“Digital cameras are doomed!”

[DrLex]

I bet you have also seen them before, the doomsday predictions that digital cameras are going to be displaced by new technology. While skimming through comments on the latest DPR announcement, I bumped into this one:

But the elephant in the room is a close to dying digital camera mass market. Phones, drones, AI, generated pictures are on the rise.

I thought to post the following as a reply, but then I realized it would be a waste of time given the uncertain future of the DPR forum content, and this forum is probably a better place. Maybe this can serve as inspiration for others when they need ammo to debunk similar claims.

Phones are only comparable to professional cameras in a limited set of circumstances. There's only so much one can do with a tiny sensor and lens crammed inside a device that has to do a gazillion other things, before one bumps into fundamental limitations, which a device whose every component is designed for the sole purpose of taking photos does not have.

Comparing drones to digital cameras is… weird. Drones are flying machines that may optionally have a special digital camera attached to them. They are good for aerial shots and not much else. I wouldn't recommend trying to use a drone to take someone's portrait, or a macro photo of a flower. Are you going to hold the drone in front of you, or are you going to try to maneuver it such that its propellers just don't shred that person's face or the flowers? I can see it before me, a couple sitting in a restaurant: “honey, you look really nice in this candlelight. Let me grab my drone.”

As for the AI things, how will they replace the act of taking real photos? When on vacation you see something nice, or when you want to take a photo of family members, are you then going to grab your nearest computing device, launch some AI app, and start describing what you're seeing, in the hopes that the system can somehow replicate what you're looking at, or are you simply going to take that photo yourself? Again, try to apply this to the restaurant scene… “No, her nose is too big, enhance, enhance!”
Also, those systems need to learn from input material to be able to generate output. Only if the input is good, the output has a chance of being good, otherwise it's GIGO, Garbage In Garbage Out. What do you think these things are being trained on? Photos made by real photographers using real cameras, right. If you see really impressive demo photos generated with an AI system, the demo was likely rigged to basically replicate some of the training images with few modifications. Ask the system something that cannot be pasted together from bits of its training data, and the result will be awful.

Digital cameras aren't going away. Their market might shrink, but it will not disappear.

 

[RBIV]

I agree. Leica is still building rangefinders. And [hopefully] Pentax will hang in there with OVF bodies.

But the market share leaders will be mirrorless, maybe some consolidation there as we go along, but probably better EVFs and [hopefully again] better capacity in smaller batteries.

I personally think this is a great time to purchase camera. So much choice, and all of them quite good.

 

[Lee Jay]

My primary purpose for photography is documentation. You can't document anything with an AI-generated image. Making stuff up is the opposite of documentation.

Phones today are pretty good in a wide range of conditions, but far from all. Since most of my photography is edge cases (at least, for most people), phones leave a lot to be desired for my type of photography. But I still use mine a lot for simple stuff.

 

[Tom Caldwell]

I bet you have also seen them before, the doomsday predictions that digital cameras are going to be displaced by new technology. While skimming through comments on the latest DPR announcement, I bumped into this one:

I thought to post the following as a reply, but then I realized it would be a waste of time given the uncertain future of the DPR forum content, and this forum is probably a better place. Maybe this can serve as inspiration for others when they need ammo to debunk similar claims.

Phones are only comparable to professional cameras in a limited set of circumstances. There's only so much one can do with a tiny sensor and lens crammed inside a device that has to do a gazillion other things, before one bumps into fundamental limitations, which a device whose every component is designed for the sole purpose of taking photos does not have.

Comparing drones to digital cameras is… weird. Drones are flying machines that may optionally have a special digital camera attached to them. They are good for aerial shots and not much else. I wouldn't recommend trying to use a drone to take someone's portrait, or a macro photo of a flower. Are you going to hold the drone in front of you, or are you going to try to maneuver it such that its propellers just don't shred that person's face or the flowers? I can see it before me, a couple sitting in a restaurant: “honey, you look really nice in this candlelight. Let me grab my drone.”

As for the AI things, how will they replace the act of taking real photos? When on vacation you see something nice, or when you want to take a photo of family members, are you then going to grab your nearest computing device, launch some AI app, and start describing what you're seeing, in the hopes that the system can somehow replicate what you're looking at, or are you simply going to take that photo yourself? Again, try to apply this to the restaurant scene… “No, her nose is too big, enhance, enhance!”
Also, those systems need to learn from input material to be able to generate output. Only if the input is good, the output has a chance of being good, otherwise it's GIGO, Garbage In Garbage Out. What do you think these things are being trained on? Photos made by real photographers using real cameras, right. If you see really impressive demo photos generated with an AI system, the demo was likely rigged to basically replicate some of the training images with few modifications. Ask the system something that cannot be pasted together from bits of its training data, and the result will be awful.

Digital cameras aren't going away. Their market might shrink, but it will not disappear.

I also agree.

Put a 4/3 sensor in phone and make it do all the things that a M4/3 lens can do seems ridiculous. Thy can bend rays to fit a telephoto zoom inside but we might remember that they were doing exactly the same thing with compact cameras and the idea was not exactly a roaring success.

To say that the mobile phone can make the same image inside a phone that is still regular pocketable phone size is to make us wonder just why camera companies have been forcing us to buy such large lenses for our M4/3 kit when with a bit of mobile phone manufacturing tech we could easily fit (say) a 300/4.0 lens inside any M4/3 camera body and have so much room left over that the parts would rattle around inside.

To set the mobile phone up as a true systems camera this would entail a mount system - it is hard to think that the flange focal distance would be any more compact with so many M4/3 lenses already begging to be mounted and used .... really the GM series was also hardly a run-away success and we might be thinking of a protruding mount that would ace the pocket-size idea straight away and we would need our jolly big bag of lenses and accessories over the shoulder and good luck in getting that into your pocket. But even if somehow a flush mount could be organaised then it would be yet another new ML mount system with no lenses on the horizon.

Just try your tiny "100/16.0" lens that can be attached without grumbling about the lack of grip. No so sure where this would get the dedicated 4/3 systems phone user in the equivalence argument.

Most phone users have no idea on how to compare their wide angle lenses to the lenses in real camera bodies - probably just as well - even though wide-angle and computational photography seem to satisfy the occasional snapper and some of those with skills can make very worthwhile images from their phone.

What regular cameras cannot offer is the marketing medium of mobile pones which allows that the need of a phone (the latest and greatest) is such that they will pay quite a lot of money for the phone and get a 'free' camera with it.

Out at a band "do" just recently I was the only one with camera kit and the lighting was 'difficult' to say the least but there were mobile phones there snapping away but I did not get to see just how well they performed under those conditions. I also had a special tool up my sleeve - I had the knowledge, means and inclination to post-process my raw file captures ....

However AI and all things auto simply take away the learned skill element and by making things easy and automatic are going to result in accepting the common denominator as 'good enough'. Self drive cars? Is nobody ever going to feel the need to learn proper driving skills and pride themselves that they can drive well?

Similarly I have tended to practice a lot with all sorts of camera gear - testing myself in different conditions. This came partly from the type of person that I am and partly when I first took up photography seriously I missed some 'moments' that I could have captured - even with camera in hand - because I took too long fiddling with the camera to 'get my settings right'. Now I can go to a band event and more or less get my camera set to the venue without apparently thinking about what I am doing.

For the most part I am normally just testing myself to see what I can do and the journey is more satisfying than the arrival at the destination. I take a lot of very ordinary pictures in the process. But that is just the handmaiden of learning a skill well.

 

[clewis010]

I suppose AI might be helpful in some situations if you know exactly what you want always assuming the quality is there. But, let's say, you want a photo to illustrate a street scene in New York. Do you know enough about what you want to instruct AI to produce it? Or do you prefer to go to Adobe or Shutterstock and scan through a few hundred of their stock images to find something that really works for you?

I remember when I was working and we needed a photo to illustrate the cover of a report or photos. We would tell marketing that we need a shot of, say, a swimming pool. That's all we really knew so marketing would send us a dozen or so swimming pool photos from one of the stock agencies and we would choose the one we wanted. Could AI do that better? I don't really know.

Then, of course, AI isn't going to generate those photos of the grandkids. Yeah, a phone can do it but I can do a way better job with my camera.

I do use my phone to take photos from time to time. These are usually ad-hoc shots because I don't carry my camera with me all the time. The photos are okay but can't be compared to the camera for quality. First of all, I have more difficulty holding the phone steady than I do a camera. Second, it's all on automatic with no adjustment of aperture, ISO or shutter speed. Maybe you can do that with a phone. I haven't bothered to find out. I think of the phone as the equivalent of the 'point and shoot' cameras I owned before DSLRs became a thing.

 

[Tom Caldwell]

I suppose AI might be helpful in some situations if you know exactly what you want always assuming the quality is there. But, let's say, you want a photo to illustrate a street scene in New York. Do you know enough about what you want to instruct AI to produce it? Or do you prefer to go to Adobe or Shutterstock and scan through a few hundred of their stock images to find something that really works for you?

I remember when I was working and we needed a photo to illustrate the cover of a report or photos. We would tell marketing that we need a shot of, say, a swimming pool. That's all we really knew so marketing would send us a dozen or so swimming pool photos from one of the stock agencies and we would choose the one we wanted. Could AI do that better? I don't really know.

Then, of course, AI isn't going to generate those photos of the grandkids. Yeah, a phone can do it but I can do a way better job with my camera.

I do use my phone to take photos from time to time. These are usually ad-hoc shots because I don't carry my camera with me all the time. The photos are okay but can't be compared to the camera for quality. First of all, I have more difficulty holding the phone steady than I do a camera. Second, it's all on automatic with no adjustment of aperture, ISO or shutter speed. Maybe you can do that with a phone. I haven't bothered to find out. I think of the phone as the equivalent of the 'point and shoot' cameras I owned before DSLRs became a thing.

Yes there are times when you want that certain look - low key or high key, or even wedding cake capture .... you can get scene modes to maximise the camera for these types of scenario but how many scene mode might there be? And surely managing the scenarios can become even more complex than A-S-M + EV + ISO.

The old adage comes to mind - "it is better to teach someone how to fish than give them a basket of fish". The basket of fish is obviously Ai (which is quite popular really).

For quite a few years I have been trying to teach myself how to manage different capture scenarios with a selection of gear. Improving a bit, but maybe another 10 years and I will have mastered it .... . But I enjoy the process of learning and mastering my subject matter - what is wrong with me?

 

[grsnovi]

I don't own a drone (and probably never will) however there are certainly advantages to having an aerial-mounted platform. There are shots you can get with a drone that would be impossible otherwise. Surfing, kite-boarding, skiing all offer great opportunities for airborne selfies. Maybe if my iPhone could fly by itself?

 

[RBIV]

My primary purpose for photography is documentation. You can't document anything with an AI-generated image. Making stuff up is the opposite of documentation.

...

Absolutely, and a point lost on many.

 

[Tom Caldwell]

I take it further - there are two main branches of photography - the pictorialist or painted art style of photography which tries to emphasise on the feeling of what the image shows - or 'the impression' I suppose. The precise detail is not necessary for the mood. Of course early photography was by nature more dreamy and soft focus. But many of these image made in early photography are still revered. This is a hang-over from what the artist with his easel and box of paints once did. The fact that the artist could take liberties in portraying his subject was not only understood but probably encouraged by his ultimate audience and the subject of the portrait if it was in fact a portrait.

As the gear and the medium improved it became possible to make very realistic images which resulted in the f64 movement. Photographers have been on the high road to ever greater precision ever since. One might wonder just where this realism might end up. Considering how much work goes into improving the model's looks in post processing advertising (which is never admitted) we might wonder which fake-art is really being truthful about what it portrays. Even such greats as Ansel Adams fiddled with his prints in the darkroom.

This is not to say that a realistic image cannot also move the soul - they surely do and I accept that. But I wonder why the more artistic dreamy, even painterly images have now received what appears a bad rap. I would propose that it is hard to make a bad image better by a painterly wash. In fact we should agree that to make a great image into a faux-painting is actually harder then making a 'snap' into a perfect one with modern gear. Even if we go to the digital darkroom and enhance dynamic range, and fix any exposure inaccuracy and render what we think was the correct white balance.

Of course images as documents need to be realistic but if we de-clutter background (or even remove it) for documentary clarity does this make our image less useful as a document?

Ai - what is that? Does this mean pointing the camera at a subject with great lens in "A" mode and having the camera with its impressive expensive lens work out shutter speed for perfect exposure, adjust for perfect white balance , grab more dynamic range out of shadows and highlights, and of course glide quickly to perfect focus in split parts of a second then effortlessly keep that perfect focus of that moving object. All this and perfect life-like reproduction. What a wonder, and all internal camera body Ai. We don't all like it but the market demands it.

 

[RBIV]

.....

Ai - what is that? Does this mean pointing the camera at a subject with great lens in "A" mode and having the camera with its impressive expensive lens work out shutter speed for perfect exposure, adjust for perfect white balance , grab more dynamic range out of shadows and highlights, and of course glide quickly to perfect focus in split parts of a second then effortlessly keep that perfect focus of that moving object. All this and perfect life-like reproduction. What a wonder, and all internal camera body Ai. We don't all like it but the market demands it.

I'd say in this example, AI is influencing the tool. Clearly, many of us would enjoy a camera that automatically delivers "perfect" exposure, white balance, focus, sharpness and saturation," DEPENDING of course on our subjective definition of "perfect" AND whether or not we want it perfect in this case.

In terms of the difference between "journalism/documentary" and "fine art/ pictorialism" well, isn't that always subjective? I think so. HCB decided upon the perfect moment. Walker Evans where to point the camera. Ansel how dark to make the background. And me, should I blur that waterfall as it falleth, or snap it at a higher shutter speed to look more "natural."

So there's art in art and art in documentary as well, right? I suspect AI is a long way in determining much meaningful in this regard, other than knowing that X% of the population likes blurred water waterfalls, or Y% don't care if a photographer clones a beer can from the sidewalk as his subject glides by.

I wonder if the ease with which users of real cameras can override AI may be one advantage over computational cell phones.