Edit: with Flux, you can't even tell the difference: https://blackforestlabs.ai/
Edit: with Flux, you can't even tell the difference: https://blackforestlabs.ai/
Photo is an image but also a record. The fact something really did exist and captured is probably more valuable than ever.
So wedding/event photographer really don’t have to worry about lose their job to AI
But in places where photo, as an image just to express abstract idea, without concerning where and when it happened, then that part of value goes to AI already
"give me an elephant under a tree" "make it later in the afternoon". "not that late". "emphasize eyes just a scooch and make it look sad and pensive". "Not quite that much". "Clouds could be a bit whispier" Like you'd talk to a photographer, but with instant updates and no retorts like "great, are you going to pay me to camp out for days waiting for the clouds to move in the sky and then somehow hoping the elephant revisits this tree?"
Beats scouring a huge catalog (which, sure, will have AI powered search, but still), and suddenly, it isn't stock anymore, it is very particular to your specific needs. Custom to your needs, faster than getting a stock photo, and so, so much cheaper.
There is a difference between out of necessity and out of appreciation. Those are two completely different economies, later IMO, is much less predictable and reliable.
SDXL + LoRA easily eclipses Flux in realism, but prompting is 100% more difficult for complex scenes.
It's pure slop, of the non-AI kind.
Are we talking a human subject? Nature?
I would do even better at this if we limited it to pictures of "realistic" settings.
I'm not even sure how we could implement a real-life test without bias. Maybe if there was a complete feed of your internet browsing, where it asks you at the end of the day "ballpark the % of media that you think was AI?". Then go through the entire feed, and scrutinize it one by one.
Bringing it back to the topic of stock photography: A large percentage of stock photos are of real things, people, scenery. So, when someone says I'll have a hard time spotting generated stock photos, I kinda go uhh, well, no, not generally, because stock photos are very often of people and real life scenes, the thing that is the easiest to spot as a generation.
We still don't know whether or not you're good or bad at picking out AI images used in actual campaigns, because we have every reason to assume at least a reasonable proportion of AI images used in actual ads will have been through an editorial process that'd rule out a lot of the easily recognized shlock, and so a test that does not use images that have been through the same selection process is meaningless.
I have no doubt you can recognize some. You may well be able to recognize all of them perfectly for what I know. The point was not to argue you can't, but that your impression can't reliably tell you, because you'd be likely to think the same whether your accuracy is high or low.
1. This idea that "you can't tell if you mindlessly scroll past it" isn't a very good measurement.
2. Given IG is slowly filling up with AI slop, I actually do spend a decent amount of time going "is this AI-generated".
3. I'm not discrediting "advancement of realism" in AI at all. I'm just saying it's much, much easier to detect AI when a generation is supposed to be of something real.
How many more do you think would get past you if the person running the hypothetical campaign was someone with a similar experience at picking images to you spending the same amount of time they would picking stock photography on ruling out any picture that looks like it's AI-generated to them, or editing them to remove things that'd tip you off?
Maybe it's not rational. Maybe I can't tell the truly good AI images form the cheap slop ones. But that's how I feel, and ultimately a lot of commerce runs off customer feelings. The faker, cheaper, and more soulless we feel a company is being, especially in marketing, the more negative perception we have of them. That's just me though
It's just that now there's an even cheaper way to do that.
Both low talent AI use and stock photos have their own look about them and neither is premium.
Frankly speaking they are getting so good I can hardly tell by first glance
I'll agree that people who don't care about sewing and calligraphy probably won't notice, but there's a difference between "you can't even tell" and "you can't even tell as long as you don't care too much about the result".
In 2 (time units) we'll be doing computer analysis of lens distortion or something to try to suss out the AI. At which point it won't matter for the stock image use case, of course it matters for legal matters and such. And then in 1-2 more units we're going to need public/private key signing implemented in 'cameras of record', because detection will be practically, if not actually impossible.
Is that 'unit' days or years? Dunno, but I bet it is a lot closer to the former.
Shutterstock used to have a program called "Red Carpet" where they endorsed independent photographers to help us get in to events as press. Then like all good things, it was shut down, no explanation given. ¯\_(ツ)_/¯