I see deep dreaming as a new, big and growing thing. I am surrounded by people who not only want to try it out but have the capacity to do it. As in the programming ability to fiddle. That puts me in a uniquely exciting place to see several deep dreaming experiments popping up all over facebook. My question is—where is it going? What is it leading to?
Do you have a good prediction of what comes next out of deep dreaming RNN systems? I believe its something big, and with the confidence and ability to predict it; it might be possible to see into the future, ride the wave or be ahead of the curve. (basically I chase opportunities). If there were a time to change the world; new technology would be assisting in making it happen.
Either:
.1. people play around with it and decide it has no real-world applications
or
.2. shit gets real and the world changes really quickly.
Ideas?
Edit: will settle for; “wouldn’t it be great if deep dreaming could do X”
I have given a lot of thought to this subject before, but I didn’t think the technology was good enough yet. Now with deep dream and facebook’s Eyescream, DRAW, and Google’s DeepStereo, generated images are becoming practical.
Someday we will have something like a totally automated photoshop. You can just ask it to make a photo, and it will produce it. Then you can tell it how to tweak it. Or it can give different versions and you can pick which you like most, and repeat (see http://picbreeder.org/, which also uses neural networks, but imagine it has an actual model of natural images instead of starting from scratch.)
It need not be limited to static images and could eventually do videos.
Generated porn that meets everyone’s weird fetish is the most obvious application.
I also think colorizing black and white photos will become a popular niche application. And all sorts of other simple photo enhancements and tools.
These are very good ideas:
Things that I half knew already but didn’t really admit: artwork generation and it’s viability for sale.
Thing that I hadn’t extrapolated to yet: video generation.
neat application: pornography—neat! I hear it makes lots of money!
Colourizing already exists, but its not that great/we don’t care enough to have seen it take off.
I am looking into artwork generation to see where it can go.
According to my reading of the license it should be fine to sell images (generate an artwork and sell it); as long as you don’t try to sell the code to anyone.
According to my reading of the license it should be fine to sell images (generate an artwork and sell it); as long as you don’t try to sell the code to anyone.
Even then, Google published enough research papers on the topic that it’s trivial to reverse-engineer and write your own program to do the same thing—in fact, several people have done it already.
I see deep dreaming as a new, big and growing thing. I am surrounded by people who not only want to try it out but have the capacity to do it. As in the programming ability to fiddle. That puts me in a uniquely exciting place to see several deep dreaming experiments popping up all over facebook. My question is—where is it going? What is it leading to?
Do you have a good prediction of what comes next out of deep dreaming RNN systems? I believe its something big, and with the confidence and ability to predict it; it might be possible to see into the future, ride the wave or be ahead of the curve. (basically I chase opportunities). If there were a time to change the world; new technology would be assisting in making it happen.
Either:
.1. people play around with it and decide it has no real-world applications
or
.2. shit gets real and the world changes really quickly.
Ideas?
Edit: will settle for; “wouldn’t it be great if deep dreaming could do X”
See what Kaj_Sotala posted above: http://kajsotala.fi/2015/07/deepdream-today-psychedelic-images-tomorrow-unemployed-artists/
I have given a lot of thought to this subject before, but I didn’t think the technology was good enough yet. Now with deep dream and facebook’s Eyescream, DRAW, and Google’s DeepStereo, generated images are becoming practical.
Someday we will have something like a totally automated photoshop. You can just ask it to make a photo, and it will produce it. Then you can tell it how to tweak it. Or it can give different versions and you can pick which you like most, and repeat (see http://picbreeder.org/, which also uses neural networks, but imagine it has an actual model of natural images instead of starting from scratch.)
It need not be limited to static images and could eventually do videos.
Generated porn that meets everyone’s weird fetish is the most obvious application.
I also think colorizing black and white photos will become a popular niche application. And all sorts of other simple photo enhancements and tools.
These are very good ideas: Things that I half knew already but didn’t really admit: artwork generation and it’s viability for sale. Thing that I hadn’t extrapolated to yet: video generation. neat application: pornography—neat! I hear it makes lots of money!
Colourizing already exists, but its not that great/we don’t care enough to have seen it take off.
I am looking into artwork generation to see where it can go.
According to my reading of the license it should be fine to sell images (generate an artwork and sell it); as long as you don’t try to sell the code to anyone.
Even then, Google published enough research papers on the topic that it’s trivial to reverse-engineer and write your own program to do the same thing—in fact, several people have done it already.
Thanks! Thats a good point.