AI and Photography

boojum

Mentor
Local time
8:28 AM
Joined
Jan 23, 2021
Messages
2,247
AI and photography have been discussed out on the net a lot recently. And now with ChatGPT-4 it has some amazing tools for creating "real" images and surreal images and all in an instant on your laptop. And according to this video the rate of progress in the field is accelerating. And getting more efficient and cheaper. So what does it mean to those of us lugging film cameras around or our cherished CCD sensor cameras or even the hot-shot new cameras whose BSI-CMOS sensors are really, really good? After all, it is possible to prompt your laptop to generate an image as good or better. True, it may not be the image of something before you, yet.

So what have we here? Is it the Gutenberg press threatening calligraphic books? Are we to be Luddites by the loom? Or are we being presented with a wonderful new tool? Or is it HAL?

Thoughts?

()
 
I find the projected effects of AI (in the VERY near term) to be terrifying. Given our society's inability to rein in the effects of social media platforms, I'm very concerned that the AI developers are playing with fire!
 
I find it laughable.

Thirty years ago "The Big Boss" would only believe that a picture from me was real if I handed it to him after coming out of a Polaroid Camera.

I see opportunity for Wedding and Portrait Photographers shooting more Polaroid for clients that "want to keep it real".

Computers are deterministic. At the physical layer- run on a computer executing simple instructions. Very quickly- but simple. It might take a while, but you can track down why a piece of code produces a specific output. Read more Asimov.



HAL, I could get along with. ChatGPT- propaganda machine, too many politically correct governors artificially biasing the results. One time I rewrote the error messages for a compiler to emit HAL like messages.

For the DIY part- I've written codes with routines spanning over 40 years. Every so often the code gets itself in trouble, and they pop up. Sometimes the code makes a decision when encountering something new and unexpected. That is cool.

Some Day... Some time when least expected, the Infamous "Non-Existent Error Diagnostic Number 511" will rear its ugly head. But I've only had two computers burst into flames due to software that I wrote.
 
Last edited:
I find it laughable.

<snip>

Some Day... Some time when least expected, the Infamous "Non-Existent Error Diagnostic Number 511" will rear its ugly head. But I've only had two computers burst into flames due to software that I wrote.
Yeah, right. That's what you would have us believe but I understand that there have been numerous mysterious and unexplained exploding computers in your area. It may just be synchronicity, it may be. But then again, . . .
 
The moment human beings started comunicating with written language 'anything' became possible.
 
I was thinking of the classic example of psychological synchronicity. A dog walks into the back yard and pees against the tree. At that same instant someone throws a bone over the fence. The dog's mind sees a connection but there is no connection. This is synchronicity that is not engineering. It abounds. Like exploding computers, accident and no connection or some really tight code loops? Who knows? The Shadow. Anyone else?
 
Similar discussions come up time after time again. It's like it was with chess computer programs. Every discussion about chess was dominated by the question about whether a chess program could beat even the best chess players. It was such a relief when that was finally accomplished, because that meant we humans could leave the competition with computers behind us and focus on the game again. Sure, a chess program can be and is used to hone our skills, but we don't have to necessarily compete with it anymore to prove a point. Basically, a chess program has now become something like a forklift; it's a tool to be used when appropriate, and it hasn't replaced people playing chess.

This will go the same; for a while it will be AI imagery versus photography. But this will pass, and photography will find its own way again..
 
This will go the same; for a while it will be AI imagery versus photography. But this will pass, and photography will find its own way again..
In a way but in the meantime, very many photography jobs will have disappeared; first in North America and years later, in Asia. What has taken people by surprise is how fast AI is developing and with the speed AI commercial applications are launched in the marketplace. Cheers, OtL
 
Last edited:

Even 45 years ago a friend told me that Weddings were photographer's bread and butter. Staff photographers for magazines and newspapers are way down, but AI had nothing to do with that- more like Cell Phones shooting Video. According to the above site, the median salary of photographers is about what you get working at Starbucks. It is less than what my work pays for a college Intern in IT.
 
Last edited:
So what does it mean to those of us lugging film cameras around or our cherished CCD sensor cameras or even the hot-shot new cameras whose BSI-CMOS sensors are really, really good? After all, it is possible to prompt your laptop to generate an image as good or better.
Well, I would think that photographers actually like to use cameras and like using reality. I can't imagine using text to generate an image will be as fun as using a camera out in the world. It is just different. AI is not photography.
 
Just about any travel site that you can visit has already been photographed and the images published in a Book, Magazine, or Online. Yet people have brought cameras and taken pictures at those same places since Kodak Brownie Days.

AI generated imagery is just the new buzzword. "It's a thing", more like "It's a nothing that's been around for a long time". They have their 5 minutes of fame. The "AI will kill Photography" is the least threat to the Photographic Industry. THAT is the incorporation of Cameras into Cell Phones.

35 years ago I used fractal generated imagery in a simulation model. Several people seeing the simulation refused to believe I did not use real images. Show them the code.
 
ChatGPT-4 has emerged as quite powerful. It can be used to generate very good images. I am sure that painters voiced similar complaints and observations when photography showed up. The YT video I linked was provocative. Current news blurbs on the power of ChatGPT-4 are stimulating. Predicting the outcome can be difficult but for sure AI will change things in photography and many other places. A recent news item is that MS is using it to thwart hackers. As it seems able to learn, just how much will it learn?
 
ChatGPT-4 has emerged as quite powerful. It can be used to generate very good images. I am sure that painters voiced similar complaints and observations when photography showed up. The YT video I linked was provocative. Current news blurbs on the power of ChatGPT-4 are stimulating. Predicting the outcome can be difficult but for sure AI will change things in photography and many other places. A recent news item is that MS is using it to thwart hackers. As it seems able to learn, just how much will it learn?
It can be all of those things and still not be photography. No doubt about it, it will change the way images are used. That does not mean it is photography.
 
When you can feed ChatGPT a machine readable executable file and it tells you exactly how the program works I'll be impressed, can retire guilt-free, and set up the Darkroom again. Until then- my wife laughs and tells me "yeah, give it to ChatGPT- like that is going to do any good."

I don't get why everyone is so spun up about AI recently. To me, it is just the SOS. My Nephew told me some of his friends were submitting ChatGPT generated term papers for Masters level courses. Told him WW-II techniques for "Specific Emitter Identification" could be used to identify papers that were not written by a Human. Just modify an Open-Source word processor, such as from Open-Office, to time-stamp keystrokes. Cryptographic checksum generated by the Word processor to tag the file. Students are given a uniquely keyed word processor to use for the class. I of course use Wordstar, but don't have to write term papers anymore.

DIY ChatGPT detection. Saw it here. Something similar could be done for detecting images from specific cameras based on non-uniformity in detectors, probably. You would need the raw files. No raw file- didn't happen?
 
Last edited:
Wordstar!? Wordstar!? Didn't that one come with ink in case the CP/M system locked up? LOL Would only print dot-matrix onto parchment.

About ChatGPT-N - I am reminded of the great Adm. Grace Murray Hopper's comment on computers and how they were amazing folks back in the IBM 1401 days when you could get a 12K machine or the big 16K machine, "We are just at the edge." What I found interesting in the AI discussions in general and ChaptGPT in particular is the accelerating pace of development which is a result of ChatGPT itself. Forget the labels, forget that its term papers can be identified, this thing is like yeast. It is expanding in scope and power on its own. Sixty-five years ago Ray Bradbury had a short story about something like this, back when the only computers took up whole rooms. Will this be like Mickey Mouse in The Sorcerer's Apprentice? This is way hipper than Pong. This is worth watching.

MS is using ChatGPT to thwart hackers. Imagine what it could do when induced to hack. And then reflect that what we are looking at here in the civilian sphere has probably been surpassed long ago elsewhere. You can be sure that some "unnamed government agencies" have some hotshot coders. Just like how phone companies will make a generous offer to really slick hackers to come work for them. "Captain Crunch" was not the only one to put on the white hat. Back at the dawn of time when I wrote code I had some buddies in tech who were doing "impossible" things on mainframes to bypass security and protections within a huge and sophisticated major insurance company's shop. The real grey area tech guy used to always say, "It's all in the manuals, it's all in the manuals." He was operating at a level below machine code so you know he had one of those pointy hats with stars on it.

My whole point is that this could grow into something huge altering the image industry and others unless it is somehow caponized. MS and Google have put up barriers but as my grey area tech guy knew, you can do a lot of cute stuff if you just read the manuals. This should be a fun ride.
 
Last edited:
… It was such a relief when that was finally accomplished, because that meant we humans could leave the competition with computers behind us and focus on the game again. …
Your chess analogy is perhaps closer than you think in regards to photography.

As recently as last year and even as far back as 2006 (the “Toiletgate” incident), there have been allegations of grandmaster-level players receiving coded signals over wifi from an accomplice with access to a near-unbeatable computer program.

You will have analysts say that a certain player‘s moves in a game are matching that of Stockfish (the strongest computer chess program today) to such a high percentage that he cannot possibly not be cheating - especially when the player cannot adequately explain his choice of move in an interview later.

The photographic equivalent might be artists whose photos aren’t entirely real or, more maliciously, photos of events or situations that aren’t entirely real. That has been happening even without AI - AI just makes it easier.

For most of us, either in chess or photography, none of this will matter.
 
Back
Top