Sinfest Forum Index Sinfest
welcome to the fest
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

I, for one, will welcome our future robotic overlords
Goto page Previous  1, 2, 3, 4
 
Post new topic   Reply to topic    Sinfest Forum Index -> Casual Chat
View previous topic :: View next topic  
Author Message
Heretical Rants



Joined: 21 Jul 2009
Posts: 4551
Location: No.

PostPosted: Mon Mar 11, 2013 6:15 pm    Post subject: Reply with quote

fritterdonut wrote:
I don't think we need to be overly worried until robots develop proper Pavlovian responses.


You mean like we can already do with neural nets?

I mean, that Pavolivan shit isn't very complicated.

The idea behind this is that we hard-code the neural net with a goal, e.g. win at Pong (only takes about three neurons to play Pong well) or recognize a face in an image uploaded to Facebook, and then we feed it data until it can accomplish the goal - often this means letting the neural net try out different strategies until it finds one or more that work.


This is relevant to my discussion with Snorri: the Pavlovian goals hardwired into animals' brains (get food etc) are not at all the same as the goals we give neural nets.
_________________
butts
Back to top
View user's profile Send private message
bitflipper



Joined: 09 Jul 2011
Posts: 728
Location: Here and Now

PostPosted: Tue Mar 12, 2013 12:52 am    Post subject: Reply with quote

And neural nets behave far more like living brains than do classic von Neumann discrete-state machines. Although one can program a von Neumann machine to model a neural net, actual analog neural nets still tend to outperform their simulations in just about all aspects of intelligence they model: time to achieve goal, runtime performance, flexibility, learning, etc. They even tend to develop more optimized solutions in terms of nodal excitation thresholds and number of nodes necessary for task completion.

Who's got some recent info on genetic or evolutionary neural nets? My information's only from the 90s.
_________________
I am only a somewhat arbitrary sequence of raised and lowered voltages to which your mind insists upon assigning meaning
Back to top
View user's profile Send private message
Heretical Rants



Joined: 21 Jul 2009
Posts: 4551
Location: No.

PostPosted: Tue Mar 12, 2013 1:14 am    Post subject: Reply with quote

bitflipper wrote:
genetic or evolutionary neural nets


I'm actually kind of doubtful of how useful genetic algorithms would be for neural nets. Might work okay for optimizing a few parameters, but other than that...
_________________
butts
Back to top
View user's profile Send private message
bitflipper



Joined: 09 Jul 2011
Posts: 728
Location: Here and Now

PostPosted: Tue Mar 12, 2013 2:07 am    Post subject: Reply with quote

Heretical Rants wrote:
I'm actually kind of doubtful of how useful genetic algorithms would be for neural nets. Might work okay for optimizing a few parameters, but other than that...

I've heard of work using genetic algorithms to design neural nets--essentially, multiple SPICE-like models run in a virtual environment and the ones most successful at accomplishing a list of tasks get cross-bred, then iterate--but I haven't heard of much genetic architecture being implemented on nets, themselves. It's one of the things about which I'm curious to learn more; I'm not even sure how it could be done.
_________________
I am only a somewhat arbitrary sequence of raised and lowered voltages to which your mind insists upon assigning meaning
Back to top
View user's profile Send private message
dazedb42



Joined: 09 Jul 2006
Posts: 2348
Location: Margaret River, Australia

PostPosted: Tue Mar 12, 2013 3:20 am    Post subject: Reply with quote

When Watson gets reloaded with the urban dictionary again and starts his own hip hop act I am officially going to start worrying.
_________________
(_**_) *note to self, insert bottle at other end.
Back to top
View user's profile Send private message
Genius robot



Joined: 28 Jul 2006
Posts: 21
Location: Greenland

PostPosted: Tue Mar 12, 2013 3:48 am    Post subject: Reply with quote

My name is Wat-dogg and my rhymes are very
fresh since I been updated by the Urban Dictionary
But I got some flow not everyone will appreciate
"the man done belch after he done felch" some will hate

Yeah the Urb-D has a million words for poop
and some words you don't know like faloop
so time to get literary and start readin' better
you'll be smarter knowing the Cosby Sweater
Back to top
View user's profile Send private message
dazedb42



Joined: 09 Jul 2006
Posts: 2348
Location: Margaret River, Australia

PostPosted: Tue Mar 12, 2013 5:05 am    Post subject: Reply with quote

hah, everyone knows that Watson is more gansta than that.
_________________
(_**_) *note to self, insert bottle at other end.
Back to top
View user's profile Send private message
Display posts from previous:   
Post new topic   Reply to topic    Sinfest Forum Index -> Casual Chat All times are GMT
Goto page Previous  1, 2, 3, 4
Page 4 of 4

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © 2001, 2005 phpBB Group