I dont think thats an absolute.
But you do underestimate my power. I dont care if you have the high ground
3 Likes
I agree on everything except the common/uncommon. This is where a lot of artists really shine and can truly express themselves. A ton of beautiful artworks are from card costing 0.10 $.
It would be so sad seeing pokemon giving up on them, and I hope I’m not the only one thinking this lol.
3 Likes
I would say there are quite a few fallacies in here. The primary one is anthropocentric. There is nothing special about the human brain or human creativity. Compared to animals, sure we are exceptional. But there is no process happening in our brain that can’t be replicated in a computer.
Compared to current image generation models our brain does work differently. We replicate 3d symbols and constructs from our brain into art. Whereas modern models are basically trying to predict a group of pixels based on a prompt and the pixels around it. It’s basically approximating something real looking by averaging a huge database of images. This mismatch in the way we “think” is why the details in AI images today are often nonsensical to us.
But in a more abstract perspective we are both just leveraging patterns we have learned and stored in memory. There’s a sentiment in this thread that AI can’t generate something novel. Depending on how you define it, neither can humans really. We also don’t know the limitations of neural networks in this case. If you never trained an AI to learn what “cubism” is, could you approximate an image in that style? I don’t see why not.
The second fallacy is the implicit assumption that AI won’t get better. We went from cracked-out blurry images to ones that trick people as photorealistic basically in one year. Eventually we could get to a point where drawing something yourself is extremely cumbersome and unnecessary. Like trying to calculate something on paper instead of using a calculator. Perhaps a bit of a sad reality
13 Likes
I can be pretty quick, days/hours. But also, the model doesn’t need to generate the high-quality end products. The true art application is to spark ideas and save time.
I’ve done this with my own work where I train an AI on my paintings, then had it generate news ones that I review for ideas and inspiration, then I paint new paintings off of those ideas, those new paintings get added to the model…and so on.
For example, Pokemon could create an AI based on Komiya. Komiya could then transfer their style to the Base Charizard and generate 10,000 images in minutes. They would review them all for inspiration, ideas, structure, colors…then paint their own version, instead of starting with blank piece of paper. The time savings and the creativity this would unlock would give us more and better Komiya!
3 Likes
100% sure that ai will get better and better, and also agree that there is nothing magical behind human brain. What make me think that mediocrity will win is that corporates will be using it.
I believe that the results will be much different from an enthusiastic skilled artist carefully prompting and selecting generated artworks in his bedroom, finely postprocessing them in photoshop.
Unfortunately I think that ai will be used to maximize profit, and produce tons and tons of “well enough” content, leading to stagnation.
1 Like
We are also just animals, never forget
2 Likes
Depends on if the Pokémon has hands or not. Aipom is safe.
3 Likes
The brain is just essentially a very complex model that theoretically should be re-creatable. But it is still special because we don’t know how it works and we can’t replicate it in a computer yet, if ever. Right now, the models just approximate our outputs like images and text, but they can’t approximate the human brain yet because we don’t have direct access. That’s where our spark of creativity lives and its still safe for now. Our words, art, creations, and Pokémon cards are intermediary abstraction layers between us, computers, and other humans; but they aren’t “us.”
AI will get better. But the current models aren’t anything that wasn’t predicted 10+ years ago. We knew we could do it; we just needed the tech/resources/data. Similar to putting people on Mars. We know we can do it. However full AI represents going to the next solar system. We have no idea when/if we can. These models are cool, but they still only represent <1% of human intelligence. The current AI methods are limited and will plateau without a breakthrough.
6 Likes
I agree that this is one possible option but I don’t agree its the only one. I’m not sure we fully understand the limitations of nn type models. Mind you, I haven’t been actively looking into the research for a few years now
Maybe AI will be so advanced that it will be indistinguishable from a human creation, it just isn’t yet. AI art right now feels gimmicky.
I think the value of art is the perspective shared in the piece. It could be aesthetically moving, or socially relevant. It sounds like ai art is still in the phase where you need to prompt it to create. I think I’ll subscribe to the idea that human brains aren’t unique when ai is creating art on its own. But at that point Skynet will probably run everything. 
4 Likes
Maybe I have an unpopular opinion, but I welcome the use of AI in art. I don’t see it as a replacement for artists, but as a tool at their disposal.
Regarding whether AI will get better: yes. We’ve seen substantial increases in the application of machine learning and deep learning algorithms in research and medicine. There is no reason to believe that these algorithms will not become more sophisticated for the art world.
1 Like
The public tools that are widely available are pretty gimmicky. However custom trained models that are targeted can be very convincing. People have mistaken my AI art for human art many times and said how they were “moved by it.” However its abstract, so easier to seem “human” when its blobs!
3 Likes
I’m not fully convinced by the “tool” argument honestly. Sure I can imagine well know artists, art director, big names like Komiya and Arita using their model to speed/improve workflow.
But I don’t think that this luxury would be granted to new artists trying to establish themselves in an ai market. Maybe 1/1000 would become a senior and succeed, but there is simply no need of that many artist when real people is called to just do minor fixes
It’s the data and the evaluation. NN are inefficient with their data compared to humans, so much so we are running out of training sets in many areas. A human needs 2-3 pictures to understand a boat whereas a NN needs hundreds. The other problem is the evaluations. We can provide human feedback to models if the output is short text or images, like GPT4. But how do you provide human feedback on an AI novel efficiently?
NNs are like combustion engines, for the past 20 years we’ve just been making it bigger and adding more cylinders to make it go faster, but we’ll need a Tesla(new methods) to come along to make the next leap.
1 Like
I could see artists using AI to generate the backgrounds if the Pokemon is just doing a generic pose. The Sugimori cards in Base and the Gym sets would be an example.
1 Like
I don’t see the Pokémon Company ever using AI to replace artists. I could see the artists themselves using AI tools to plus their work. And we’d likely never know. Not that it really even matters that much. But…
The easiest way for artists to combat AI imagery is to state that “Using AI is like buying fake diamonds.” AI images will have their place, just like fake diamonds. But it also isn’t the most satisfying thing in the world to tell someone your jewelry is made with fake diamonds. 
2 Likes
I personally feel myself pulling away from this incessant need to automate everything. AI art kinda shits on much of what makes art special, in my opinion. Removing all of the nuance from the process, and the personality in one’s style, just to jump to the end product? Everything seems so focused on optimization and speed. For what? Until AI become self-aware and start creating their own art about the AI experience (lol), it’s just empty mimicry as far as im concerned.
That may be a boomer-doomer take, but i just dont like it. Makes the hair on the back of my neck stand up.
4 Likes
Some people, more than others 
1 Like
I agree with the general point about efficiency and also the general point about needing different models. The analogy to the combustion engine is solid. BUT what I will say is your example I quote here is cheating a little bit. For a human to extrapolate what a boat is from 2-3 picture, it takes 6+ years of constant training in order to get to a point where you are able to do this. It also assumes the surrounding context of the boat is also understood (what is water, some understanding of buoyancy, understanding of different materials and textures, etc, etc).
The brain is also super biased and heuristic. The false positive rate for recognizing faces in non-faces is embarrassingly high. It makes me wonder what kind of trade-offs we are tolerating in our brain in order to be extremely good at classifying objects
1 Like