6 ms·

There are infinite worlds to explore

113 points by apsec112 5y ago

uejfiweun 5y ago
Pretty cool stuff - I eagerly await the day when the same generative models can spit out fully explorable 3D environments that I can then explore in VR.
- andybak 5y ago
  Yeah - that's my wish too. I think all the pieces already exist but you'd need a supercomputer to get realtime frame rates.
  - DethNinja 5y ago
    You don’t need a supercomputer. You can actually hook algorithmic art from Houdini to GAN models and bake 3D art for later use. Problem is that GAN models are still not good enough to produce human level realistic images. It can generate trippy images but these are not very useful for commercial products. We still need better models or a completely different approach for computer generated realistic art.
    - ShamelessC 5y ago
      Diffusion models are better - but their inference times are orders of magnitude slower. GAN models are really cool. Don't get me wrong. But they are popular _because_ they are fast at inference. They are still challenging to train (hopefully projected GAN helps with this) and have a very tough time representing highly diverse datasets with lots of data.
    - andybak 5y ago
      > You can actually hook algorithmic art from Houdini to GAN models and bake 3D art for later use. I've seen some stuff like that but it's not really what I'm talking about. I want fully dynamic, spatially consistent generative worlds. And that is pretty hard to do at 75 or so FPS in stereo.
PeterisP 5y ago
Really interesting work (and the discussion it), and very nice visuals. Thank you!
zwass 5y ago
This is some of the most compelling algorithmic art I've experienced. I hope the artist (guide?) continues.
mrjangles 5y ago
I remember back in the old days we all thought that it wouldn't be long before computers and artificial intelligence would learn to do advanced science and reason about the world. But, we thought that things like art, that involve some unknown hidden 'human' or spiritual element, would be something machines might never be able to deliver. I find it remarkable that it turned out to be the other way around. Turns out machines can do amazing art very easily. They are still nowhere near being able to replicate human intelligence though. I guess it is because art and vision occur at such a deep level we aren't aware of it, so we used to not really understand how it worked. It's quite hard to build a machine that does something when you don't understand what it is. We built them anyway because apparently even though we didn't understand it, it wasn't very complicated compared to other tasks that we perform.
- jjj123 5y ago
  I have seen a ton of AI art, a lot of it impressive, but I’ve never seen a piece generated by a computer that had any significant meaning behind it that the human artist didn’t give to it. So AI hasn’t solved art at all really.
  - beaconstudios 5y ago
    Yeah AI can only really mass-produce a hardcoded aesthetic right now. There's much more to art than that.
  - stevenhuang 5y ago
    Are you sure that knowing it was generated did not affect your impression? If you were told instead it was made by a human, you'd likely reach harder for a meaning without knowing it, and find it. Meaning is subjective enough as it is.
    - pasquinelli 5y ago
      i think that's the point. who cares about generated art? the point of art is that a human being is saying something to you, the human audience.
jabthedang 5y ago
what a mess. lightyears away from a deeply composed masterpiece. (fawn on, and keep those downvotes coming!)
moultano 5y ago
Author here. Glad you all are enjoying this. It was a lot of fun to make. You might also like my previous piece that had an accompanying narrative. https://moultano.wordpress.com/2021/07/20/tour-of-the-sacred-library/ https://moultano.wordpress.com/2021/07/20/tour-of-the-sacred...
- dusted 5y ago
  Hi! Thanks for making these! They're brilliant! I enjoyed your "Tour of the Sacred Library" a lot! At first, I was just reading a nicely written, very "ambiance heavy" story, then little by little, I realized that the accompanying images were a bit "off", and finally that they were generated. Very inspiring experience.
hazbot 5y ago
I found the generation technique fascinating, but the images felt 'wrong' - I felt like I didn't know where to look to get an overview of the scene, and when I tried to focus on details it was like me eyes kept 'slipping off' because nothing would resolve itself.
pontifier 5y ago
I'm most intrigued by the use of manual modification of the images. The hybrid approach amazed me where the author entered the loop to erase and blur one of the images that the AI then expanded on. This style of interaction actually fits very well with my current theory of consciousness. It allows "echoing" of ideas back and forth with the AI. Repeated echoes allow an idea to form using the best inputs from all involved.
indigochill 5y ago
> Wish I didn’t have that shadowy figure there. He shows up a lot for some reason. Is this leading up to "The Ring: Revenge of the AI"? Jokes aside, great images and the technique is interesting as well.
BLKNSLVR 5y ago
Meta commentary: The code that I started from was written by RiversHaveWings based on a method by advadnoun with further modification by Eleiber and Abulafia. Metaverse or otherwise, we're definitely moving towards self-chosen names that only really work in the electronic-life realm. I'm aware this has been a slow shift over decades, but the sentence above is 33% online nomenclature by letters.
- ShamelessC 5y ago
  RiversHaveWings (Katherine Crowson): https://twitter.com/RiversHaveWings https://twitter.com/RiversHaveWings advadvnoun (Ryan Murdock) https://twitter.com/advadnoun https://twitter.com/advadnoun Don't know the other two. The ML art community is weird. They like to cite a lot due to presumably academic backgrounds but no one _really_ wants to be hyper-famous amidst the NFT crowd (recently hacked Banksy, for instance) and so they all use handles pretty frequently.
sisjohn 5y ago
I do really love the art. I wonder if someday, we can walk and explore inside of these artworks.

113 points by apsec112 5y ago

Pretty cool stuff - I eagerly await the day when the same generative models can spit out fully explorable 3D environments that I can then explore in VR.

andybak 5y ago
Yeah - that's my wish too. I think all the pieces already exist but you'd need a supercomputer to get realtime frame rates.
- DethNinja 5y ago
  You don’t need a supercomputer. You can actually hook algorithmic art from Houdini to GAN models and bake 3D art for later use. Problem is that GAN models are still not good enough to produce human level realistic images. It can generate trippy images but these are not very useful for commercial products. We still need better models or a completely different approach for computer generated realistic art.
  - ShamelessC 5y ago
    Diffusion models are better - but their inference times are orders of magnitude slower. GAN models are really cool. Don't get me wrong. But they are popular _because_ they are fast at inference. They are still challenging to train (hopefully projected GAN helps with this) and have a very tough time representing highly diverse datasets with lots of data.
  - andybak 5y ago
    > You can actually hook algorithmic art from Houdini to GAN models and bake 3D art for later use. I've seen some stuff like that but it's not really what I'm talking about. I want fully dynamic, spatially consistent generative worlds. And that is pretty hard to do at 75 or so FPS in stereo.

PeterisP 5y ago

Really interesting work (and the discussion it), and very nice visuals. Thank you!

zwass 5y ago

This is some of the most compelling algorithmic art I've experienced. I hope the artist (guide?) continues.

mrjangles 5y ago

I remember back in the old days we all thought that it wouldn't be long before computers and artificial intelligence would learn to do advanced science and reason about the world. But, we thought that things like art, that involve some unknown hidden 'human' or spiritual element, would be something machines might never be able to deliver. I find it remarkable that it turned out to be the other way around. Turns out machines can do amazing art very easily. They are still nowhere near being able to replicate human intelligence though. I guess it is because art and vision occur at such a deep level we aren't aware of it, so we used to not really understand how it worked. It's quite hard to build a machine that does something when you don't understand what it is. We built them anyway because apparently even though we didn't understand it, it wasn't very complicated compared to other tasks that we perform.

jjj123 5y ago
I have seen a ton of AI art, a lot of it impressive, but I’ve never seen a piece generated by a computer that had any significant meaning behind it that the human artist didn’t give to it. So AI hasn’t solved art at all really.
- beaconstudios 5y ago
  Yeah AI can only really mass-produce a hardcoded aesthetic right now. There's much more to art than that.
- stevenhuang 5y ago
  Are you sure that knowing it was generated did not affect your impression? If you were told instead it was made by a human, you'd likely reach harder for a meaning without knowing it, and find it. Meaning is subjective enough as it is.
  - pasquinelli 5y ago
    i think that's the point. who cares about generated art? the point of art is that a human being is saying something to you, the human audience.

jabthedang 5y ago

what a mess. lightyears away from a deeply composed masterpiece. (fawn on, and keep those downvotes coming!)

moultano 5y ago

Author here. Glad you all are enjoying this. It was a lot of fun to make. You might also like my previous piece that had an accompanying narrative. https://moultano.wordpress.com/2021/07/20/tour-of-the-sacred-library/ https://moultano.wordpress.com/2021/07/20/tour-of-the-sacred...

dusted 5y ago
Hi! Thanks for making these! They're brilliant! I enjoyed your "Tour of the Sacred Library" a lot! At first, I was just reading a nicely written, very "ambiance heavy" story, then little by little, I realized that the accompanying images were a bit "off", and finally that they were generated. Very inspiring experience.

hazbot 5y ago

I found the generation technique fascinating, but the images felt 'wrong' - I felt like I didn't know where to look to get an overview of the scene, and when I tried to focus on details it was like me eyes kept 'slipping off' because nothing would resolve itself.

pontifier 5y ago

I'm most intrigued by the use of manual modification of the images. The hybrid approach amazed me where the author entered the loop to erase and blur one of the images that the AI then expanded on. This style of interaction actually fits very well with my current theory of consciousness. It allows "echoing" of ideas back and forth with the AI. Repeated echoes allow an idea to form using the best inputs from all involved.

indigochill 5y ago

> Wish I didn’t have that shadowy figure there. He shows up a lot for some reason. Is this leading up to "The Ring: Revenge of the AI"? Jokes aside, great images and the technique is interesting as well.

BLKNSLVR 5y ago

Meta commentary: The code that I started from was written by RiversHaveWings based on a method by advadnoun with further modification by Eleiber and Abulafia. Metaverse or otherwise, we're definitely moving towards self-chosen names that only really work in the electronic-life realm. I'm aware this has been a slow shift over decades, but the sentence above is 33% online nomenclature by letters.

ShamelessC 5y ago
RiversHaveWings (Katherine Crowson): https://twitter.com/RiversHaveWings https://twitter.com/RiversHaveWings advadvnoun (Ryan Murdock) https://twitter.com/advadnoun https://twitter.com/advadnoun Don't know the other two. The ML art community is weird. They like to cite a lot due to presumably academic backgrounds but no one _really_ wants to be hyper-famous amidst the NFT crowd (recently hacked Banksy, for instance) and so they all use handles pretty frequently.

sisjohn 5y ago

I do really love the art. I wonder if someday, we can walk and explore inside of these artworks.