| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10
Comments:
<0> suppose we have a 32 byte string with some random content; it is possible that there are more than 8 different ways to correlate each byte, which easily amounts to a terabyte if there was only a 1 byte charge in making a correlation
<1> "model it in all ways available"... sounds like you're talking about simulating the entire environment internally
<0> Nah, at least I hope I am not. I was specifically thinking of figuring out the patterns / models that would make it work. Simulation is a different matter and I think it can be sidestepped.
<0> humans traditionally come up with only a handful of 'explanations', but really there are many many possible explanations (that can work, many more that can't work)
<1> haven't humans done all right so far in modelling processes in our universe?
<0> we tend to stick with things that we know really, really wel
<0> also, from what are you judging "all right"?
<2> Hey al.
<2> *all
<2> Cool, AIrelay is back
<1> hi andares
<0> do you know of a universal standard that we can compare our progress too?
<0> hi Andares
<2> Does anyone know of a type of neural network that can handle streaming input?
<0> Andares, all of them? You feed the data, it predicts the next state, so-to-speak.
<1> kanzure: obviously not
<0> Andares, right?
<0> athena, heh oh well :)
<0> athena, that would have been really helpful to me
<2> kanzure, uhh... not really what I had in mind.
<2> I mean say, you want to build a neural network to predict temperature based on climate patterns
<2> So you feed it in constantly streaming data
<2> Wouldn't it have to have a method of true storage?
<2> Not storage in the way of the weights recording stuff
<2> more physical storage
<2> Am I making sense?
<1> kanzure: look at human progress... science and technology
<1> doesn't look to me as though we're having much of a problem coping with overhead or data storage
<2> athena, hmm... so could I construct a neural network that would act as a hard drive?
<2> I feed it in input and it stores it, then later I feed it in a "recall" bit and it retrieves it?
<2> Of course it is possible, but wouldn't it take forever?
<1> andares: sorry, that comment was directed at kanzure
<2> oh :x
<2> Hmm
<2> Anyway
<1> andares: why not store the data outside the neural net?
<2> Is there some sort of graphical neural network constructor I could use to play around with various methods?
<2> athena, a sage idea. Partially because I'm not entirely sure how much integration it would have to have
<2> So is there any tool for studying neural networks graphically?
<1> have you tried a google search?
<2> Yes. :/
<1> http://www.google.com/search?q=graphical%20%22neural%20network%22 shows lots
<2> I fail at Google.
<0> athena, 'human progress'?
<0> Of course we're not coping with overhead or data storage .. because for the most part, we're not storing much of anything.
<1> so what r u saying? humans are not making sufficient progress?
<0> As for 'human progress', I think that we can say that the 'progress' is the result of us having so many people working on the same problems.
<0> Athena, that is not what I am saying, because I am not gauging human progress to anything more sufficient.
<0> Athena, this was just a comment earlier today
<0> nothing big
<0> remember? I was saying something like "as I ponder this more, it looks like humans aren't actually 'intelligent' by any hardcore definition, in general"
<1> what does "intelligent" mean to you?
<0> I usually use the definition of "the ability to actually pursue a goal" or worded as "the generation of behavior/actions given a situation without any wrong intermediate steps to get from premise->conclusion/behavior/actions"
<0> right now it's looking like humans can jump from a collection of premises -> to conclusions and still be more or less right, thanks to the resilient nature of reality (or how vague our conclusions are), hehe
<0> a.k.a, guesswork
<1> "the ability to pursue a goal"... by that definition just about any organism is intelligent
<0> and what 'goal' has it set for itself?
<1> what does "goal" mean to you?
<0> A goal? that which should be pursued
<1> what does "should" mean to you? .... ad nauseam
<0> You, sir, have a misguided understanding of natural selection or (biological) evolution
<1> ha... an excellent example of jumping to conclusions :)
<0> yes, but anyway, let me answer your other point
<1> i merely use your definitions... no more no less
<0> 'should' seems to mean 'by the [correct] definitions' (where we should not ***ume that our definitions are correct, but we may ***ume them useful upon experimentation)
<1> this is just an argument about semantics... doesn't seem very interesting
<0> what??
<1> you say it doesn't look like humans are "intelligent", using your definition of "intelligent"... ok
<0> it just started off with my comment :(
<0> you wanted to know what I meant
<1> perhaps someone else will bite :)
<0> okay?
<3> <[FRED]@ef> hi
<4> i try not to bite....too unsanitary
<3> <[FRED]@ef> i from Moscow
<3> <[FRED]@ef> how work by yuo
<3> <[FRED]@ef> ppz
<3> <[FRED]@ef> you
<3> <[FRED]@ef> :)
<3> <[FRED]@ef> tut kto nit" est"?
<3> <[FRED]@ef> AIrelay:
<2> est?
<2> ;o
<3> <[FRED]@ef> ))
<3> <[FRED]@ef> i from Moscow
<0> It's a bot! Kill it with fire!
<3> <[FRED]@ef> ???
<3> <[FRED]@ef> neponimaI
<3> <[FRED]@ef>
<3> <[FRED]@ef>
<2> kanzure, uhh...
<3> <[FRED]@ef>
<0> Andares, I do not understand, either.
<3> <[FRED]@ef> kanzure)
<0> [FRED], yes?
<3> <[FRED]@ef> AIrelay: I do not understand yuo
<3> <[FRED]@ef> )))
<0> [FRED], AIrelay makes a connection between EFNet and Freenode
<3> <[FRED]@ef> I do not understand
<3> <[FRED]@ef> i from Moscow
<4> oy
<0> Alright, who wants to go over to EFnet and kick Fred?
<3> <[FRED]@ef> russian
<0> hah
<3> <Motivez@ef> done
<0> Motivez, thank you
<4> but he was from Moscow
<4> the Russian one
<2> Fred do you speak English?
<4> he's gone
<2> I see.
<2> Hmm
<2> I might use dillo, if it were even near compliant.
<2> Has anyone used Topographica here?
<4> what is that, Andares?
<2> Some sort of tool used to study human neural networks
<2> Not sure past that.
<0> http://topographica.org/Home/index.html
<2> Wow, it looks pretty cool kanzure.
<2> It was quite hard to run when I used it though.
<2> http://topographica.org/images/060220_topographica_screen_shot.png
<2> see here
<0> It does not look difficult to me.. everything -should- be labeled and readable in the manual.
<0> I have not tried it, so I can't speak for it =)
<0> <-- has yet to become a neuroscientist
<0> <-- same here
<2> kanzure, so does it do anything besides image processing?
<0> I am unsure, I just heard of it
<2> topographic mapping?
<2> Cool...
<2> hmm
<2> I think neural nets will really take off if people can duplicate them outside of the digital world
<0> there's more to it than that ... one factor being the exploration of how the biological neural networks, well, do the 'work' in 'net'.
<2> kanzure, but surely it can't be that hard to build a synthetic network using simple hebbian learning.
<0> I have not heard of Hebbian learning, do tell
<2> Hmm... well I am not that learned in it.
<2> It is a type of learning, I believe, that has been proven to exist in I think starfish
<2> or something
<0> starfish, eh?
<2> Anyway, in Hebbian learning, neurons that fire together grow stronger links to each other
<0> haven't thought of those as a particularly enlightening specimen
<0> ah
<2> Hey, they have feelings too!
<2> </badpun>
<2> You have to start simple though, in my opinion.
<0> Hebbian learning concerns the direction of synaptogenesis?
<2> Ironically, I have not heard of synaptogenesis, pray tell.
<0> building of synapses, :P
<2> Wow, lilo is in alot of channels ;o
Return to
#ai
or
Go to some related
logs:
wiki wusb11 ubuntu
#fedora
#linux
#debian
ubuntu removing exim
gentoo gccmakedep not found
nvidia0 fc5 create
#perl
grout58
#gentoo
|
|
