/sci/ - Science & Math

bamp

>>1377827
all we have is faggots building human like robots which cannot even move and then call it AI-Engine because it can recognize patterns.. SUCKS

I entered an AI degree program a year ago, and the ideas and concepts were taught to us. Basically, AI is really, really hard. Harder than people imagine. Even getting them to do the parlour tricks they do today has taken decades of experience.

The brain is only a bunch of well connected neurons.. THATS IT!! THATS FUCKING IT!!
couldn't the same architecture be created virtually?

>>1377883
yes

>>1377881
hmm I see..
what happens when you just create a "brain structure" like this? >>1377883

but this is like a death brain, am I right? there is no life in it and it's just doing nothing

>>1377888
so why is nobody doing it?
also how would the "life" be created in such brain structure? (it needs some tasks)

>>1377890

Well, we have those, they're called neural networks. But a neural network, no matter how complex, is nothing without input - they need to learn to do stuff with the network they have. If you want them to output a certain value if you stimulate a couple of inputs, you need to tell them 'bad!' if they don't do that, so they can adjust their signal weights. For such a trivial task that's easy. For difficult stuff like machine vision or any real cognition, it's much more fuzzy about how to do that. Figuring out how to make them learn in useful ways is difficult. Also, the brain has a *lot* of connections, more than we can simulate atm.

>>1377892
its not a question of not having the design, its a question of not having the materials.

>>1377914
I see, sounds very interesting.
is it possible that we will see an AI soon which is more smart than a human brain and can solve for example scientific myths which human scientists just can't figure out? Or is something like this still Sci-Fi?

>>1377934
so we need better supercomputers?

>>1377951
see the last line of
>>1377914

As far as i know we just can not build something like a brain with what we have.

>>1377951
/g/ has your back there.

>>1377944

I understand a university somewhere has actually developed an AI that can come up with physical laws, if fed enough information. Like, if you showed it a ball moving in a parabola, it would be all, hmm, there seems to be some kind of force dragging that down with a constant acceleration, I think I'll call that 'gravity'.

>>1377959
lol

scientists invent machine that agrees with them

>>1377959
wow..
its a really difficult topic though
since maybe intelligence is not really existing, think about it, actually it's all just logically.. go back in evolution, first there were just nerve cell which became a brain later, the nerve cells where just in charge to move the body, which actually requires a sort of "intelligence" (-> where do I move?).. the human brain is not different, just more developed, therefore it might be just an illusion that we are intelligent, just like our consciousness is actually not existing

>>1377827
well maybe if quantum computers are viable we could possibly create AI an virtual life

>>1378002
what I'm trying to say, the human brain is just action and reaction,
I mean eventually is everything what we think and imagine 100% predictable, and if really so then there is really no intelligence existing, it would be all predictable action and reaction

>>1377827
/thread

they haven't done it because they don't have 'god'

http://en.wikipedia.org/wiki/Timeline_of_quantum_computing
anyone has a clue WHEN it will be ready???

>>1378006
and what about a small, dumb AI Brain? that of a mouse?

You want a smart AI, well heres how to do it. Start off a program, give it hard facts about different subjects and all of the definitions of words. Give it a personality, which is attatched with the ability to get off track occasionally, but it can still return to the original conversation FUCKING CLEVERBOT Then, let it learn about how to take certain things seriously, and others light heartedly through trial and error like a child. Good luck programming that shit, I only have methods of teaching in mind.

>>1378095
Cleverbot isn't just a chatbot, it'll occasionally switch you to talking with another person.

To address the topic, with how rapidly neuroscience and computing power is progressing, it's likely that we'll see true A.I. in our lifetime. That is of course, if it doesn't become illegal.

More neurons in the middle layer.
Better programming of large neural network simulations
and more processors more RAM.

Should be possible with current graphics card technology

>>1377951
I think we would rather need computers that work and are constructed more like brains. Not just faster computers that add ones and zeros.

>>1378015
>I mean eventually is everything what we think and imagine 100% predictable
implying chaotic effects (many body system) and quantum mechanics (uncertainty principle) do not influence our thought processes, which we don't know yet for certain.

>>1378234
just found a reference:
http://en.wikipedia.org/wiki/Quantum_mind

>>1378244
ohh
that might be a big clue.
everything went more complicated than expected..
we need the breakthrough in quantum mechanics first I guess

Quantum corrections are negligible at the scale of the brain.

>>1380490
everybody agree?

>>1377892
they are, the trouble is the human brain is so much more powerful and complex than any computer. I believe recently they developed a model of a cats brain (~10% of a human brain) which could run at 1/1000th of real time.

>>1380603
There have been no neurological studies that indicate quantum features. Its just an ass pull that's popular because it keeps the nature of consciousness more distant and mysterious.

>>1380649
This, the processing power just isn't available yet for strong AI, there has been ubiquitous application of narrow AI but that never counts, people will whine till they get their robot buddies.

>>1378063
>LED quantum entanglement demonstrated
sounds pretty cool

>>1378131
sauce?

>>1377883
Oh fucking really?

Have you considered that the folds of the brain might have a purpose?

>>1380790
wat role?

Yeah small glitch, Neurotransmitters do shit ton of stuff that we're not all that sure about. Networks, its all in the communication. Whats the point of fiber without the right protocols?