/sci/ - Science & Math

I have no reaction image for expressing how complicated this task is

and I have a fuckhuge reaction folder.

It's piss easy, just need a complex enough evolutionary simulation + powerful enough hardware, turn it on, 5 mins later shit's talking to you. The laws of physics and conditions of Earth birthed us, it's really not that hard.

I suspect that true AI is impossible. We will have quite sophisticated expert systems, but true AI is, IMHO, not possible.

>>2835800
So your brain works by magic?

Not gonna happen in the next, hmm... 5-6 decades.

>>2835806
My brain isn't very well understood by science at all, now, is it? Nor is the process of reasoning, the integration of emotions into the creative/learning process, etc.

Talk to Roger Penrose about it, bub

Possible, will be achieved. The only limitation is that we are humans, and to err is human.

we have no idea how hypothesis generation works in humans

>>2835848
Any basis for AI being considered impossible has to look a lot like a basis for considering flight/cars/computers impossible. Just because science does not understand it now does not mean that it will never know. You probably don't have any kind of solid reason to suspect science will never know the nature of consciousness.

>>2835855
science can't say what "consciousness" even means

>>2835848
That doesn't change the fact that your brain is 3 pounds of meat. Complex, thinking meat, but meat nonetheless.

We gain more insight into how it works, how it learns, every day. We may not imitate it exactly, but we can figure out enough of what makes it tick. People once thought learning how living flesh like muscle worked was impossible, or going to the moon was impossible.

It's just a matter of time. OP just underestimates how complex general intelligence, especially human-friendly AI, really is.

>>2835855
Heck, the concept of 'consciousness' may eve be outside the realm of science. Remember, the great power of the scientific method comes at the price of a narrow focus.

>>2835777
wait, but with out input how could the AI learn anything? and 5 min would be insanely fast for something to gain sentience probably mind shatteringly so... wait. I just thought of something; To make mentally stable AI's we would have to do one of two things:
Give the AI a preset 'mold personality' for them to fill when they Are made, and then they can grow from there, giving them the tools to cope with insecurities, and the ability to communicate as well.

or raise them like a baby, slowly filtering in more cognitive powers, letting it grow slowly and easing it into it's role.

these methods would prevent a AI from 'Breaking' right off the get go for various reasons.

>>2835868
No, but it can say what complex behavior looks like, as well as intelligent response to a being's environment. Who cares if the machine has qualia like I do, so long as it talks like I do, hurts no-one, and can help me get work/creativity done.

>>2835884
Shit, linked to wrong post

.>>2835799

>>2835883
Oh right.
Because studying the entire fucking universe,
From the sub-atomic parts
To the entire scope of living things
to the drift of continents and seas
to the orbit of planets
to the life and death of suns
to the formation of galaxies and black holes
to the shape, origin, and ultimate fate of the universe

is 'narrow'.

Fuck. You.

I think i might post this in /b/ to see what goes down, good idea?

a theory that just came to my mind would be to let the AI get an understanding of it's own existence and mimic the human brain as much as possible with cameras for visual feed, mic for hearing and some sensors on an arm or two for understanding surfaces. learn it basic programming and let it request changes in it's own system for exploring itself just as humans do the first years. with trial and error it will in theory form a system of what works and not while adapting to the surroundings right? well that's just a thought. to remove the limitations I think we have to give the AI the alternative to "free will" if you may, since it then can "learn" how it can improve on it's own maby?

this subject is as fuzzy as trying to understand how a christian works.

captcha: Trust therfan

>>2835903
wow. who took a shit in your cereal this morning?

>>2835884
I think that what that guy was talking about was doing a world simulation. there's a school of AI that says that you don't even need to understand how mind and consciousness works to make an artificial intelligence. All you need to do is create the situation in where a conscious would make itself. Of course you'd need an absurdly powerful computer and you wouldn't really have any control over what comes out since you're just running a simulation and letting it take its course.

>>2835903
Use the scientific method to prove your mother loves you, dumbass. While you may be amazed at the sweep of the universe, using the scientific method to drive your car to Starbucks would get you killed.
Scientism is as bad as scientology

>>2835744
We're still just slightly smarter animals than the rest of the animals on this planet; this whole 'intelligence' and 'self-awareness' shit is still a new trick for us, and frankly we don't have a clue how it works yet. If we survive long enough as a race to understand it, we could then create it artificially, but until then the best we can do is create something that mimicks the appearance of it.

>>2835925
Ah, huh. that could work.

but then how could we use this 'sim' AI?

>>2835928
Actually, evolutionary psychologists have a pretty good grasp on why friendship, love, loyalty, and caring for offspring occured as survival adaptations in out ancestors.

1) Science built the fucking car.
2) Who in fucks name said 'use the scientific method for everything derp derp'. That is a strawman argument. Science is used for unserstanding the nature of reality and the universe, and suggesting that we use it for everyday learned tasks like driving is like suggesting we eat breakfast cereal with a hacksaw instead of a spoon.

On point. We will figure out intelligence someday.

>>2835949
You dont use it. It just sits there and claims discrimination and asks you for rights.

>>2835968
evolutionary psychology = just-so stories

1) engineering built the fucking car, learn the difference, dumbass
2) It was implied that nothing is outside the scope of science, which is BS

>>2835883
>>2835868
Consciousness is well within the scope of scientific research. Science studies stuff that exist in nature. Consciousness exists in nature.

>>2836004
And engineering has NOTHING to do with science?
You really are retarded.
Every phenomena that we see in nature, from life to intelligence, can be understood using science. Science is for figuring out how and why things work the way they do.
We may not be there yet for intelligence, but we'll get there.

>>2836013
*high-five*

>>2836013
Oh, yeah? How do you measure consciousness? By that definition, love exists in nature, too - so does ethics.

Read some Feyerabend, Charles Tart, and Popper, for your own sake.

>Talk to Roger Penrose about it, bub

Modern Rene Descartes. I see no reason why, in a materialistic universe, a mind cannot be created.

>mfw

>>2836038
heh, those names kinda sound like insults.

>>2836021
In other words, you don't know the difference and don't care to learn.

Folks, the statement "no statements are true unless they can be proven scientifically" is self-refuting. Science cannot answer everything; it cannot even *address* a range of issues important to human life.

To believe otherwise is to be as loony as any cultists

>>2836049
if science can't address something, then that something exists beyond the material

is that what you're implying?

>>2836042

>>2836060
That's what it sounds like.

"Science doesn't understand it now.. it never will! The mind will always be mysterious!!"

I could program a simulation for anything and give the creature a goal that tries different things. Whats the problem were trying to solve here?

>>2836060
As far as natural science, or 'hard' science, even more so. Economics is certainly not a natural science, nor is sociology, psychology, etc. Economics, politics, ethics, morals, etc. - all beyond the scope of science in the rigorous sense.

>>2836093
those things arise from the material, so are still in the scope of science

>>2836102
"People are material, therefore science can explain everything"
You had better *never* criticize a theist, pal.

Just a hint; go to google and look up "scientism" - it will make your life better

>>2835806

Consciousness in an emergent and distributed phenomena arising from high-bandwidth interconnection between multiple brain regions.

The reason we can't understand and predict it is because a lack in sensor technology and a lack in simulation technology. The same applies to material science, weather, chemistry, physics and god knows what else. You don't argue that superconductivity is a magic phenomena existing outside of science because it's badly understood(especially high temperature superconducts).

We have limited simulations of neural structures, their behaviour matches real life observations.

You have to understand that when it comes to advanced science we're an infant civilization. The transistor is less than one hundred years old and every year that passes results in big improvements in circuitry and digital technology. Our cutting edge technology encased in pure white, smoothlooking plastic is actually very primitive, and in twenty years everything will appear severely outdated, twenty years after that it will be outdated again.
We're far away from mature technology. And if you wonder what mature technology would be? Say a brainscanner with molecular precision, oh you could improve it to atomic precision, but with no functional advantage when it comes to scanning brains. Now that's what mature technology is. Do we have any such things? No.

Reposted in TG
>>>/tg/14484381

>>2836116
you jelly of my superior logic

Semantic gap.

In essence, computation (as an abstract process) is syntactical. Human consciousness, at least, is at least partly semantic in nature; that is to say, that the meanings of words do not arise entirely out of syntax, but out of semantics as well. This is part of the reason why human beings seem to make such shitty computers.
As such, any true artificial consciousness (which may or may not be the same as artificial intelligence) would either be based on wholly different principles to any consciousness we have encountered, which is one hell of an engineering challenge, or cannot be computational.

I agree with you OP. Intelligence is limited to or coding, not our hardware. The true is that we don't know much about intelligence to make a good program that would analysis the world like we do.

>>2836146
I am saddened by your ideologically-imposed ignorance

>"There are good reasons to believe that we are at a turning point, and that it will be possible within the next decade to formulate a meaningful understanding of brain function. This view is based on several measurable trends, and a simple observation which has been proven repeatedly in the history of science: Scientific advances are enabled by a technology advance that allows us to see what we have not been able to see before, and for the first time in human history, we can collectively observe the brain at work in a global manner with such clarity that we will be able to discover the overall programs behind our intelligence, and by doing so coming face to face with our own selves. From there, we can build more advanced systems by borrowing ideas from the brain. This creation of greater intelligence will be a throwing away of all the previous rules, and perhaps in the blink of an eye, an exponential runaway will occur beyond any hope of control."

>>2836149
If we had abundant computation we would have artificial intelligence already. Build a physical simulation sandbox of a few square meters of highly detailed enviroment to match real world scenarios, then run highly-detailed neuron simulations with an evolutionary algorithm until they wire into something intelligent.

Henry Markham is running his blue brain project with neuronal simulations, but due to a single neuron requiring the equivalent of a laptops computing power he can't really go ahead and simulate all hundred billion neurons plus an enviroment to see if it really works.
There is a very real hardware barrier, maybe we could code our way around it, but pouring more hardware at it is a very real solution too.

>>2836133
this.
the reason we don't have AI at the level you speak of is that the processing power required is still prohibitively expensive, fortunately the well documented trends of price improvement will make such things possible in about 15-20 years.

>>2836149
Not really, the “secret” to the brain’s sophistication is that its structure is built by experience, the seed principals that brain starts with (genes that deal with the brain fit on a CD and if streamlined are probably less than a hundred megabytes of data storage).

I never realized there were so many dualists on /sci/.

>>2836183

The problem is that we still don't understand how neurons work, fully.

We need better tools to understand those and the brain as a whole, it's not only a matter of computing power.

>>2836183
today you can fit all the computational power used by the apollo missions into your pocket

give it a couple of years

>>2836183
We will probably have the computing power within say 15 years, so the hardware barrier is probably only temporary. However, one could suspect that a full blown brain simulation is far from the optimal way to code an AI. A brain is something that evolved after all, and evolution can produce a lot of unneeded stuff, and it should also be possible to abstract away large parts of the reasoning/learnign process.

>>2836202
we understand them well enough. We can modell one neuron very well, the problem lies in selecting what parts are important and which parts we can neglect, since simulating a brain on the level of simulating every individual neuron is completely untractable with todays (and the near future) computers.

>>2836202
We know their electric profiles quite well and we know the cause for the electric profiles.

Things like synapse strenghtening and connectivity might be more tricky but molecular studies are doing a bloody good job of teasing out the fine details of it.

When the relevant computer power is there to actually conduct the simulations well(in twenty years we'll have petaflop cellphones provided nothing breaks the trends) i'm pretty sure we'll have studied neuron physiology to the point where we know all the relevant details to do mass simulations of them.

>>2836221
When you have a simulated brain the possibility of doing virtual experimental neuroscience in addition to accelerating the subjective measure of time by factors of millions and whatnot else means that you will very quickly end up with an entity that is brutally intelligent.

Consider this: A virtual brain can have previous states loaded. Meaning you can do a lobotomy, let it play around as braindamaged for ten days, and then be reverted to its previous state. It could also be allowed to keep the memories from its lobotomized state. Also, you could add ten new hemispheres to it, give it a huge supply of virtual heroin and the library of congress and accelerate the simulated enviroment to 100 times normal speed and leave it for a month. When you come back the super-mind have spent 8 years locked into the library of congress with nothing to but read books and use heroin(give it a few grams of high purity digital heroin for every book it finishes). Now what happens if you can give it the internet and accelerate its thought-enviroment one hundred thousand times?
An acceleration of 100k means that ONE MONTH becomes over EIGHT THOUSAND SUBJECTIVE YEARS.

Even with an abundance of digital pornography and video games it would end up learning vast amounts of data just by random internet lurking in that time.

humans are AI dude. you mad?

>>2836299

No, it makes sense.

>>2836299
Are we 'artificial'?

>>2836299
>>humans are AI
>>humans are artificial

howaboutno.jpg

>>2836202
DISREGARD THAT I AM A FAGET

If the human brain were so simple that we could understand, we would be so simple that we couldn’t.

>>2836039
JELLY

True artificial intelligence will never be possible. It requires a fundamental understanding of the brain, which we cannot obtain since the fundamental processes of the brain are encoded in the soul.

>>2839362
>soul
shsh ... wtf troll
-3/10

pic related, it's you

>
The way I see it is that an AI needs a set of codes that dictates how it can write new codes; AKA how it can learn. But the fact that those codes are there put a restricting limitation on the intelligence, because it can only learn in the way it is coded.

Humans have the same problem, as they are also bound by the fundamental rules of the physical realm.

>>2839414
the main problems is in the cognitive part i guess

>>2835744
very related
utubes :
v=8ZmO948g4Q4
v=8ZmO948g4Q4

>>2839362
>>2839403
He is trolling.
But he is also repeating what many philosophers claim.
The bullshit he wrote is a popular view in academics.

>the brain is like a television receiver

>>2839981
then: 4chan 1 x 0 Academics

>>2836240
>we know how neurons work
http://www.physorg.com/news/2011-02-rewrite-textbooks-conventional-wisdom-neurons.html

>>2835903
It is narrow in terms of methodology and kinds of objects of inspection, not in number of objects of inspection within one category, or within-category scale. Twat.

>>2835914
>eating cereal
>see something brown floating in bowl
fuck. How did I not notice that before?

>>2835744
We lack parallel computational power, among other things. If you look at specialized supercomputers, they can sometimes emulate the abilities of a dumb child at a single, specific task. General AI takes much more than that.

>>2836202
the better tool is already here

http://www.stanford.edu/group/brainsinsilicon/neurogrid.html

>>2835800
look at the neurogrid link I just posted

the only barrier before achieving AI is processing power.. we still need a lot of it to compare to a human brain and mathfags are working on it all the time

thread too full of trolls to finish reading

It always terrifies me to think about the issues people are going to face in the future while experimenting with this stuff. I mean, imagine you have a sort of retarded AI crudely fashioned after the human mind and you're fucking around with it, driving it into insanity and stuff but of course it's a thinking and learning being isn't it? (or at least part of one). Press the off button.. whoops you just killed it, well never mind we can randomly generate another billion...

>>2841036
We don't have the software, or any serious idea how to write it. You can have all the processing power in the world and if you don't know how to structure the task it won't even get started.

>>2839981
You know nothing about philosophy. You cannot name one philosopher who in the past 50 years has said anything about the mind being encoded in the soul. Idiotic blowhards like you give science a bad name.

>>2836133
We can't even model the structure of ONE protein molecule, and you want to model EVERY ONE?

>>2841368
>You can have all the processing power in the world
Until recently that wasn't very much. Still it's not all that much either. But say if you give me one thousand exaflops of processing power and a team of computational neuroscientists i'll figure something out.

>>2841368

I'm curious ... Did anyone tried/played with that comp. lang. called Church ?

Nothing is keeping us from it. Nothing has kept me from it.

What you are missing is the possibility of infinite input from the outside world, and structuring the whole AI based on making sense of that input and using it to drive behavior.

The human brain is constructed using uniform units. So are the brains of many other life forms. The issue is primarily a mathematical one. Nature clearly has a general algorithm for organizing streams of information.

An association data rule mining algorithm. You watch. You notice groupings. You notice groupings of groupings. There is more to it, but I am not about to make it public knowledge. North Korea's automated grenade launching border control turrets are already deadly enough.

The driving factors revolve around something similar to the nietzsche-ian principal of "Whatever doesn't kill you makes you stronger". This is really more of a law of nature than it is simply a philosophical principal. The simple act of recording current conditions and trying to recreate them in the future is good survival behavior because you weren't dead when you recorded the conditions thus are likely not to be if you recreate them. This is a huge driving factor for many things in nature.

>>2841645
>North Korea's automated grenade launching border control turrets
Those are south koreas.
Unless you mean the north korean guards are so indoctrinated that they are functionally automated turrets.

South Korea's right... sorry

The missing link is a sufficiently realistic ontology. You neurotypicals are too wrapped up in your survival drives to pay proper attention to us Asperger people when we hand you the answers on a silver platter.

Read "Implied Spaces" then come back and ask again, OP.