About AI … 7


In recent weeks, we have heard a lot about Artificial Intelligence (AI). A bit of everything. It started with ChatGPT. Then we started talking about software like Midjourney. I know almost nothing about these programs, but I hear about them and John Harper passed me a few links with pictures made by AI. It’s pretty amazing and impressive. John has written a very good article on the subject here. Go and have a look.

On my Google feed I got quite a few notifications with articles on the topic. Photography related articles of course. I still don’t understand how Google manages to collect enough data on me to make these kinds of suggestions, but it’s a safe bet that AI is behind it too! Maybe I should reconnect my VPN to surf more discreetly… Anyway, I came across an article talking about Street Photography and AI generated images. To read it, it’s here !

I don’t like these photos. They are too clinically clean and you can see that they are AI generated photos. The problem is that we are only at the beginning and soon I doubt we will be able to tell if a photo is real or not. For several years now, with the advent of digital technology, there have been debates about photos post processed on LR or Photoshop. There are those who consider that a photo that has been retouched too much corresponds to the work of a graphic designer and no longer of a photographer. What is the limit? Some purists swear by film because apart from dodging and cropping, there was very little to retouch the photos. You could double print on film, but nothing like the possibilities offered by software like LR and Photoshop.

Clearly, with the new apps like Midjourney, we are radically changing the era! I’m not resistant to change because in my opinion it’s inevitable and it’s better to go with it than to fight it. What bothers me is that these images are only created from a compiled database. These pictures do not exist. They have been generated. The machine was inspired by thousands of existing photos to recreate another one in the same style. A lazy machine that didn’t even bother to go out on the street, to see things, to observe the scene, to frame, to capture the moment. No. This machine just has great calculation skills to produce something.

But this machine, however efficient, is not yet perfect. Will it ever be? I hope not. For the moment, what is certain is that it has problems in correctly modelling certain elements of the human body. A dummy’s guide to detecting AI-generated images:

  • Look at the eyes. They are often different sizes.
  • The hair is also a problem for him. It is apparently very complex… Nature does things well!
  • The skin is often too clean with faces that seem to be made of wax.
  • When smiling, you will see that the mouth contains well over 32 teeth!
  • Funny thing, AI has trouble generating text that makes sense! Whether on billboards, posters, books … It’s really stupid in the end.
  • Look at the hands. That’s where the focus should be. The AI has a lot of trouble generating them well. Very often the fingers will be too long and there too, as for the teeth, you will often have hands with more than 5 fingers!

In short, the AI still has flaws. But I am also aware that these flaws will be erased in future versions. I’m posting a small collection of photos I found on the web.

#1

This photo was generated in the style of Vivian Maier. Several disturbing details in this photo. The buttons that have random spacings and the holes that don’t match in front. But the worst part is his left hand with 6 fingers!!!

#2

This one was generated in the mind of Garry Winogrand. I’ll skip the hair and skin texture, but again I’ll focus on the hands. The lady in the middle has 6 fingers and the guy in the foreground has deformed fingers. You’ll also notice that the texts don’t mean anything on the tram. It’s even more obvious on other pictures I’ve seen with posters in a shop…

#3

This is a far cry from the 32 teeth that adults are supposed to have! They are cute, but I find them a bit creepy with their smiles!

If you are interested in photography, you will have heard about the German Boris Eldagsen who managed to trick the jury of the Sony World Photography Award. Not only did none of the jury (sic) see the trickery, but he also won the award! Which he of course refused, as he admitted to having voluntarily submitted this AI-generated photo to open up a debate on the challenges facing Photography. Clearly, it’s a great job, but I wonder about the professionalism of the curators… You’ll say, it’s easy to say that, once the deception has been demonstrated (here admitted)! Yes, it’s not wrong, but with the recent advent of all these AI apps, we have to be a bit more vigilant, don’t we?

Here is the photo in question. When I saw it (and I knew it was an AI-generated photo), I went looking for errors. And quickly I saw the image and it seemed to me that several fingers looked weird… On the left we have a little finger that is almost the size of the ring finger and on the right side of the image, an index finger that has a weird angle … If we look at the eyes, it seems to me that the pupils are of different sizes … As I said, it’s easy to say that after the deception has been revealed. The thing is, we’re not yet formatted to ask ourselves the first question: Is this photo a real photo or an AI-generated photo? Unfortunately that’s going to be the norm now.

The consequences will be terrible. Not only for the world of photography as I practice it. But soon we’ll be able to put anyone in the picture and make them do anything. I’m not going to start rambling on about it, because we’re essentially on a Street Photography site, but you can see the danger of AI-generated images if we were to stage public, political people…

I am sure that some ill-intentioned people will have no qualms about crediting themselves with Street Shots when they have been generated by AI. There is nothing to gain from Street Photography. No money. Finally, there are still photo contests with some prizes that are worth it (money, GEAR, …). And then there are the social networks, where long before AI, photographers were working hard to get the spotlight on them. This was done on merit, but now there will be a new element that could change the game. Again, I’m going to repeat myself, but this is just a consequence of our social networking world. Did you know that many young people have plastic surgery because of the filters they use on Instagram, Snapshat, TikTok or other apps, the image they see in the mirror does not match the image they see on their social networks with their filters!

The only person you should impress with your photos is yourself! Frankly, if other people’s opinions agree with mine, so much the better. Otherwise I don’t care. It’s been several months since I shared anything on my website. Am I still doing street photography or have I given up completely? If you know me at all, the answer is obvious. I never stopped… I’m doing less, that’s true. But I still document the streets and I don’t need to fool myself with AI photos.

All the pictures on this blog (except the ones with weird fingers) were taken with the Ricoh GRD IV

Share Button

Leave a comment

Your email address will not be published. Required fields are marked *

7 thoughts on “About AI …

  • Deborah Swain

    Great to read your thoughts Jeff … and loving these latest Ricoh GRD IV shots! Never has street photography felt more important. All we can do as street photographers is remain true to ourselves and keep on doing what we’re doing, out there actually recording the real world as we see and experience it.

    • Jeff Chane-Mouye Post author

      People who will use this kind of app will miss all the beauty of street photography. On the street, I don’t create anything. I just capture a moment that happens in front of me. It doesn’t have to be a decisive moment but a moment I saw and that I experienced as a witness. People already cheat with staged scenery and pretend it’s candid so I expect to see glorious shots generated by AI without people claiming it…

    • John Harper

      Strangely when we look at the definition of ‘Artificial’ then we find it is either a ‘copy’ or ‘insincere’. That tells you a something about the potential controversy surrounding AI imagery. Incidentally this is a much better blogpost than mine. You’re showing examples and we can all see the glaringly obvious. Like you I’m certain it will advance and we’ll be unable to spot the difference. Apart from it’ll still be made up of fragments of others work, that some photographer will shout “hey, I took that shot of the guy in a car”. You couldn’t pass of a landscape photo and announce that you painted it, AI has to be distinguished somehow from the real thing.
      The Real photography here is much better too, I can definitely see someone typing “Street photograph in the style of Jeff Chane-Mouye”.

      • Jeff Chane-Mouye Post author

        I think that there is nothing wrong copying someone’s work. Well let’s formulate it differently. I’d rather have someone getting inspiration from someone’s else work. We all do that. But it has to be done your own way. What bothers me is that they just some database to build a photo. And in the case of photos in the style of Vivian Maier, Winnogrand or Alex Webb, is that the photos are codified. Light, shadows, bench, portrait…
        I feel that it is much more than that. One thing that comes to my mind is “Soul”. Something AI is still not able yet to reproduce. I hope that AI will never be able to do it…

        • John Harper

          AI is already smarter than every human on the planet with ‘general knowledge’, but it’ll never have “soul”, that can’t possibly be replicated or copied. If doesn’t have soul it can’t have a style. It’s completely soulless.