4 ms·

These tools are only weeks or months old. At the moment they’re a novelty. ‘A picture of a dog riding a skateboard by James Gurney | Trending on Artstation’ is

These tools are only weeks or months old. At the moment they’re a novelty. ‘A picture of a dog riding a skateboard by James Gurney | Trending on Artstation’ is clever and astonishing, yes, but we’ve nowhere near scraped the surface of what they can do or how they can be used as part of the creative process. It’s going to take time and a good few rounds of development and pushback, but eventually we’ll wonder how anyone did anything without it. Current trends in AI art are pretty boring, but once we properly get the hang of it it’s going to be fascinating. Reminds me a lot of the early days of Photoshop - adding a filter was cool and novel so most people added a filter and called it a day. But over time people have figured out how to do all kinds of creative things with it. As an artist, AI tools are like having a paintbrush that’s possessed by a drunk genie and that’s very exciting.

mkaic 4y ago
This 1000%. I'm currently working on my own slightly more complex/customizable interface for Stable Diffusion because I firmly believe that the value lies not in simple prepackaged text-to-image, but in the wealth of information, structure, and hierarchy that is compressed into its 4GB latent space. SD encodes only 4GB of information, but that information is so highly structured and compressed that it represents a sizable chunk of all images on the internet. The true value Stable Diffusion brings to the table is the sheer level of semantic meaning encoded into the vector spaces it samples from to generate images. The fact that it's open-source means anyone can write meta-algorithms to walk through and explore that latent space, too, and as Stability and other companies start releasing better and better models, the latent spaces of those models will become less and less entangled and more and more meaningful and—— We're on an exponential curve and it's so exciting.
Bakary 4y ago
Using any Photoshop tool still requires some work. AI is an order of magnitude more multifaceted and autonomous. What happens when we realize that AI art generators consistently and instantly bring forth better contributions than what we can make ourselves over a much larger period of time? The artist will look at their paintbrush and wonder what the point of holding it is when the drunk genie is doing much better on their own. There will still be artists, but there will be space only for the most exceptional of them. Art software is similar to cruise control. AI art will in a few years become the equivalent of a fully perfected autonomous car that is safer than a human driver in all but the most contrived scenarios.
- mattkevan 4y ago
  My phone camera can take much better pictures of things than I can draw. Doesn’t mean that my drawings have been invalidated. Human art is going to adapt, similarly to how it adapted to photography. If a camera can make perfect reproductions of reality, AI is a camera of the imagination. I wonder if there’s going to be a trend towards craft -tangible, clearly handmade things that an AI can’t do.
  - Bakary 4y ago
    Photography is a distinct medium with its own expectations. People looking for 2d non-photography art generally aren't looking for a perfect representation of reality, so photography does not supersede what they want to see. In fact photo-realistic drawings are viewed as a bit of a meme, especially if it's a picture of some pop-culture character like the Joker. It's worth asking at this point what exactly the adaptation will look like. I definitely agree that there will be more interest in non-2d, non virtual art. But for 2d/virtual artists? We need more than just an "underpants gnome" sequence where adaptation is considered a given. The work of those artists won't be invalidated in a philosophical sense, but the competitiveness of the space will skyrocket. The capacity to live as an artist will be much more restricted and it's already a struggle to begin with.

mkaic 4y ago

This 1000%. I'm currently working on my own slightly more complex/customizable interface for Stable Diffusion because I firmly believe that the value lies not in simple prepackaged text-to-image, but in the wealth of information, structure, and hierarchy that is compressed into its 4GB latent space. SD encodes only 4GB of information, but that information is so highly structured and compressed that it represents a sizable chunk of all images on the internet. The true value Stable Diffusion brings to the table is the sheer level of semantic meaning encoded into the vector spaces it samples from to generate images. The fact that it's open-source means anyone can write meta-algorithms to walk through and explore that latent space, too, and as Stability and other companies start releasing better and better models, the latent spaces of those models will become less and less entangled and more and more meaningful and—— We're on an exponential curve and it's so exciting.

Bakary 4y ago

Using any Photoshop tool still requires some work. AI is an order of magnitude more multifaceted and autonomous. What happens when we realize that AI art generators consistently and instantly bring forth better contributions than what we can make ourselves over a much larger period of time? The artist will look at their paintbrush and wonder what the point of holding it is when the drunk genie is doing much better on their own. There will still be artists, but there will be space only for the most exceptional of them. Art software is similar to cruise control. AI art will in a few years become the equivalent of a fully perfected autonomous car that is safer than a human driver in all but the most contrived scenarios.

mattkevan 4y ago
My phone camera can take much better pictures of things than I can draw. Doesn’t mean that my drawings have been invalidated. Human art is going to adapt, similarly to how it adapted to photography. If a camera can make perfect reproductions of reality, AI is a camera of the imagination. I wonder if there’s going to be a trend towards craft -tangible, clearly handmade things that an AI can’t do.
- Bakary 4y ago
  Photography is a distinct medium with its own expectations. People looking for 2d non-photography art generally aren't looking for a perfect representation of reality, so photography does not supersede what they want to see. In fact photo-realistic drawings are viewed as a bit of a meme, especially if it's a picture of some pop-culture character like the Joker. It's worth asking at this point what exactly the adaptation will look like. I definitely agree that there will be more interest in non-2d, non virtual art. But for 2d/virtual artists? We need more than just an "underpants gnome" sequence where adaptation is considered a given. The work of those artists won't be invalidated in a philosophical sense, but the competitiveness of the space will skyrocket. The capacity to live as an artist will be much more restricted and it's already a struggle to begin with.