Hacker News Reader

Comment by zachmu | Hacker News Reader

What's interesting to me is how quickly AI generated art came to be expected, unexceptional. Every random blog post has a couple AI generated images now, they are completely unremarkable and uninteresting. The novelty wore off basically instantaneously. They're just the new corporate art style.

View on HN

Maybe people have had enough of worrying about copyright infringement, licenses, attribution etc. and just want a quick sketch to get the idea across. Art is a tool for expressing ideas after all.

It’s only been a year. As SORA shows, state of the art images and videos will soon be undetectable by human eye and it’s very soon coming to proliferate every “content” you interact with.

Does it really matter? AI will generate the images or video according to the introduces prompts so the creativity falls in the one entering them, you could be an excellent piano player but that doesn’t make you a composer.

>> What's interesting to me is how quickly AI generated art came to be expected, unexceptional. Every random blog post has a couple AI generated images now, they are completely unremarkable and uninteresting. The novelty wore off basically instantaneously. They're just the new corporate art style.

> It’s only been a year. As SORA shows, state of the art images and videos will soon be undetectable by human eye and it’s very soon coming to proliferate every “content” you interact with.

Even if the "AI" generated images will be "undetectable by human eye," they'll still be at least "completely unremarkable and uninteresting" and more likely crappy and stupid (even if all the people have the right number of fingers now).

I predict the floodgate of that "AI" generated stuff will reduce the production of actually good and creative art and images and make what is produced harder to find.

Sora costs a fortune, like dollars per minute. It's DOA. Also when handpicked demo content shows cats sprouting and losing limbs in a 15 second clip and that's the best they got, not sure anyone has to worry about undetectable fakes, certainly not with the total lack of creativity I've seen on the image side of things.

And generated pictures are easy to identify because hands are all messed up, right?

This is just the first attempt and you're criticising details. Last time they worked them out pretty quickly.

I'm perplexed that there are a significant amount of people on a tech forum that can't do a simple linear extrapolation from last year to one year into the future. That would get you in the right direction, except that we're on an exponential curve.

Somebody post the Disco Stu graph

Not all problems are linear - people have been saying for over a decade now that self driving cars are imminent, but at this point it seems like most have actually given up on them.

Well… it’s possible we’re on an exponential curve there too. The early part of exponential can be pretty slow. And it depends on how you measure progress. We are seeing exponentially more training data collected in terms of miles driven every year. And likely exponentially more processing power applied to the problem space.

The stock photos accompanying every random blog post we had up until now were also completely unremarkable and uninteresting.

I think it is like if you gave everyone spray paint cans and said go paint, you wouldn't expect all the buildings around to have giant Basquiat inspired murals everywhere.

It would just be ugly, shitty graffiti, everywhere.

Reply on news.ycombinator.com