| Author
|
mind to computer -- is it ever possible?
|
medir
Inactive User
Started Topics :
113
Posts :
1193
Posted : Jun 28, 2004 14:00
|
 
187 is just a nice number, isn t it ?! 1 + 7 = 8...
...serious...
it was hooked in my brain since i played magic-the gathering ( a trading-card-game ). there creatures which got lots of useful abilities called 187-creatures ( at tournament-jargon ). n i guess there s a movie with samuel l.j. called the same...
bomski
  experiment !
make it your motto day and night.
experiment,
and it will lead you to the light.
the apple on the top of the tree
is never too high to achieve,
so take an example from eve...
experiment ! |
|
|
|
florin
Started Topics :
3
Posts :
16
Posted : Jun 28, 2004 18:04
|
Quote:
|
i think the major problem with such a statement is that it leaves out the environment as a evolutionary shaping tool .
|
|
running evolution implies an environment needs to be modelled for it. we need only model a small volume and introduce information from outside on the surface (like cubemaps in games).
i buy your point that ending with something like _our_ brain will be hard (it's chaotic: weather simulations diverge after a while). but, i think intelligent structures will readily emerge.
Quote:
|
ooh, and more's law is a self fulfilling prophecy.
|
|
right. are you willing to bet against it ? exponential increases are around the corner (quantum computers).
i don't buy your computational complexity argument. the numbers i offered were from a back of the envelope calculation of atoms / volume & cpu needed to simulate that. higher level structures are irrelevant.
let's take this to pm if you want to argue further. |
|
|
|
Top-down
Inactive User
Started Topics :
7
Posts :
119
Posted : Jun 28, 2004 19:37
|
Quote:
|
On 2004-06-28 12:22, medirium wrote:
...but i guess turning "ideas" ( not just commands like at the jet ) into impulses is pritty hard...
|
|
You can see an idea as list (a long one) of commands,something that could be easy with assistance of a strong computer...You will build "Idea 2 reality" presets or use manufactor's ones.
P.S.
I am not sure creating brain-like machine is usefull for MMDI at all.
|
|
|
|
medir
Inactive User
Started Topics :
113
Posts :
1193
Posted : Jun 28, 2004 21:15
|
...afaik the problem isn t a stronger computer, the problem is to divide the thougths into "single signals".
if u concentrate on "one" thought, like steering smt, it s easier than to "create" smt...bla bla 
bomski
experiment !
make it your motto day and night.
experiment,
and it will lead you to the light.
the apple on the top of the tree
is never too high to achieve,
so take an example from eve...
experiment ! |
|
|
|
ZilDoggo
Started Topics :
4
Posts :
663
Posted : Jun 29, 2004 02:48
|
"but, i think intelligent structures will readily emerge. "
yeah but what kind of intelligence will that be?.,
how do we recognise it?
how would it recognise us?
how do we communicate with it?
we have all these things already shaped into our intelligence.,
greets.,
aka., |
|
|
|
zooter
IsraTrance Junior Member
Started Topics :
44
Posts :
771
Posted : Jun 29, 2004 07:04
|
offtopic:
anybody stumbling into this topic by mistake would think this is some kind of a computer research forum  |
|
|
|
eliran17
IsraTrance Full Member
Started Topics :
51
Posts :
168
Posted : Jun 29, 2004 21:53
|
or a secret highly-intelegent-life-form-technology
from another galaxy
<One learns people through the heart, not the eyes or the intellect> |
|
|
|
Amygdala
Amygdala
Started Topics :
12
Posts :
175
Posted : Jun 30, 2004 12:29
|
Top-down, thanks for trying to bring the actual subject back on the track  - I'll give it another go.
I believe it's true, that we don't need any kind of intelligence to interpret brain signals, no matter which kind. As stated earlier, interpretation of simple motoric commands can be learned in a matter of weeks by a machine (asuming this is true, since I haven't the articles to support my claim), so why not the thoughts (in lack of better word) that constitutes ideas of sound...?
After all, the device wouldn't need to think for itself, just take some input and process it. Granted, the input is somewhat complex, but if sheer size of problems had frightened programmers before, we wouldn't have computers as we know them...
- A |
|
|
|
Amygdala
Amygdala
Started Topics :
12
Posts :
175
Posted : Jun 30, 2004 12:32
|
Oh, I forgot to ask...
ZilDoggo, what do you mean by Moore's law being self fulfilling...? Just wondering
- A (again) |
|
|
|
ZilDoggo
Started Topics :
4
Posts :
663
Posted : Jun 30, 2004 13:53
|
amygdala,
two things,
first, why motoric commands are possible and thought not.
motoric commands (and sensory input) have a good and more or less clean interface with the outside world.
there is a limited number of nerves that deal with this stuff and they are easy to isolate.,(since they go directly to the part of the body thet needs them.)
but thought is the result of the whole brain working together.,
we haven't got a clue where we should look in the brain to find a thought process.,
we dont think there is one place where you can find it.,
so you would need to consider the brain as a whole before you can measure things like thought.
but for that you need a full understanding of the brain in the first place., and we dont have that.,
so basically, we can hook technology to the peripheral functions of the brain.,.,
but we dont know what signals inside the brain are used for thought and what their meaning is in the process.,
and even if you find all the right places to measure and you are sure you've only captured tought then you still need to interpret it., and this is a completely different beast., you would need to know what a particular thought means to a brain.,
since YOU cant explain your thought completely to me I would be unable to install a device that would interpret your toughts
a brain doesnt work like a computer., there are no nice linear area's that can be designated a function., there is no specific list of commands., no linear memory., everything is complex .,
it's one big chemical balancing thing that keeps itself alife.,
very hard to interpret.,
we are pretty dumb when it comes to these things.
we would need to run extensive simulations of the brain., but we dont know what is needed in the simulation because we need to run more simulations.,. etc etc etc,
about mores law,
the only reason it is still working is because computer tech companies keep it that way.,
it's not a law, it's a work schedule.,
greets.,
aka., |
|
|
|
UIU
IsraTrance Full Member
Started Topics :
88
Posts :
238
Posted : Jun 30, 2004 13:58
|
|
Slightly off topic. Did you know that Kurzwell, the same guy that made the synthesizers also does research in A.I.? Maybe you should ask him! |
|
|
|
zooter
IsraTrance Junior Member
Started Topics :
44
Posts :
771
Posted : Jul 1, 2004 07:36
|
Amygdala,
One more thing. From what i know, nobody has yet found out how memories are stored in the brian. It is not like a computer memory (flip-flops) or a hard-disk (electric impluses). If man still hasn't found out the secret of memories (which are very much present in the brain), then imagine how difficult it would be to capture thoughts.
I may seem very pessimistic, but at some point of time, we all need to face reality |
|
|
|
Amygdala
Amygdala
Started Topics :
12
Posts :
175
Posted : Jul 1, 2004 12:09
|
Zooter and Zildoggo (what's with the Z's here )
- Very good arguments.
Of course motoric commands are easier to get a hold of, that's completely true. There are, however, also a limited (finite) number of nerves dealing with thought - I don't buy into the "infinite memory" theoris that some people have of the brain.
"a brain doesnt work like a computer, there are no nice linear area's that can be designated a function, there is no specific list of commands, no linear memory, everything is complex,
it's one big chemical balancing thing that keeps itself alife, very hard to interpret"
-The same holds true for artificial neural nets (except about the chemical stuff ). These emulate a collection of the single neurons in an organic brain, although at a simpler scale - I think it would be sufficient for something as non-complex as sound.
Moore's law is indeed not a law of nature but more of a working schedule as you say. But to call it selffulfilling is to go a bit to far - just nit-picking...
And more nit-picking just to honor my education in computer science: computer memory is implemented by latches as well as flip-flops, and storage on harddisks by magnetizing some treated magnetic surface. In the latch/flip-flop case, why shouldn't neurons function the same way? At an abstracted level, neurons form memory by synapsing to other neurons, and by "reconfiguring" themselves as to what to react to. Where flip-flops and latches are discrete binary devices (on-off), neurons can be thought of as a sort of continuous latch...?
Although human memory is "not like a computer's memory" I am pretty sure it can be effectively emulated, and as you probably have guessed, I believe neural nets is the way to go.
- A
|
|
|
|
blueOrb
IsraTrance Full Member
Started Topics :
100
Posts :
1698
Posted : Jul 1, 2004 14:01
|
|
ZilDoggo
Started Topics :
4
Posts :
663
Posted : Jul 1, 2004 15:42
|
" I don't buy into the "infinite memory" theoris that some people have of the brain"
not infinite, just very very distributed and time dependant.,
"The same holds true for artificial neural nets (except about the chemical stuff ). These emulate a collection of the single neurons in an organic brain, although at a simpler scale - I think it would be sufficient for something as non-complex as sound"
well, the neural nets are indeed very simple compared to a brain (any brain)., you loose more braincells in one day than are present in most simulations.,
our brain has about 100.000.000.000 braincells, for instance.,., all capable of dealing with (more or less) analogue input.,
" But to call it selffulfilling is to go a bit to far - just nit-picking..."
hehe, yeah, we couldd argue over it for ever.,
all i know is that moore sometime said that it was funny how computer tech seems to double in speed every 2 years or so.,
since then the whole indusry started running on that frequency.,
"In the latch/flip-flop case, why shouldn't neurons function the same way?"
because neurons are very different kind of switches.,
they allow for aggregation of signals, allow for muliple inputs and these inputs can be weighted differently (different threshold level).,
(in a way neurons behave a little like compressors )
the output can also be weighted depending on either feedback or other aggregated signals,.
or it could be influenced by hormones and stuff.,
the point is realy that a neuron is waaaay more complicated than a latch flipflop.
it allows for spatial configuration, for instance.,
they are not lineary adressed, they are spatially addressed memmory., !!
so one memory unit depends mostly on the output of anoher (or multiple) memory unit.,
and all this paralell!, <- very important!!
this allows the brain to make all kinds of feedback stuff inside which would be a real problem for a linear computer.,
the brain is more like an analouge modular synth than like a computer.,.,
" I am pretty sure it can be effectively emulated, and as you probably have guessed, I believe neural nets is the way to go. "
i, personaly am not that sure about just neural nets.,
the most important argument against plain neural nets is evolution.,
the neural nets we posses in our heads have been subject to environment for billions of years.,
shaped to be more efficiently configured.,
everything that we see as intelligence is the result of a very slow and tedious process called evolution.,
i dont belive that you can get intelligence just by making big neural nets.,
they would lack any basic suff that we see as intelligent, like motivaion, emotions, language, and a few more which i forgot.,
we would need to simulate evolution to come up with something similar., otherwise there would be no reason for it to be there (whatever it is) .,
so there is a limit to what you can achive with a plain neural net., the structuren is just too simple.,
also, remember that neural net simulations do not have a notion of time., they are not 'life' systems.,
there is only a simulation of the information flow through the net., firing off a random neuron on a random time is impossible., (on a significantly large scale)
you would have to simulate an environment with a suitable time resolution (what is the nyquist frequency of a brain? ) and have a neural network live in that environment., you would have to calculate the state of a neural unit even when it is inactive.,
otherwise you quantise out the little details that make brains so complex and able to deal with reality.,. (that's imo)
anyway., nuf said.,
hope i made sense.,
greet.s,
aka.,
|
|
|
|
|
