonymous No.12158613]

>>12158603
If instead of just doing a problem you need to do a huge lot of additional actions, that just shows that you are terribly inefficient compared to someone who just does the problem. Efficiency of problem-solving isn't measured by additional work and time.
> will takes your brain billions of floating point calculations.
No, it does not. Once again, you shouldn't use a clear "floating point calculations" term when the reality is pretty different.

[Anonymous No.12158617]

>>12158611
>solved games like chess
I don't think that you know what "solved" word means. By the way, humans can still suck at games which are, in fact, solved.
>constant background calculating
That's irrelevant.

[Anonymous No.12158623]

>>12158606
>If instead of doing that task your brain is doing unrelated stuff, then it's like you procrastinating instead of doing work.
Human survival and socialization takes a lot more unrelated background calculation than computers background operating system tasks whereas computers get to dedicate all their calculating to specific tasks, so of course humans will be easier to distract because only they can understand and be distracted, you can not get the robot all flustered by bringing up Drumpf while they are trying to do their thing.

>How many objects will you be able to remember after closing your eyes? How often? Even here modern computers can have an edge over you.
Yes because those are very specialized, specific calculations where human output relies on fuzzy math floating arithmetic to get general estimates since there is so much constantly being calculated through the senses and most of it is superfluous.

How is it possible for a human to walk around in life while breathing without falling or chocking if they are only able to make miliflop calculations in the range lower than the first digital calculators?

[Anonymous No.12158628]

>>12158617
>I don't think that you know what "solved" word means
I explained it already here >>12158425

>humans can still suck at games which are, in fact, solved
which is why specialized computation algorithms that think only about chess are able to beat out human brains that have a lot of other things to calculate just to stay alive.

>irrelevant
It adds to the flops in any given second of brain work since you can't turn most of that other stuff off while the brain calculates a single floating point problem like you can with a computer.

[Anonymous No.12158631]

>>12158171
This guy is literally dumb as bucket of water tho

[Anonymous No.12158637]

>>12158623
>Human survival and socialization takes a lot more unrelated background calculation than computers background operating system tasks whereas computers get to dedicate all their calculating to specific tasks,
Can you temporarily shutdown these tasks and run Doom? Ok, forget Doom, can you at least achieve one FLOPS? No, because brain is too specialized and can't be adapted to the very different tasks, despite all its power.
>Yes because those are very specialized, specific calculations where human output relies on fuzzy math floating arithmetic to get general estimates since there is so much constantly being calculated through the senses and most of it is superfluous.
That means that if we are talking about raw sensory processing, then humans are already bad compared to modern cameras/software. They were almost always better at image recognition (using their very complex and old neural networks), but during the last decade that was changing and is continuing to change now.
>How is it possible for a human to walk around in life while breathing without falling or chocking if they are only able to make miliflop calculations in the range lower than the first digital calculators?
They have a brain which is
a)Very slow
b)Very parallel
c)Tailored for specific survival tasks by millions and billions years of evolution
Because of b) and c) brain is still able to do some tasks well. But computers are way better at a) and can achieve b). Current software progress also forces them to go above c), so when they will achieve that, previous factors will immediately make them better than any brain.

[Anonymous No.12158641]

>>12158628
No, chess is clearly not a "solved game, like checkers or tic tac toe". That's not even a philosophical question, that's factually wrong.
>which is why specialized computation algorithms that think only about chess are able to beat out human brains that have a lot of other things to calculate just to stay alive.
Being constantly on the brink of death isn't really a good quality, and it prevents you from playing games.
>It adds to the flops in any given second of brain work since you can't turn most of that other stuff off
That once again shows that brain is too specialized and can't be adapted to the significantly different tasks.

[Anonymous No.12158658]

People who worships the machine are the exact type of people who does not understand machines to start with

The biological body already boasts an extremely powerful information network that is self-maintaining, self-sustating, self-learning, and self-replicating. All at a 99% efficiency

AIs can never surpass that.
In fact, it is more likely that metals, semi conductors, and electricity are bound to hit its maximum efficiency because relying on electricity increases heat, heat ruins everything, and now have to make even more complex safety features to keep it moving.

On the other hand, the use of carbon-based system that uses glucose as fuel and employs microorganisms to do the hard labor is not just extremely efficient. It also outright cheats the system via mutual symbiosis. The entire process produces very little heat for the amount of data it recieves and releases

I'm sorry, but that's just how it goes.
Even super conductors cannot match the power of the neuron.

Moravec's paradox, my friend

Despite the common belief, AIs doing complex calculations are actually simple and predictable
However, thinking creatively, asking questions, solving problems, and even just maintaining balance is far more complex than you think

Glucose is better than electricity
Blood beats metal
And neurons are greater than transistors.

Machines can only reach as far as the nanolevels (virus size) but the neurons are theorized to be producing conciousness at the quantum level

[Anonymous No.12158659]

>>12158637
>Can you temporarily shutdown these tasks and run Doom?
You can consciously ignore most everything else and get in a zone playing doom where it seems like everything has else is shut down and you are able to run doom more efficiently.

> can you at least achieve one FLOPS?
Your brain achieves billions of floating point operations in a second, its just that the neural representation of the points and operations that your brain is using is vastly different from the symbolic arithmetic in a computer or a society so the process of translating the raw neuron floint point arithmetic to symbolic language makes in real time brings down your method of measurement significantly.

a and b are practically mutually exclusive the more you make operations parallel in a processor, the faster you will make calculations.

>will immediately make them better than any brain.
So they aren't yet better than every brain and brains have more than one flop since computers have been capable of more than one flop for decades?

[Anonymous No.12158663]

>>12158641
https://en.wikipedia.org/wiki/Solving_chess
https://www.chess.com/forum/view/general/chess-has-been-solved

[Anonymous No.12158666]

>reading this thread
Water does calculations what?
Comparing electrons in water to neurons what?

How does this make sense
And how the fuck does psuedocrap undermine the brain and elevates the machine?

[Anonymous No.12158670]

>>12158666
The universe doesn’t care about how you define a computer, you just need to find the right mapping between physical states and game states and you’re golden.

[Anonymous No.12158671]

>>12158670
Shut the fuck up.
If it cannot do calculations, it is not a computer, fucktard

[Anonymous No.12158674]

If you guys wanted to know how far biology is against the machines, look no further than the European robins

For a bird so small, it uses quantum entanglement to know its position on the earth's magnetic field

>European robins may maintain quantum entanglement in their eyes a full 20 microseconds longer than the best laboratory systems, say physicists investigating how birds may use quantum effects to “see” Earth's magnetic field

Machines can never achieve this precision and efficiency because we can only go so far before the transistors become as small as an atom. Which is impossible

[Anonymous No.12158677]

>>12158666
>Water does calculations what?
Water does perfect liquid simulation computation, o mighty Satan. Just like neurons perfectly simulate neurons, nicely do some pattern recognition, yet suck at generalized flops.

[Anonymous No.12158679]

>>12158663
Thanks, now I am converted!
>37... Ka2!!
Truly a work of genius. You can't stop the progress!

[Anonymous No.12158680]

>>12158677
Oh wow
You know fluid dynamics and chaos theory

*Round of Applause*

Now tell me how does this make your AI better than the brain, you talking nutsack

[Anonymous No.12158685]

>>12158677
Why does it have to be perfect, who said anything was perfect?

[Anonymous No.12158686]

>>12158032
Yeah. John Nash had 60 IQ

[Anonymous No.12158687]

>>12158659
>You can consciously ignore most everything else and get in a zone playing doom where it seems like everything has else is shut down and you are able to run doom more efficiently.
You can't run Doom in your brain, you can just play it on some computer.
>Your brain achieves billions of floating point operations in a second
No, it doesn't (not even a flops). Brain does not achieve flops (despite doing some tasks well) and bucket of water isn't equivalent to the googolflops computer which is able to simulate it.
>So they aren't yet better than every brain and brains have more than one flop since computers have been capable of more than one flop for decades?
Computers still can't successfully simulate any large-scale liquid flow. Should I continue?

[Anonymous No.12158694]

>>12158680
No, you tell me about all the petaflops (not milliflops) brain is able to achieve. Then I will tell you about the water bucket computing magnificence. Or better read the thread to avoid repeating this stuff the nth time, o my nutsack connoisseur friend.

[Anonymous No.12158697]

>>12158685
If it shouldn't be perfect, then simulating every electron inside the neuron is not needed too (no matter if you do hypothetical mind upload, or simply do some math).

[Anonymous No.12158699]

>>12158694
Why avoid the question?
You know damn well that if the brain cannot calculate it, your AI won't.

[Anonymous No.12158701]

>>12158694
The bucket of water does not compute anything, though a brain is constantly and measurably generating its own internal movement, power and brain waves while the bucket will just sit there until acted upon.

[Anonymous No.12158703]

>>12158699
>if the brain cannot calculate it, your AI won't.
It's clearly and demonstrably false: AI already does many tasks much better than brain and did it since its inception (floating point operations and liquid simulations included).

[Anonymous No.12158708]

>>12158697
Electrons are just how we model packets of charge that can randomly pop in and out of existence or occasionally exist in a superposition, we don't even have a perfect model of them to know if they are uniform particles that simulate themselves perfectly or if the model needs tweaked.

[Anonymous No.12158709]

>>12158701
>The bucket of water does not compute anything
The bucket of water perfectly computes the position of every atom of water inside. Sure, it's an useless tautology, but it is no worse than "brain computes brain waves"
> while the bucket will just sit there until acted upon.
Some old buckets break on their own and let the water out. That clearly shows that a bucket of water has immortal soul and free will.

[Anonymous No.12158711]

>>12158703
>AI can do math
Meanwhile: the brain is figuring out the meaning of life, creates art, and questions authority and nature

Wow, almost like AIs are just tools meant to provide assistance and not another brain.

[Anonymous No.12158714]

>>12158708
>or occasionally exist in a superposition
Ah, quantum stuff! Brains are really magnificent. But there are even better news. Water in a bucket has quantum effects too!

[Anonymous No.12158717]

>>12158709
>bucket of water has immortal soul and free will.
Might as well just say that the stars are alive because of quantum superposition that allows for nuclear fusion.

You pretentious fuck

[Anonymous No.12158718]

>>12158711
Brains had millions of years for that and did a pretty shitty job. Now let's see if AI will do better.

[Anonymous No.12158720]

>>12158717
True, but that's what happen when you start to count every electron inside a neuron and then to call it "flops".

[Anonymous No.12158721]

>>12158714
Yes which means they aren't perfect and your entire premise of comparing things to perfect computation is a load of horse shit by your own admission.

[Anonymous No.12158724]

>>12158714
Oh shut up over water and speak to us over how binary can be better than DNA

>>12158718
Considering that quantum computing is the maximum possible efficiency of machines which is still expected to be less than the brain, I'm afraid not

>>12158720
Nothing that makes AI better than the brain

All what you are doing is called False Dichotomy.
>Water has flops and electrons therefore AIs are better than the brain!

You stupid fuck

[Anonymous No.12158726]

>>12158711
>meaning of life
https://philosopherai.com/

>art
https://www.artaigallery.com/

>question authority
https://en.wikipedia.org/wiki/Tay_(bot)

>question nature
https://philosopherai.com/

AI does all of that already.

[Anonymous No.12158733]

>>12158720
That isn't how power is calculated.

Why would they need to count every single electron that goes through the energy grid for it to work rather than measuring kWh output at access points?

[Anonymous No.12158735]

>>12158721
Happily there is a well-defined and perfect (almost perfect, but you can achieve a very high probability of success) process. It is called generalized computation and it is measured in flops (floating point operations per seconds). Water in a bucket can't really do it. Humans can, but way slower than even one FLOPS. Meanwhile computers easily achieve many gigaflops. So if you speak in such terms, you have a clear winner.

[Anonymous No.12158736]

>>12158726
That's a search engine and a text generator.
Way to showcase your stupidity in full glory
LMAO

[Anonymous No.12158740]

>>12158724
You should enjoy your new bucket waifu, you dork. Enjoy your gurgling wet love with her! But don't hurt her immortal soul!

[Anonymous No.12158741]

Quantum computing is expected to calculate 10^18 calculations per second

A single neuron calculates at 10^24

~Blood beats metal~

[Anonymous No.12158745]

>>12158726
Nice. Even with modern imperfect AI humans are defeated.
>>12158736
Your brain is a search engine text generator (albeit pretty shitty).

[Anonymous No.12158747]

>>12158741
>Blood beats metal
But water and metal conquer everything!

[Anonymous No.12158749]

>>12158747
Rust
It's called rust

[Anonymous No.12158750]

>>12158733
Because that's how you get your 10^24 absurdity and get bucketmputers a bit later.

[Anonymous No.12158752]

>>12158735
>Humans can, but way slower than even one FLOPS.
Please describe this measurement process and point me to the research that provided you with the baseline categorization you are using for human brain FLOPS.

[Anonymous No.12158754]

>>12158736
>That's a search engine and a text generator.
Which one?
None of them are either of those, though.

[Anonymous No.12158755]

>>12158749
Surely there is an ideal bucket heaven where every bucket is stainless and every drop of water is clear and shiny, and together they do the liquid quantum flopsy liquid simulations to praise the Bucket God.

[Anonymous No.12158760]

>>12157573
>The goalposts have always been a robot that is able to act and think like a human.

No, the GOAL is ethically OK slaves.
AI robots should fill this role.

[Anonymous No.12158761]

>>12158658
>carbon-based system
>blood beats metal

Computers are made of silicon which is a more complex and versatile carbon lattice than the metal iron in blood

[Anonymous No.12158762]

>>12158754
It literally says right there
>it merely mimics opinions, proven by how it can produce conflicting outputs on different attempts.
> However, their output was not truly AI generated art as the user is required to choose images that already exist. This is why these algorithms were criticized for essentially being just like a fancy Instagram filter. In fact, this is very similar to how Instagram filters work.
>Type Artificial intelligence chatterbot

[Anonymous No.12158765]

>>12158750
No, you measure the output power and estimate the number of electrons flowing to make large calculations. You never count anything that large, its absurd, its like adding up the same number thousands of times rather than multiplying but on a much larger and dumber scale.

[Anonymous No.12158766]

>>12158752
1)You get a random single-precision floating point operation task
2)You take the pen (or do it in your mind) and compute that task
3)You clearly fail to do that in one second, but I hope that in a minute it will be finished
4)You calculate your FLOPS rate (hopefully it will be at least tens of milliflops).

[Anonymous No.12158769]

>>12158761
False
Transistors are made out of silicon, magnesium, zinc, and gold
Iron, aluminium, nickle, and cobalt are the main elements of a computer.

[Anonymous No.12158771]

>>12158766
That is terribly articulated, why not just post the original research you are citing instead of trying to poorly translate the information in real time and severely slow down the rate you can communicate the information that has already been performed by the research you are referencing?

[Anonymous No.12158773]

>>12158771
But do you have any research which shows that I should provide you research? Come on, don't be shy.

[Anonymous No.12158774]

>>12158769
I meant semi-conductors not computers.

[Anonymous No.12158777]

>>12157018
> ai beating us at chess is the final frontier. We've made no major advancements since. All progress has stagnated.
ai beats us at go and StarCraft now, both of which are significantly more complex. There's also the many ml algos we use daily. There's also the generative ai stuff and algos that play hide and seek in a 3d environment. I mean Jesus, have you been under a rock? I'm just naming random shit here.

[Anonymous No.12158778]

>>12158773
You are the one claiming specific numbers and procedures, not me, I just want to see the things you claim you are citing without your poor translation in the middle.

[Anonymous No.12158781]

>>12158774
Which also rely on electricity that produces heat.
Semi conductors needed to break their dielectric breakdown in order to process data.

Neurons only needed less than 1watt to produce the right chemical reaction for every stimulus

[Anonymous No.12158783]

>>12157036
Even using neurons, you could drastically increase intelligence by increasing numbers of dendrites and adding new neurons. Depending on the architecture, a brain with even twice as many neurons could be far more than twice as intelligent.

[Anonymous No.12158786]

>>12157069
>we haven't so we cant
dope

[Anonymous No.12158788]

>>12158778
That's pretty articulated again. Sure a link to your peer-reviewed published research will be shorter.

[Anonymous No.12158790]

>>12158781
All materials including carbon can have that problem given silicone and semiconductors are carbon-based.
Neurons have had billions of years to evolve to that state of efficiency.

[Anonymous No.12158792]

>>12158788
For what claim?
You never asked and I didn't throw out specific numbers and protocols.

[Anonymous No.12158793]

>>12158790
And it has long beaten quantum computing - the final state of machine computing

Please tell me how you plan to undermine the brain and make AI look better.

[Anonymous No.12158796]

>>12157811
well sure if you want it to act human, but the goal of a general ai is that it would be a single algo that adapts to any context a human can adapt to. You only need to define the context you want it to operate in.

[Anonymous No.12158797]

>>12158792
Provide me a peer-reviewed research for the demands and claims you made in >>12158771

[Anonymous No.12158798]

>>12158793
cybornetics

[Anonymous No.12158799]

>>12158798
And be a human that is less efficient than a robot and fully available for being hacked and enslave?

Shut the fuck up.

[Anonymous No.12158800]

>>12158701
Nigger you are so fucking stupid its unbelievable. You have no idea what computation is. Your brain has one cell and runs at 10^-(24^24) FLOPS.

[Anonymous No.12158804]

>>12158068
Not always true, depends on the level of abstraction your net operates on. Same reason a human can apply the same strat to different games. However, even if you're right, it just means we need a new architecture, like a neural net hierarchy.

[Anonymous No.12158805]

>>12158797
The only demand was the source for your procedural research claim in >>12158766.

Grammar, syntax and general articulation are not determined by peer review and research publication, but I can just point to your inconsistency to show it does not follow standards that imply well articulated thoughts.

[Anonymous No.12158824]

>>12158805
No, your posts are such a vast source of wisdom so they clearly imply a new paradigm in articulation, communication and peer-reviewing. But it would be a pity if such a wisdom will stay without a review. That's why I ask you to provide some.

[Anonymous No.12158826]

>>12158777
It beats humans in Starcraft cause it has instant reaction (level of micromanagement unachievable by humans). His overall planning and economy planning aren't anythung special and in some cases very dumb

[Anonymous No.12158829]

>>12158824
Here is how researchers estimate petaflop output of the human brain for computer interfacing devices.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6450227/

tl;dr
>One estimate of maximum computational speed required to handle the electrical data in the human brain is 5.52 × 1016 bits/sec

[Anonymous No.12158921]

>>12157338
truth. Stockfish 12 now incorporates NNUE, a small neural net for evaluation in the framework of a normal alpha-beta search. This tech was invented in the computer shogi arena.

Stockfish 12 is playing its first tournament in TCEC 19, where it is currently a few points ahead of LeelaChess Zero, both way ahead of their competition.

[Anonymous No.12158935]

>>12158826
They beat humans with limited micro ability just as well they simple have better overall planning and economy.

[Anonymous No.12159069]

>>12158318
people do that all the time. Have you never heard of memory competitions?

[Anonymous No.12159097]

>>12158663
You're a fucking idiot.

[Anonymous No.12159111]

>>12158800
There are multiple different people itt.
I am the smartest in here btw

[Anonymous No.12159117]

>>12158935
I mean when AlphaStar sees any enemy entering its field of view anywhere on the map it reacts instantaneously. Humans cant do it cause they can only concentrate on small part of the map while checking other parts through minimap. The AI kinda have whole map projected directly in its mind ignoring all the middlemans humans have to struggle with. But thats not why it wins.
And it micromanages ideally in the sense ideally smart, not because humans can't come up with such strategy, but because humans can't switch units fast enough in the mess of tge battle to make it work.

[Anonymous No.12159127]

>>12158745
>Nice. Even with modern imperfect AI humans are defeated.
Wrong
>Your brain is a search engine text generator (albeit pretty shitty).
Wrong
You dumbfucks don't even know how the brain works or how it is considered in modern neuroscience but you pretend fucking GPT3 (which is fucking stupid as shit) is anything comparable to a biological brain.

None of you are intelligent and none of you know what you're talking about.

[Anonymous No.12159138]

>>12159117
I know what you meant, I corrected you saying you were wrong. It can't see the entire map or react instantaneously and it's APM is limited to human speeds as well and it's still grand master.

[Anonymous No.12159167]

>>12158195
It was already shown how it presents a meaningful operation, and you ignore it because you don't like it.

[Anonymous No.12159183]

>>12158740
You are the dumbest poster itt

[Anonymous No.12159184]

>>12158741
No, a single neuron computes 10^11. 10^24 is the total computation in the brain.

[Anonymous No.12159185]

>>12159138
>>12159138
Im not saying that it has map hack, im saying that it sees whole open part of map instantly, while man only sees what is right in front of his eyes, or else AI wouldn't be able to react so fast throughout all the map.
It reacts instantaneously in the sense that it ignores all manipulations that human must do in between thinking about clicking somewhere and actually clicking somewhere. Human must move mouse. It has to be done with presition. AI just clicks wherever it want even if does less clicks than human

[Anonymous No.12159192]

>>12159185
>Im not saying that it has map hack, im saying that it sees whole open part of map instantly
I can read, this is just wrong, it sees what a normal human can see.
>It reacts instantaneously in the sense that it ignores all manipulations that human must do in between thinking about clicking somewhere and actually clicking somewhere.
It has to move it's camera like a human would to issue commands and it has minimum reaction time of 70ms with mean of 350 which is much slower than a professional players it regularly crushes.

Again it plays by the same rules and beats the crap out of humans not because it's mechanically faster but because it simply has better strategy and execution. In fact it plays by harder rules because it must conserve it's actions due to artificial limits while human don't need to. On a third point I don't even understand why the playing field would have to be even, it beats humans on human rules but it can also play by AI rules and be even stronger which just speaks volumes of how much better it is at the game than any human could ever be. You can keep moving the goalpost and the AI's just keep beating them at some point you run out of field.

[Anonymous No.12159253]

>>12159192
Lurk moar

[Anonymous No.12159255]

>>12157018
>t. never seen WarGames

[Anonymous No.12160216]

>>12159183
Don't be upset that your wife is more clever than you.

[Anonymous No.12160248]

>>12158777
I remember the openAI thing in dota was a huge faillure.
Within five hours of it going public, people figured out how to beat it reliably. If you can force game states the AI has never seen before, it just breaks. Humans can still intuit what the right move is, but the AI just does things that make no sense.

[Anonymous No.12160250]

>>12159192
>but because it simply has better strategy and execution
You never played high level Starcraft did you?
It's not about "being better", but about being more consistent. Machines don't get tired, like people do, machines don't get frustrated like people do. This is why people lose.

[Anonymous No.12160257]

>>12157060
AI that single-handedly killed Go

[Anonymous No.12160264]

>>12160250
What if we'll play some game where you can think a lot between moves and they are discrete? Like, you know... chess?

[Anonymous No.12160286]

>>12158829
That's an interesting article. Still the number listed is like order of orders smaller than previous claims ITT.

[Anonymous No.12160287]

>>12160250
And that's AI's fault how? And how are the humans so tired in show matches despite having better reaction times and higher APM, you would think people would sleep before those.

[Anonymous No.12160298]

>>12160287
You never played any high level rts. Fatigue is the main reason people lose.

[Anonymous No.12160346]

>>12160298
And how are people playing their first game of the day in a showmatch fatigued and how is that AI's fault?

[Anonymous No.12160355]

>>12160298
You can then play chess. Actually make it that way: you can think about every move for hours, sleep between moves and even consult with other humans (but not computers). Then the fatigue reason will go away.

[Anonymous No.12160372]

>>12160346
Because you get fatigued as you play.
Go play some Starcraft. I recommend ZvZ, you notice it the quickest there.

[Anonymous No.12160373]

>>12159167
It is a meaningful operation in a sense as water molecule moving is meaningful.

[Anonymous No.12160423]

>>12160372
>nooooo I fatigued so much during the first 5 minutes that I can't play the rest of the game out it was unfair!!!

[Anonymous No.12160427]

>>12160423
Anon, what are you trying to say?

[Anonymous No.12160442]

>>12160372
If the game you play demands you not being fatigued, then losing because of the fatigue is a weak excuse.

[Anonymous No.12160444]

>>12160427
You are saying the players are fatigued. If it's your first game of the day how are you fatigued and how is fatigue the AI's problem in the first place?

[Anonymous No.12160454]

>>12160442
But I am not making excuses, I am pointing out the trivial fact that computers are more consistent over time.

>>12160444
Well, it depends on how the game goes, if it's very active you'll be pretty tired 20 to 30 minutes in.

[Anonymous No.12160477]

>>12160216
You are the dumbest poster itt

[Anonymous No.12160539]

>>12160286
There are multiple different people itt. You are not talking to one person.

[Anonymous No.12160601]

>>12160373
Wrong, this is a non sequitur as was already pointed out, multiple times, by different anons.
The total amount of atoms in the brain is ~4*10^26. The total amount of relevant compute in the brain (when we look at the total thermodynamic information of ATP consumption, which is used to run computations) is 10^24 compute per second.
All of this compute is relevant for the general intelligence of the brain.

[Anonymous No.12160676]

>>12157018
OP I have sad news: you are retarded

[Anonymous No.12160819]

>>12160601
If you will declare "but it's not every molecule! it's 1% of molecules! or 0.1%" it will not become more true for brain or for bucket. It will also not become more true if said by two anons.

[Anonymous No.12160824]

>>12157018
General AI is a pipe dream. It will never happen.

[Anonymous No.12160834]

>>12160824
Never (in 2020s).

[Anonymous No.12160957]

>>12157018
There are multiple paths of creating a super intelligence, right now hardware makes it impossible. Quantum computers are a necessary next step. Check out the emperor's new mind.

[Anonymous No.12161001]

>>12157018
you think some faggots who couldn't beat a computer a chess can program intelligence?

[Anonymous No.12161046]

>>12161001
It's called Moravec's paradox.
Doing complex math and logistics are simple to program but the things that we do for granted like balancing and thinking creatively are extreely complex.

[Anonymous No.12161049]

>>12161046
But it also means that as long as computer learns how to do something, it very soon learns how to do it much better than humans.

[Anonymous No.12161061]

>>12157018
Nowhere in sight? Are you retarded?
It is extremely within sight.

[Anonymous No.12161076]

>>12161049
Nope.
You can only get so far with machine efficiency.
It would hit its maximum efficiency at quantum computing which is expected to still be less efficient than the brain.

[Anonymous No.12161532]

>>12157018
Narrow AI =/= General AI

[Anonymous No.12161543]

>>12161076
Brain is shit, it works nicely only because it was tailored by literally billion of years.

[Anonymous No.12161551]

>brain has billions of years to evolve
>we can't match it after 2 decades
skynet fucking obliterated lmao

[Anonymous No.12161585]

>>12161049
yes, but that's the difference.
A brain doesn't do "something", it does everything.

[Anonymous No.12161606]

>>12157036
I believe it's not a matter of computation units but about how are those units wired and connected to each other

[Anonymous No.12161608]

>>12161543
No, that's literally it.
You cannot get further than quantum computing because if you go further, data would no longer be coherent.

It is unknown how the brain manages to surpass this but it is believe that it does so by simply not storing memories at all. Everything just get remade whenever needed.

>>12161551
You transhumanists always say those
>mind would be uploaded to the web
>machines would beat humans
>VR is the future.
>just give it time
Yeah, shut up

It's called Argument from ignorance
>it cannot be proven false therefore it must be true
Our current models say it is impossible. It is your duty to prove otherwise

[Anonymous No.12161621]

>>12161608
>It is unknown how the brain manages to surpass this
By simply not surpassing it.

[Anonymous No.12161623]

>>12161608
>You transhumanists always say those
And continue to be more and more right. See >>12158004 pic.

[Anonymous No.12161627]

>>12161623
Prove it then.

[Anonymous No.12161673]

>>12161627
Here: Q.E.D.

[Anonymous No.12161676]

>>12161608
The current models don't say it's impossible. In fact they say AI is pretty much a given and we just haven't coded it out yet.

[Anonymous No.12161693]

>>12161623
so where is the AI that all of these things at once?
the problem isn't teaching a computer how to do one thing, but all of them

[Anonymous No.12161695]

>>12161676
Yes it does, mister
Mind upload is impossible because of no cloning theorem
Machine augmentations make you prone to being hacked
And the neuron is far more efficient than transistors which would max out at nanolevels

Allow me to repeat
Quantum computing is the limit and theorized to calculate at 10^18 per second
The brain calculates at 10^24

The future will NOT be metal
It would be trying to synthesize gray and white matter as a source of data storage and process fueled by glucose and super meth

The future is flesh

[Anonymous No.12161698]

>>12161695
Wrong

[Anonymous No.12161700]

>>12161698
Prove it

[Anonymous No.12161703]

>>12161693
>so where is the AI that all of these things at once?
Optimistically in 2020s, realistically in 2030s. Neural network quick development implies the former.
>but if something did not exist before, it will never exist!
That's demonstrably false.
>the problem isn't teaching a computer how to do one thing, but all of them
That's not a problem, if computers can do something, every computer can install a program doing that.

[Anonymous No.12161707]

>>12161700
Brain doesn't calculate at 10^24
Computers aren't limited to 10^18
One of these is by definition true. If you think water makes calculations inside the brain then just sprinkle some water on the processor and now it's going at 10^30

[Anonymous No.12161711]

>>12161703
Anon, putting a bunch of neural networks together to do everything trivially does not work. Unless you approach the problem in a different way, you're not building a system that can do all the things.

[Anonymous No.12161712]

>>12161695
Quantum effects in brain are irrelevant noise.
>The brain calculates at 10^24
So does the bucket of water.

[Anonymous No.12161716]

>>12161707
>>12161712
We already had a good laugh at you insisting that water is a computer
Please save your dignity and just shut the fuck up

[Anonymous No.12161721]

>>12161695
You can trivially get to 10^45 range with power output of a small star at Landauer limit with a Matrioshka brain

[Anonymous No.12161722]

>>12161711
Anon, if you have one program which does one thing and another program which does another, then making a computer which does both is trivial: just install both programs on it.

[Anonymous No.12161724]

>>12161716
But you are the one insisting that.

[Anonymous No.12161725]

>>12161721
You can trivially get to 10^100 with the power of a corn field by combining a few hundred brains.

Get out

[Anonymous No.12161727]

>>12161716
Yes, anon: if brain does septillions of flops, then bucket of water is probably able to do octillions. Of course that's nonsense, but you started first.

[Anonymous No.12161730]

>>12161725
You just got proven wrong care to try again or are you all out of ideas already?

[Anonymous No.12161732]

>>12161725
When you combine several brains, you get decreased productivity (as /sci/ clearly shows).

[Anonymous No.12161736]

>>12161727
Please shut the fuck up
Your water has nothing to do with AI.

[Anonymous No.12161739]

>>12161732
Just destroy the frontal lobe and be reduced to just a flesh computer without sense of conciousness or whatever.

Eitherway, the future would be trying to beat the brain but instead immitate it

[Anonymous No.12161743]

>>12161722
Can you really not see the trivial problem?
For every thing you want your computer to do, you have to make a program to do it.
this is easy, if you want your computer to do two or five things, but if you want it to do every thing, including things it and you do not know about, that isn't going to work.

[Anonymous No.12161750]

>>12161739
*not* be trying to beat the brain

[Anonymous No.12161949]

>>12161736
Water is full of flops and soul, see >>12158755

[Anonymous No.12162012]

>>12161743
That's still partially unsolved problem. But judging by the fact that AlphaZero was created for Go and then was used for chess with almost no tweaks shows that there is a progress. In fact, underlying architecture is widely used not only in board games.