Now they just need to feed it its own source and turn it into Skynet source.
Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
11–20 of 146 posts
Re: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
#12The next step is "turning pencil drawings into photos" and using it to fabricate evidence on grand scale. Why bother catching politicians doing something when you can just draw them in? Will wreak havoc on societies with weak politics/reporting culture.
How long until personal testimony and non-repudiable crypto signatures are the only admissible evidence in court?
Maybe it's the good thing. Law worked like this for ages and in some ways they managed.
Of course it relied heavily on everybody knowing everybody else in much smaller societies, but that we can do also by doing an AR face lookup on every person we ever interact with.
Re: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
#13The next step is "turning pencil drawings into photos" and using it to fabricate evidence on grand scale. Why bother catching politicians doing something when you can just draw them in? Will wreak havoc on societies with weak politics/reporting culture.
This is very impressive tech and has a lot of good applications for movies and art but I'm with you that it does scare me that it will find it's most common use case as a tool for propaganda by governments.
Re: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
#14Some of that functionality looks a lot like Pikazo.
Re: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
#15Very cool. What's the max resolution (on a 12GB GPU)?
Re: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
#16At this point, I'm wondering "What's real any more..."
Seriously, if this continues, I don't know how to keep up with this field. I spend at least an hour a day just reading about the work that has been done (i.e. reading the research).
Re: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
#17I'll believe it if it's still around in 48 hours
Re: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
#18The project page is perhaps a better place to directly link: https://junyanz.github.io/CycleGAN/
Re: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
#19My favorite of the results tended to be [anything] -> Ukiyo-e/Cezanne. I think because these are easier problems, lots of detail to less. The transfiguration and painting -> photo have me firmly in the uncanny valley, but I suspect this harder problem will be solved given more training.
Re: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
#20it feels like there's a new deep learning paper each week, ever so slightly bringing me closer to an existential nervous breakdown.
Yeah, I feel that. I tell myself they're only tools, really no stranger than time frequency domain transformations to someone unfamiliar with the fourier transform.
Yet there really is something disturbing about seeing a computer resurrect so much of the mind of an artist who's been dead for nearly a century. I know the GPU doesn't understand what it's doing, but did Monet?
When I paint I don't really understand. There may be occasional moments of clarity, I like to think that I make deliberate choices based on the emotions and thoughts within me that I'd like to reflect to the people experiencing my art, but... Honestly am I so different from a neural network?
I am a neural network. Much of what goes into creating a piece of art is based on intuition, on experience and practice, pathways eroded into my mind from years of thoughts traversing the same landscape. The art I create is unique to me, not in that it cannot be reproduced but in that every action I take is a reflection and an echo of all the moments remembered and forgotten that create me.
How much of Monet's mind is in a GPU in Berkeley?