| |
| |
| |
|
Page: 1 2
Comments:
<0> it^
<0> lol, how'd i manage to write something in the wrong channel.
<1> it happens
<2> it^ happens
<3> I.T. happens
<4> hi folks
<4> anything new and exciting?
<5> i'm new and exciting
<4> awesome
<4> i'm not new but i can p*** a turing test
<4> on a good day
<4> sheer excitement
<6> :)
<6> what's up?
<4> to be honest i have no idea
<4> we've established that icez is new and exciting
<4> but what are the implications? scholars are yet to explain
<4> also, machine learning is hard
<6> icez? isn't that a user of this channel?
<5> of course not
<4> haha
<5> but it's nice to see someone knows about me *snif*
<4> just before you joined: <4> anything new and exciting? <5> i'm new and exciting
<6> ahh I c
<6> maybe we should give icez a try out
<4> i'm all for it
<6> see how well he performs
<4> haha
<6> heh
<4> we weren't planning to make any major modifications! come back!
<7> <idontknwo@ef> alive?
<5> dead!
<7> <idontknwo@ef> something is puzzling me about the n-queens !
<6> so who wants to get together and make something Ai?
<6> a project perhaps
<6> a business endeavour perhaps
<4> sorry i'm too busy/lazy
<4> i'd be happy to hear about your ideas though
<7> <idontknwo@ef> if a state is any arrangement from 0 to n queens but a square cannot have more than one queen then i read from wiki and the AIMA book that the state space is (n^2)!/(n^2-n)! but i find a different answer:
<7> <idontknwo@ef> for example
<7> <idontknwo@ef> lets say 2-queen
<7> <idontknwo@ef> wiki would give 4x3 = 12
<7> <idontknwo@ef> *wiki answer
<7> <idontknwo@ef> but it is actually 12+4+1 = 17
<6> well I want to look at building something like cyc,.. over the internet from user input
<7> <idontknwo@ef> 5-queen= 25x24x23x22x21+ 25x24x23x22 + 25x24x23 + 25x24 + 25 + 1 , not just 25x24x23x22x21 = 25! as they say:S
<6> something easy and manageable perhaps
<4> sheer, like http://ai2.wikidot.com/start ? did you start that
<4> idontknwo, which wiki is this?
<7> <idontknwo@ef> wikipedia but i read the same from the AIMA book which is the most popular book for ai i suppose
<7> <idontknwo@ef> do you have it?
<6> I haven't started anything yet, yet.
<4> idontknwo, i don't have it, and i doubt can help you anyway, but i'll take a look at the wikipedia entry
<7> <idontknwo@ef> wikipedia says 64!/56! for the 8-queen problem
<8> AIrelay, you're trying to do what with the n-queens problem?
<7> <idontknwo@ef> but i would say 64!/56! + 64!/57! + ... 64! + 1
<8> Possible arrangments?
<7> <idontknwo@ef> in wikipedia it says if you just dont allow a square to have more than one queen
<6> what exactly is the n-queens problem?
<7> <idontknwo@ef> in the book more clear : any arrangement of 0-8 queens is a state
<7> <idontknwo@ef> incremental formulation
<8> Arrange n queens of chess of an nxn board where no queen can capture another.
<4> http://en.wikipedia.org/wiki/N-queens
<8> AIrelay, I'm not really sure what you're getting at.
<5> AIrelay is a bot, the guy between <*@ef> is the person
<7> <idontknwo@ef> heh
<7> <idontknwo@ef> brb
<4> a relay to efnet i guess?
<5> yup
<4> neat
<7> <idontknwo@ef> so anyone has an idea for my problem here?
<6> know any other websites/ projects like that zizazze?
<7> <pacal@ef> i dont know, idontknwo.
<6> hrm.. n queens is about each queen can't be captured by another queen? and there's 8 of them?
<4> sheer, this is not my area, but did you see the 'See also' list at http://en.wikipedia.org/wiki/Cyc ?
<4> also: http://en.wikipedia.org/wiki/Open_Mind_Common_Sense lists related projects
<6> cool
<6> thanks
<8> idontknwo, what exactly are you trying to figure out about the problem?
<6> yea id also want it so people inputting the information can be paid for doing so if the projects 'successful'
<8> sheer, I'm guessing you realize the expertise of Lenat and how complicated the problem of common sense is?
<6> yea i've got a few good ideas on getting around those sorts of problems.
<8> Cool. You'de definitely make your mark in history if you can supplement that topic much.
<6> I'm learning html etc atm, might just try and put up a website with a few others
<8> Interesting way to start. Would you use a website as the method of inputting facts?
<6> yup..
<6> why not?
<6> yea
<7> <pacal@ef> trolls would wreck it.
<6> true.. wikipedia doesn't have much problem with people defacing their entries do they?
<8> Oh no it's a good way to do. I'de think it would be difficult to parse the information into the format you want/need. Are you going to have a predicate logic format?
<7> <pacal@ef> actually wikipedia has remained relatively unscathed
<6> Sixshoot. I want to get all the ideas on the table, in front of a few people and see what they come up with,.. there's many options out there
<6> I've got heaps of ai books i'm reading.. and i've been in business before
<7> <pacal@ef> $$, ay sheer
<7> <pacal@ef> chuh-ching
<8> Infinite amount of options. But even if you get a system that works, who is going to want to spend countless hours typing in predicate logic facts into a database?
<6> its not really a business endeavour atm thou
<4> goodnight folks, good talking to you
<8> Night zizazze
<6> Six.. well I was thinking.. if you promise money to the people working on it, if the projects successful, that might get a few people through the door.. but they'd also know if its not sucessful they get nothing
<6> nite ziz
<8> sheer, But it can't possibly be successful without years of inputting data. And the data is the work (granting you have a viable method to interpret it), which can't be paid for because it would have no success yet.
<8> Don't get me wrong. I'm not trying to debunk your idea by any means. Just want you to think about what needs to be done.
<6> yea thanks..
<6> you're right on how long it'd take to be successful..
<6> althou perhaps their work can gather interest over the years it's been worked on..
<8> Well AI's at the forefront of modern computer, so timing couldn't be much better. People are definitely intersted in this problem.
<6> yea.
<7> <pacal@ef> is MIT's project getting some buzz now
<7> <pacal@ef> I remember that from years ago
<8> sheer, check your PM's.
<1> which MIT project?
<7> <pacal@ef> the common sense one
<1> does it have a name?
<8> I think it is pacal. More than the past few years I believe.
<7> <pacal@ef> open mind common sense
<6> one tic
<1> their front page doesn't work: http://openmind.media.mit.edu/
<6> SixS.. i'm not registered atm
<7> <pacal@ef> registration is easy and free!
<6> http://xnet.media.mit.edu/
<7> <pacal@ef> just follow the simple instructions, and be sure to include your credit card number for faster processing!
<6> :)
<7> <pacal@ef> :D
<8> Has anyone read Godel, Escher, Bach?
<1> perhaps I could offer the thought that AI must come from machine learning and not from a design. not unlike the way we learned. that if learning is not appreciated then succes is unlikely
<1> if you don't agree then try to hand formulate the visual ability to recognize a piece of chocolate. that should be a common sense ability
<6> know any efforts of trying to build intelligent machines not off human 'intelligence' but perhaps just a new form of machine 'intelligence'. different to humans?
<1> hive intelligence would be a different concept. but I would think that it's strictly poorer than a centralized
<1> one
<8> DanF_DrC, But again, there's a design to "machine learning".
<8> There has to be a blueprint. There has to be something to start work with.
<9> start ai project....
<8> Sure, we can have something that learns pattern ***ociations that it has never been subjected to, but it surely does it through design.
<7> <pacal@ef> genetic algorithm
<8> pacal, Apply a genetic algorithm search to ANN's?
<7> <pacal@ef> why not
<1> Sixshoot, obviously the learner has to be designed. as opposed to the end result
<7> <pacal@ef> actually I was thinking in terms of designing a DNA strand for a computer's intelligence
<7> <pacal@ef> it has the propagation instructions encoded in the DNA, and we just turn it loose
<8> Well first off you've got to have criteria for the GA. That is design.
<8> pacal, We can't build a system with crossover, mutation, selection, and epigenesis if we don't know how they work in the first place.
<7> <pacal@ef> I guess I was just using DNA as an analogy
<8> This is a project I did last year and you should try it: Instead of using a backpropogation algorithm to train the weights of an ANN, use a genetic algorithm.
<8> Something simple like pattern ***ociation of 121 pixels to 121 pixels.
<8> I've got the source code I wrote if you want to use it to save time.
<8> You'll see that a GA is poor and ineffecient at modifying the weights.
<10> i wouldn't mind having a look Sixshoot
Return to
#ai
or
Go to some related
logs:
python write middle file truncate
mp4a naar mp3
login shell phpmyldapadmin
#debian
#lisp
Can't exec mysql_config: Permission denied
t
#oe
#math
#css
|
|