[NetBehaviour] Quick thought about Neural Networks
xchicago at gmail.com
Sun Mar 25 06:20:34 CEST 2018
Been building some Neural Networks.
There are two ways you can build them...
Analytical or Creative.
An Analytical NN takes a lot of data
and boils it down to an answer.
This is what the business world is so on about.
It's a replacment for anything that was formerly
random decisions, like which banner ad to display.
You can also make a Creative NN,
which taks a small amount of data
and makes up a whole bunch.
Examples of Creative NNs:
1. You press a button, and the NN creates an image. You tell it whether you
liked the image or not.
2. You select 1 of 4 colors, the NN writes a poem for you. You tell it
whether you liked the poem or not.
Techniques for Creative NNs:
(1) NN's only do what you train them to do. If you include the previous
answer as part of the input data, then you can train it to not give the
same answer twice.
(2) The higher you set the learning rate, the more "in the moment" it
Any ideas, thoughts, comments? Is there room for creative NN artworks?
Seems like this kind of thing could take over the interactive installation
space. Maybe art-tech can become surprising again?
If they make manipulation machines that drag on your worst fears, maybe we
can make happiness bots that make you fearless?
Any ideas, thoughts, comments? What if we made the entire museum into
a rube goldberg of intelligent automation? Anything else we could do?
Skynet's the limit.
((º Ω º))
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the NetBehaviour