Hacker News Reader

Comment by riidom | Hacker News Reader

I don't like it. Looks as inspiring as when someone accompanies each room description with the result of a quick image search.

View on HN

I like the concept, but on an artistic scale of 1-10, this is like a 1.

More GPU power and the pictures might get better?

I believe a bit of additional static prompt, such as appending "in pixel art style" or "in classic adventure game style" [1] to every description, to make it more consistent would solve the problem.

[1] https://www.giantbomb.com/the-hobbit/3030-9936/images/

The real problem is a lack of permanence. The fact that the house is different from other angles is frankly disappointing. Why even share this.

That is a problem. Maybe using some context :"That house again, but {new situation}"

It's one of those problems looking for a solution situation. "We have this system that can generate images. What do we do with it? Try it everywhere! It's gotta stick somewhere."

After spending some time on midjourney's discord it hit me that for the vast majority of the population, no matter the tools they are given, have absolutely 0 artistic sense.

That's why I don't believe artists will be replaced, the average Joe can't tell a good from a bad painting, or a good from a bad picture, let alone generate it.

Same for me. As once Sheldon from The Big Bang Theory said: the world's most powerful graphics chip - imagination. I guess playing this game (even with new images generated every time you play it) would resemble reading a book after watching a movie, all the characters and scenes during reading would appear as they are in the movie - no place to stretch for imagination.

They should have trained it on the weird traced stock photo illustrations from wikhow.com.

these are weak prompts though

continuity is a solved problem and so is quality and specifics

For real, an intensive dive into the full layout and structure of the game would allow for good quality rewrites of the prompts to maximize results in the image view.

what not training on an artist’s copywritten work does to an AI

Do you agree that one can train a human brain or a digital brain on as much copyrighted artist work as one wants? style is not copyrightable, so is training on copyrighted content.

the law (an common sense) implies that as long as you're not reproducing art, you can do this. and latent space doesn't save or reproduce that content at all. It mimics, combines remixes or reinvents it. The same what a human brain is doing after learning how to draw.

humanity wouldn't be where it is now when it's forbidden to learn from inventions.

Reply on news.ycombinator.com