Samsung exec says all photos are fake

Samsung phones are increasingly using AI to “enhance” the photos. I don’t think you can even turn it off, but I don’t know for sure. At any rate, an interesting if not scary comment from one of the largest camera makers (with non optional phone attached).


In my opinion what he said is true, even if he only used that comment to sway people’s opinions towards AI enhanced pictures.

I guess there’s a scale from almost real to as far away from reality as possible, people will have to chose where they draw the line.

This will bring a lot of fun discussions no doubt. A fun one I just thought about: A long exposure doesn’t capture reality in so far as humans experience it. A photo could have its details and dynamic range enhanced with AI in a way that “simulates” what we see, so that it’s imperceptible from our experience. Which ends up being the “realest” photo? (At least in relation to human experience)

Of course this is no excuse to replace photos of the moon etc.


“Real” and “reality” has nothing to do with images.
“Realistic” perhaps.


While I agree that the idea of SOOC is naive, these kind of statements make a giant leap from reality is inevitably mediated, to nothing is real and we can just make stuff up for our convenience. Surely part of the wonder of photography is the human struggle to mate beauty, however defined, and perceptual realism. If it’s just dragons and sh*t, what’s the point.


The Samsung exec was talking about pictures, not photos. I suspect that was a deliberate difference.

What he means is: “AI is good at inventing plausible but false stuff, aka hallucinations. If we can persuade you that everything is false, you won’t care that our products make false stuff.”


Well it will be hard and I am not sure how you regulate things like this but lets say you use magic eraser in google photos and you have a famous church and you want a sign or other obstruction removed…its not using just math anymore its also using other images of the church to decide how the area should be reconstructed so your final image could have elements from some other scraped photo added to your photo… who knows if something similar doesn’t go on when you let google reprocess your video… or maybe not yet but soon… for sure the image gymnastics is like going to get more invasive rather than less… I don’t really mind I guess as long as you can always turn things off…its when you have no choice that it bothers me…


The mobile phone industry is under a very specific kind of pressure. People are bombarded with photos that are well-composed, taken with good optics, and expertly developed.

Yet when they take snapshots with their phone, the quality usually does not come anywhere near a consumer-grade fixed lens film camera from the 1970s shooting on B&W film.

Sure, some mobile phones have a decent camera, but there is only so much you can do with a small sensor and small optics that has to be squeezed into a tight place. And learning a few rules basic rules of photography and composition can improve photos tremendously. But the majority of photos shot with mobiles are still crap, regardless of the kind of phone.

The answer is “enhancement”, and it will end with AI effectively painting a picture that bears some resemblance to the one that hit the sensor, but people will notice the difference. So they have to be sold the idea that this is actually a good thing. Manufactures have been will be competing on backends that enhance pictures, running on their server farms. For an initial period it will be free, after that it will become a subscription service or similar — computational power is not free and there are no incentives to provide it at no cost when the phone is a couple of years old.

Does this mean the end of photography as people practice it here? Not at all. In practice, taking snapshots with your mobile phone has been separate from any kind of camera photography for quite a while at this point. The majority of consumers will continue to be sold the illusion that one is a substitute for the other, but even an untrained eye will be able to distinguish pics shot with a decent compact zoom (!) from a typical phone camera, and a prime lens on any kind of MILC from the past 10 years will be much better than top-of-the-line mobile phone sold for more than $1000.


I strongly disagree.
More and more phones are getting ‘AI chips’, and the processing is already done on the phones, no back-end is involved.

Those consumer cameras, and the photographic skills of the average user were not all that good. At least, the ones in my family were not. One can always check out, but you need to be careful to check, as there are many pro (agency, newspaper) photos.

Not all ‘AI-assisted’ computational photography is evil / bad / fake. Capturing true HDR images, ‘sensor shift’ in the hands of the user, merging many frames to reduce noise are examples that either our ‘real’ cameras or we ourselves also do.
And the result can mean a win for the phone – see my earlier thread The phones are catching up.

1 Like

You make a very valid point, and I agree.

Just to clarify: I was talking about “heavy inpainting”, ie partial regeneration of an image. See eg I think this is an area where we should expect a lot of R&D money, and it is coming to phones which don’t have the processing power to deal with it, so images will make make a roundtrip to the backend server.

While I think you’re technically correct, in practice I don’t think it’ll really happen. Numerically, the vast majority of cell phone shooters take selfies and other types of long-term worthless / throwaway images. While they should be able to (and might if they tried) see the difference, they won’t because it just doesn’t matter to them. A typical selfie or other cell phone snap isn’t generally about IQ, no matter what the manufacturer would have you believe. There’s probably a baseline but it’s low – just look around at the poor cell phone snaps you see described by others as “beautiful”.


Note that I am not claiming that anything will “happen”. Yes, pictures from a real camera are better, and people will continue to know this but not switch because of various reasons.

Photography is a hobby that takes money (but you can get by with surprisingly little), and a significant investment of time to reap the benefits. What I think will happen is that mobile phones will use more and more AI inpainting to enhance images for social media where they get their well deserved 2 seconds of attention.

My claim is that at this point mobile phones are in practice effectively decoupled from photography practiced as a hobby or a profession. It is two separate worlds, even if phone manufactures like to claim otherwise.

1 Like

My initial take is “defense attorneys all rejoice”. AI enhancement of fuzzy photos might well open up new legal challenges, whether they’re legitimate or not.

Completely unrelated but the “Enhance” scene from Blade Runner instantly came to mind :smiley:


The “enhance defense”. It has a nice legal ring to it.

1 Like

To me, this quote echos the words of statistician George Box.

All models are wrong, but some are useful.

I learned this quote from a data scientist, but it applies equally to photography.

In photography, we’re trying to coerce a dynamic, three-dimensional scene into a static two-dimensional rendering. By definition, this process is lossy and we lose both detail and context in the process. There is no perfect methodology or equipment for achieving this, and so every photograph we take is in some way distorted from reality. At best, each photograph only conveys a single perspective at a single point in time, one of many different perspectives and points in time that might have been chosen.

Often (but not always), the goal of a photographer is not to represent a precise reality, but rather to convey some specific information (maybe a feeling) that is deemed useful, relevant, interesting, or just beautiful.

The operative question, in my opinion, is not whether or not a photograph is “real”, but whether the information is seeks to convey was manipulated with the malicious intent of misleading the viewer. Most smartphone “enhancements” are not maliciously intended to mislead the viewer.


I think Picasso said that all paintings are an abstraction even the most realistic ones

I suppose the difference there being if you have the time to learn drawing skills you will have a choice

Picasso seems out of favour because of feminist criticism which may be justified up to a point although there are dangers of judging somebody with the values of a different time

1 Like

I used to not care for Picasso, but now he is my favorite artist (for the past 20 years or so).


Knowing things about an author can make it very difficult to enjoy the work. Art and culture in general relies on speaking to the viewer. Knowledge can completely change how that voice is heard, you can’t really choose to ignore information when experiencing art. Those who can are probably not really engaging imho.

This isn’t about Picasso in particular but I think it’s unavoidable that cultural works are affected by what their authors have done.

1 Like

Me too.

I should say - there are some commercials that strike a good note.
like this one

And some that turn me off
like this one

At the end we will end up with this (no matter how much massaged by AI it is)

not this - genuine artisan craftsmanship that can be admired


Oo, yummy! (the second one!). :slight_smile:

1 Like