Boob to you too this fine morning.
I take pictures all the time and some of the tools that I use do apply AI to correct a focus near-miss, or just apply machine learning to help me sift through photos when for example I have 1445 shots to look through, like I did last night. That sort of AI is fairly innocuous and beneficial.
Some things to look at with
generative AI photos are the complexity of the background and what is motivating light sources. Generated photos will put highlights and shadows in weird places. Truthfully, I can put light just out of frame, but that's something to consider when you're looking at AI pictures. If your intuition says the lighting is off, as it definitely is with the green dress image, it's probably some fake BS. You can also question how clothes seem to be folding or wrinkling. It's an intuitive thing again, but you have decades of looking at people to look back on.
Second: Skin has texture. If it looks like plastic, it's either been smoothed to the point hat it might as well be AI, or just AI. When I use my Canon R6, which is "only" a 20MP camera, even taking a full body picture on a 50mm lens, I have enough detail that I can see pores, tiny hair and variances in skin tone. I work with models and/or dancers, and they do be wearing makeup, but even people who are paid to be professionally hairless probably didn't wax their upper arm or the small of their backs and their foundation and concealer probably stopped at their neck. Most photos aren't going to get a Cosmo Cover worth of editing, so plastic-looking skin is a huge hint.
A third consideration is background complexity. This one is a bit difficult because there really are perfect, endless beaches and plain white walls, but if a background is in perfect focus but still kinda plain, that's a HUGE red flag to me. Normally there's a focal plane in an image and the foreground and background are at least a bit fuzzy. But your brain does know what a room that actually exists looks like and decades of looking at real pictures has probably given you some idea that pictures of people will either have a candid, snapshot look about them (think Polaroid or picture from your phone) or the subject separation of pro-ish type pictures. In the case of all three of the above pictures, you have perfectly in-focus but weirdly generic backgrounds as if the subject were just composited into the frame, which it probably was.
It's only going to get harder. We used to tell people to count fingers, right? These things are going to get harder to figure out. That doesn't mean we should stop trying to take pictures. It makes the skill of taking pictures more important. Lunar goes out and takes pictures of animals and I'd have a much harder time trying to distinguish that sort of picture from something an AI made.
Purely generative video is currently not ready for prime time, to the tune of like tens of thousands of AI prompts to get enough marginally coherent shots to make up a 30 second commercial. It's getting easier and easier to paste a head on a moving body, sure, but if you're concerned about things like news, keep in mind that TV outlets regularly pull "clips" from the internet that are straight up captures or screenshots of video game footage. Some media outlets and Presidential Administrations are also EXTREMELY willing to mislabel old video clips for some new purpose. They don't have to generate fake video. They can just say the pictures from a riot in 2017 are the riot they WANT to report on in 2026. This is especially easy for regions of the Middle East, Latin America and the states on Russia's eastern border, which most Americans know exactly jack shit about.
If you're concerned about something in the pure news space, a resonable thing to do is to look for an Associated Press or Reuters source. If there's an event that's happening in real time, such as the public murder of Renee Good, you'll probably be able to find multiple streams from multiple sources. If there's a single source of dubious factual standing, you don't have to immediately give it any credence, but you should take the time to see what other organizations are saying about it. I have some thoughts about that in the
Empire in Decline thread.