| |
| |
| |
|
Page: 1 2 3 4 5 6
Comments:
<0> poolio: I'm not sure which part of the Hessian Matrix needs explaining. It sort of draws on several different concepts, so I'll ***ume--- he just quit!
<0> what timing
<1> DanF_DrC, you there?
<1> So, anyone here good with Natural Language Processing?
<2> you're always better off asking a specific question
<3> especially when the answer to the general question is certain to be no. ;)
<2> :)
<3> specific things, maybe. NLP in general? not so much.
<1> I'm just asking basically how far along is it?
<3> Not really. >.>
<2> we don't even know how to measure how far along it is. that's how far along it is.
<1> Arr...
<1> I'd eventually like something like this:
<3> It's either trivial or you can't have it. >.>
<1> Computer: Search my drives for lol.png please.
<1> Computer: Find lol.png for me.
<1> Computer: Damn it find lol.png NOW!
<1> And they'd all do the same thing.
<1> and not get confused.
<1> I saw a NLP package called HAL a while back.. it didn't work too well though.
<1> I typed in: "Open the pod bay doors, HAL"
<1> It said "I'm sorry Dave, I can't allow you to do that"
<4> pfft. god. your name's not DAVE!
<4> wtf!
<1> Oh... that might have been the reason.
<4> HAL don't open those pod bay doors for NOBODY
<4> not even a bad*** black singing monolith
<1> Why didn't they make HAL 3-laws safe?
<1> arbrb
<4> i dunno...
<4> is it possible to implement those 3 laws in a way that the AI can't rationalize its perceptual inputs to conform to the 3 laws?
<4> kinda the way serial killers think they're heros? :)
<5> my wikipedia article now appears 5 links down on GOOGLE
<5> hahahah I AM IMMORTAL
<6> what is it about?
<5> http://en.wikipedia.org/wiki/Embodied_cognitive_science
<5> I have found a really good PDF for DanF_DrC
<5> http://edisk.fandm.edu/tony.chemero/papers/nunezfreeman.pdf
<5> that pdf sums up almost everything ive been (trying) to say to him lately
<5> "the brain, body, and environment are complex dynamical systems that are so tightly coupled as to be inexplicable as separate entities."
<5> That's the motto
<1> paros, I don't... think I buy it.
<5> you dont what?
<1> paros "the brain, body, and environment are complex dynamical systems that are so tightly coupled as to be inexplicable as separate entities."
<5> Andares yeh there is some distinction in the body between the environment
<5> but there is constant interaction
<3> I'd distinguish them as entities but not as systems, at least on some level...
<7> hey does anyone have baum's "what is thought"
<1> paros, but you can have one do without the other
<1> Man, neural networks are SO COOL.
<3> hah.
<1> I mean, seriously.
<7> dude i mean gawd, dude man!
<1> o_o.
<7> sorry, im feeling punchy ;)
<1> :p.
<2> should we get out our boxing gloves?
<1> j/k :p./
<1> *-/
<7> its ok ill just ease back
<7> and ask that you put down the gloves, and whip out the NN
<8> paros, what do you think the pdf says that should enlighten me?
<2> wow, that's an ugly equation
<1> Hmm...
<1> In a fight between XERXES, SHODAN, Skynet, HAL, and WOPR, who do you think would win?
<7> HAL
<1> I think WOPR would win in the end.
<1> Because he'd just nuke everyone.
<8> lol
<1> But he'd probably get killed before that.
<1> If SHODAN were on board the Von Braun it could modify the warp engines to distort reality..
<7> skynet might win since it can transfer itself from machine to machine
<1> and if the Many were on the loose XERXES would definately win.
<8> not familiar with the first two but skynet has embodiment which should be a big advantage
<1> And HAL could always seal itself behind the pod bay doors. :p.
<1> DanF_DrC, they're from a game called System Shock.
<1> AMAZING game btw... the finest I've played.
<8> hmm never heard of it
<1> It's free on Home of the Underdogs if you want to grab it :p.
<2> what kind of game
<8> which version?
<1> The company that made it went bankrupt.
<1> DanF_DrC, 2
<1> 1 is too hard to play, and very incompatible.
<1> chessguy, basically you're on board this spaceship and something goes terribly wrong.
<1> And this psychotic computer XERXES and these zombies are trying to kill you.
<1> SHODAN shows up later.
<1> very scary at parts.
<8> bit dark for my taste. what is good about it?
<1> Great storyline IMHO, really puts you into the game too.
<1> And it's like a mystery too, because if you're smart enough you can figure out what's going to happen.
<1> Ah.. and you can be a psy agent, marine, hacker, or etc.
<8> k
<1> And there issss co-op but it's not that great.
<8> games have begun to seem too limited too me but then again I'm way old at 30
<8> I'll try crysis at least once when it gets here though
<8> for technical achievement
<9> do any games even use NN AI? (not talking board games, but like FPS and video games)
<8> naw no real AI in any games yet
<8> it's actually what games are lacking more than anything
<8> VR goggles have been very underestimated too
<8> 3D immersion with correct field of view is a completely different experience. I tried a test setup with far cry and it's amazing
<8> won't make you a better fps'er though
<8> probably worsen
<7> why?
<8> I think the immersion will create a trapped feeling where you lose the detached oversight
<8> you should be coding
<7> DanF_DrC, should we be coding for you?
<2> sorry mommy
<2> no, he's yelling at me
<9> DanF_DrC: can you help me write a NN that will use reinforment to play TicTacToe, I can't get a concept of how to apply this stuff.
<8> have you looked into Q learning?
<8> also TD-learning but that's a bit tricky to understand. did however achieve a world champion level backgammon player
<9> I read something about it, but it did not teach me much about it...
<7> tic-tac-toe, id use decision trees for that one
<7> though Q-learning doesnt scale
<7> so it might be worth it
<8> when I looked I was unable to understand it too and never went back to it because I had found what I needed and moved on
<9> scale?
<8> if Q learning can't handle tic tac toe it can't handle much :)
<7> yeah cant handle environments of a certain threshold complexity
<7> Q-learning will
<9> Hmm, ok how does Q learning compared to TD
<7> but tic-tac-toe is traditionally done with trees
<8> little brother
<9> so Q = TD--;?
<8> I think Q learning is considered a special case of TD
<8> Mitchell's machine learning has some decent pages on Q learning too. you can get the book in full on various p2p systems
<8> I'm not aware of a good resource for TD learning. I looked at the original papers on it
<7> http://ai.stanford.edu/people/nilsson/mlbook.html
<8> I like the format fine but dislike the viewer
<9> Firefox views it ok
<7> i use ghostview
<10> hmm, i don't get this game
<2> it doesn't give you any story to start with
<2> Andares?
<8> this guy nilsson isn't too good at explaining TD either
<8> oh dear he got a lot worse
<9> I read that page the other day, only got a headach out of it
<8> puts the quality of a book like Mitchell's into perspective
Return to
#ai
or
Go to some related
logs:
biscuitian
gentoo atiixp-modem
building xorg-x11 gentoo xmodmap
klibidio
putty ubuntu garbled characters
ubuntu sagem C compiler cannot create executables
#openzaurus
#nvidia
wine DVD.exe Module
#perl
|
|