| |
| |
| |
|
Page: 1 2 3
Comments:
<0> "emotions are not needed if evolution is real". WTF? The number of problems with that statement ... (the ***umption evolution only produces what is "needed", the ***umption that emotions AREN'T "needed" ... aye yi yi.) Dan just keeps drifting further and further off the deep end.
<0> (Oh yes, and the ***umption that "if" evolution is real - which it is. certain observed fact. - it's the only thing going on, which is a claim nobody makes.)
<1> verbie, what do you think about evolution?
<1> nothing i guess
<1> verbie, but you kind of evloved too, right?
<2> Magnus_RM: most research into the ant brain disproves that the left and right hemispheres dislike their electro-chemical perception.
<1> verbie, sounds complicated
<2> Magnus_RM: our faith is the holy way, any other paths are made up by satan.
<1> sure
<0> amusing bot.
<1> well, sometimes he is, sometimes he is a bit boring
<1> verbie, how do you feel today?
<0> :P
<2> Magnus_RM: i feel good today. how are you?
<1> verbie, i am ok, a bit tired
<2> Magnus_RM: take 4 liters of ketchup and mix it with 8 hg of salt, boil for 3 seconds. add 7 tablespoons of black pepper. serve with blueberry juice.
<1> verbie, that sounds horrible!
<2> Magnus_RM: the free energy in vacuum is taken into use by investing micropulses of red noise into the harmonics of electrostatic monopoles.
<1> yeah right
<1> actually he performs better if we say longer sentences to him, where he finds a word that is the same as in one of the sentence generators, because then his response will be a little more relevant
<1> like this..
<1> veribe, do you like philosophy?
<1> verbie, do you like philosophy?
<2> Magnus_RM: your philosophy is not altogether the magnificent speculation, but also the entire stance in which it gives another persepctive, and the conclusions which have not yet been understood.
<1> verbie, i am amazed, you make me understand existence now
<1> he happened to produce a sentence that was too long now, so he skipped replying, i will try that one again
<1> verbie, i am amazed, you make me understand existence now
<1> hmm, same thing, i need to fix that, havent had time
<1> verbie, i am amazed
<2> Magnus_RM: the solution of the last theorem of fermat disproves truths such as nash equilibria.
<1> verbie, i did not know you were an expert in mathematics
<1> damned too long reply again, so he stopped
<1> it is because of the 255 or whatever limit here on irc
<1> damned too long reply again, so he stopped
<1> verbie, i did not know you were an expert in mathematics
<1> same thing
<1> ok, enough of this demonstration :)
<3> nice
<3> almost got me fooled when it mentioned disproving the nash equilibrium
<1> if you want to try him out without these length limitations you can try him out at the chat function on http://sentence.bigparadox.com
<1> buggaboo, yeah he sounds impressive sometimes
<1> and by the way, if you create some sentence generator at that site i just gave, then verbie will use them here when chatting
<1> ops, i mentioned his name, so he will reply
<2> Magnus_RM: raytracing is often a stupid way to compute visualisations.
<3> yeah but I'd like to see how the nash equilibrium 'disproven'
<1> buggaboo, i think you need to ask him about that :)
<3> Verbie, I'd like to see how the nash equilibrium 'disproven'
<2> Buggaboo: the idea of infinity disproves the lack of nash equilibria.
<3> Verbie, what's infinity got to do with nash equilibria.
<2> Buggaboo: the idea of infinity disproves the lack of nash equilibria.
<3> hm... it should track it's sentences
<1> he will tend to repeat himself if you use his words in your reply, since he will do the same when replying to you :)
<1> buggaboo, yes, i want to put a last access date/time stamp on generators that he has used so that he does not reuse them too early
<1> verbie, i think you are stupid
<2> Magnus_RM: the archetype is not undividedly our stupid scheme, but the stance by which it exists, and some of the consequences which come with it.
<1> here he found the word "stupid" in a generator and used that
<3> Verbie, you don't lack no intelligence
<2> Buggaboo: hmm, i don't know what to say.
<1> :)
<3> Verbie, you don't lack no braincells
<2> Buggaboo: well, don't really know what i should reply.
<3> Verbie, you don't lack no underwear
<2> Buggaboo: well, don't know what i should say.
<3> hehe
<1> i think i have the word "don" or "don't" in the word list, but since it is such an unsignificant word i should remove it from the word list, since it is not a word worth getting hung up on, but that is what happened now
<4> Verbie, who are you?
<2> chessguy: every day i will become smarter and more confident. i possess tremendous creativity.
<4> not to mention optimism
<1> :)
<4> how does Verbie work?
<2> chessguy: girlfriends will say bizarre things to insult your ego, which could possibly be totally abnormal.
<1> he will reply to that :) after that i will explain
<4> heh
<4> good thing my girlfriend's on the phone and not paying attention, since she's sitting right next to me
<4> Verbie: how do you work?
<2> chessguy: my smallest shortcoming is not that i am best.
<1> i made the site http://sentence.bigparadox.com where one can create sentence generators, then later on when they became many, i though that i could make a mIRC script that connected that material to a chat robot here, so i did, and as i said earlier he selects a generator that contains a word that is in the sentence you said to him
<1> so it is all extremely simple
<4> interesting
<1> so if you go there and make some new generator then v e r b i e will use it here if he finds it useful
<3> obtw, Magnus_RM check this out http://www.interactivestory.net/
<3> download it play with it
<1> buggaboo, sounds interssting
<3> fun!
<3> I'm going to bed see ya tomorrow probably
<1> i had an idea yesterday night for how to make a computer program create its own stories
<5> cya
<1> ok bye
<3> bye
<6> it appears that tokyolectures has gone 404 on us
<7> may it rest in peace.
<7> did archive.org manage to back it up before it was vaporized?
<6> what do you mean by 'vaporized'??
<7> it evaporated into the ether
<7> :D
<7> it may pop up again
<7> but who's to say!
<6> ah
<6> yes
<7> so the question remains
<7> what is the url of the site that might have vanished
<6> what do you mean
<7> well, websites have urls
<7> and questions are requests for information
<7> so, i'd like the information which contains the url to the tokyo lectures site!
<6> oh
<7> hey, has anyone here played with online learning algorithms before?
<7> ones where you have a human acting as a teacher doing something, and an AI looking over their shoulder trying to find the hidden patterns in what they're doing?
<5> epitron, that would just be AI in general
<5> in the pretend-ai domain some have done some mimic software where a bot recognizes face movement and does the same. crude and useless
<8> This morning I read the pomerlaeu paper on automatic driving
<5> ok
<8> well to be fair I read the beginning and skimmed the middle since i'm not interested in transforming images
<8> what struck me was the network structure
<8> 960 input nodes, 30 output nodes and a single layer of 4 hidden nodes
<5> simple yet effective
<8> i was skimming looking for somewhere where they discussed how they arrived at that representation
<8> but i didn't see anything about that, did i miss it?
<5> haven't read it fully either. I figure it was a hardware speed limit
<8> this is exactly the kind of info that I want and the researchers seem to omit
<8> why 4 and not 6, or 2 or 1000
<8> i'd like to know ;-)
<5> you try them and pick the one that works the best given your other criteria
<5> trial and error and some experience
<8> but if they never report their trials and errors or the heuristics they develop
<5> not aware of any good techniques for determining without trying
<8> how do any of the rest of us ever benefit
<8> did they try it with 3 and it didn't work? did they try it with 7 and find it didn't benefit them
<5> most likely
<8> it seems improbable to me that their first thought was 960x4x30 job done!
<8> another question that occurs was why they went with 30 output nodes... i take the steering "hillock" pov
<8> but surely one input -1 ... 0 ... +1 on a logarithmic scale would have served
<5> you mean output. and yes BUT that function could have required more processing and precision and be heavier
<8> sorry yes, one ouput
<8> i agree with your analysis but didn't see any discussion of this in the paper
<8> yet they spent a good deal of time talking about image transforms
<8> ah well
<5> these guys aint gods. it's likely not perfect. just an example of how potent it can be
<5> the car drove 3000miles on its own on public highways
<8> my point is not that they are imperfect, my point is that they seem to have ***umed that the design characteristics of their network are not interesting to the reader
<8> this seems to me an odd choice in the context of a paper in the context of using neural networks to achieve a goal
<5> you whine over what are more or less trivial details to someone with experience in using NNs
<8> i'm not sure i care for your tone in calling me a whiner
<5> a paper will not necessarily cover details that any background would want to know
<8> but i can accept your point
<8> i'm not claiming expertise
<5> what you should be saying thought is holy cow, is it really that simple
<5> -t
<5> size of net is from trial and error and experience. deal with it
<8> Well you seem fairly convinced of your pov so lets leave it there. thanks.
<5> np
Return to
#ai
or
Go to some related
logs:
default kernel path ubuntu
sibikos
ubuntu switch resolution
ubuntu This codecs.conf is too old and incompatible with this MPlayer release
#fedora
#web
CMYK-colorspace +gd
xine-lib mov gentoo
#php
sendmail instead of root
|
|