popcar2

downvotes come at a “cost”, whereby if you want to downvote someone you have to reply directly to them with some justification, say minimum number of characters, words, etc.

I think it's the complete opposite. Platforms with downvotes tend to be less toxic because you don't have to reply to insane people to tell them they're wrong, whereas platforms like Twitter get really toxic because you only see the likes, so people tend to get into fights and "ratio" them which actually increases the attention they get and spreads their message to other people.

In general, platforms without upvotes/downvotes tend to be the most toxic imo. Platforms like old-school forums and 4chan are a complete mess because low-effort troll content is as loud as high effort thoughtful ones. It takes one person to de-rail a conversation and get people to fight about something else, but with downvotes included you just lower their visibility. It's basically crowdsourced moderation, and it works relatively well.

As for ways to reduce toxicity, shrug. Moderation is the only thing that really stops it but if you moderate too much then you'll be called out for censoring people too much, and telling them not to get mad is just not going to happen.

My idea for less toxicity is having better filtering options for things people want to see. Upon joining a platform it would give easy options to filter out communities that are political or controversial. That's what I'm doing on Lemmy, I'm here for entertainment, not arguing.

Oh nice, do you have a link for where it was posted?

Technically you're right but the thing about AI image generators is that they make it really easy to mass-produce results. Each one I used in the survey took me only a few minutes, if that. Some images like the cat ones came out great in the first try. If someone wants to curate AI images, it takes little effort.

God DAMN it

Are there any statistically significant differences between the different generators?

Every image was created by DALL-E 3 except for one. I honestly got lazy so there isn't much data there. I would say DALL-E is much better in creating stylistic art but Midjourney is better at realism.

Someone asked for the raw dataset in the other thread, here it is: https://docs.google.com/spreadsheets/d/1MkuZG2MiGj-77PGkuCAM3Btb1_Lb4TFEx8tTZKiOoYI

Have at it.

I feel for you. A few people said the human art I put in the survey were lackluster but I thought they were pretty good, not everyone is an S-tier artist.

77% of people guessed this was AI generated, and a friend of mine kept saying it was weird and inconsistent so "I doubt a real artist would put random food in the back"

It's actually a cropped image of https://www.deviantart.com/tsaoshin/art/Strawberry-Taiyaki-Cat-905271835 . I wouldn't want to be an artist right now.

No, the AI didn't try to copy the other art that was included. I also don't train the model myself, I just tell it to create an image similar to another one. For example the fourth picture I told it to create a rough sketch of a person sitting on a bench using an ink pen, then I went online and looked for a human-made one that's of a similar style.

Done, column B in the second sheet contains the answers (Yes are AI generated, No aren't)

Absolutely horrifying, thanks

No idea, but I would assume most results are from here since Lemmy is where I got the most attention and feedback.

I would say so, but the sample size isn't big enough to be sure of it.