Ai and the aim to deceive

[p.giannakis]

I did something that I promised that I wouldn't do; argue online about subjective issues. To cut a long story short here is what happened.

On a Facebook group about photography, someone posted a picture that the claim that they shot. Looks like the perfect picture, the right place at the right time. 2k likes later and handrend of comments later (by the way group rules forbid heavily manipulated images) I pointed out that the picture has some "errors" in it.

First of all, light source (sun or moon, unable to say) is on one side of a foggy bridge yet there are no shadows on the other side. Parts of the image have shadows yet others don't. There are parts of the bridge that are lit by a light source coming from the top (second sun?).

I got a lot of comments suggesting alternative scenarios. I am not convinced. The author posted another picture to prove it is genuine, I post a crop of the bridge to draw your own conclusions.



I have nothing against Ai, I am very tempted to try it myself. I know my daughter's will find it fascinating. There is just something about using Ai with the aim to deceive thought.

 

[Zuiko-logist]

Welcome to the new world. This will only get harder to detect but at least for amateur photography,is it any worse than the in camera or darkroom techniques we used for film and paper?

 

[splitimageview]

Well other than the fact that AI generated images are not photographs… 😆😆

AI-generated images have the potential to deceive the viewer into believing they are actual photographs, especially if they are not labeled or watermarked as such. This can have serious consequences for trust, privacy and security in various domains.

For example, a popular Instagram photographer confessed that his photos were AI-generated after gaining thousands of followers who believed they were real1. He admitted that he felt guilty about the deception and that he did not disclose his use of AI because he feared losing credibility and followers.

Another example is the regulation of generative AI tech in China, which aims to ban deception and promote growth. The new rules require that any AI-generated media must be clearly marked with watermarks or other labels to distinguish them from real photos or videos2. The regulation also prohibits using AI-generated media to harm national security, public order or personal rights.

There are some ways to spot an AI-generated photo, such as looking for artifacts, inconsistencies or unrealistic details3. However, these methods may not always work as the technology improves and becomes more sophisticated. Therefore, it is important to be aware and critical of the sources and origins of the images we see online.

What do you think about the potential for deception with AI-generated images? Do you think there should be more regulations or guidelines for their use?
--Bing/ChatGPT

 

[p.giannakis]

is it any worse than the in camera or darkroom techniques we used for film and paper?

Although I see your point, at the same time i wouldn't photograph a landscape, print it in canvas and argue that i painted it. I think that is where deceit comes in.

 

[p.giannakis]

Thanks @splitimageview , very interesting.

 

[telenous]

I share the sentiment and I steer away from photos that seem to me manipulated to a point of no return. Not to to moralize, they just don't excite in me that which I find interesting, or even urgent, in photography. Where that point of no return lies, though, is sometimes difficult to say, so you have to go with your instinct. At least with film there's always the negative to fall back to. A 'here it is, that's what was photographed' kind of thing.

 

[Zuiko-logist]

Although I see your point, at the same time i wouldn't photograph a landscape, print it in canvas and argue that i painted it. I think that is where deceit comes in.

I agree.

 

[Tom R]

Welcome to the new world. This will only get harder to detect but at least for amateur photography,is it any worse than the in camera or darkroom techniques we used for film and paper?

Your argument implies that these are differences of "degree" (of manipulation, intervention, etc.), whereas many argue that this is a difference in "kind." In other words: human manipulation differs from non-human manipulation. In one case we have a human being manipulating, etc., elements to create a particular object. In the case of "AI" we have a program, a "tacit algorithm" actually, assembling elements to create the object in question. This "algorithm" has no intentionality or consciousness whereas a human being arguably must have at least the intention and consciousness of the world to create the object in question. [A fine point here: such "tacit algorithms" arise during the computation and are basically unknown to entities outside of the algorithm. Depending upon the kind of algorithm we're discussing, we might describe these algorithms as "emergent behaviors" arising out of the data. It's unclear that the algorithm itself could describe the algorithm it is employing at the time of the computation... ]

To make this distinction clearer, consider the hypothetical case of how one would one respond to damages incurred by objects created solely by algorithm versus those same objects created by an individual? Does your "new world" have an ethical, legal, and moral framework for addressing these and similar issues?

Does any of this matter? You tell me. My second cup of coffee needs warming and then I have photographs to be made.

 

[Sonnar Brian]

What camera and lens did the photographer state was used?

 

[p.giannakis]

Your argument implies that these are differences of "degree" (of manipulation, intervention, etc.), whereas many argue that this is a difference in "kind." In other words: human manipulation differs from non-human manipulation. In one case we have a human being manipulating, etc., elements to create a particular object. In the case of "AI" we have a program, a "tacit algorithm" actually, assembling elements to create the object in question. This "algorithm" has no intentionality or consciousness whereas a human being arguably must have at least the intention and consciousness of the world to create the object in question. [A fine point here: such "tacit algorithms" arise during the computation and are basically unknown to entities outside of the algorithm. Depending upon the kind of algorithm we're discussing, we might describe these algorithms as "emergent behaviors" arising out of the data. It's unclear that the algorithm itself could describe the algorithm it is employing at the time of the computation... ]

To make this distinction clearer, consider the hypothetical case of how one would one respond to damages incurred by objects created solely by algorithm versus those same objects created by an individual? Does your "new world" have an ethical, legal, and moral framework for addressing these and similar issues?

Does any of this matter? You tell me. My second cup of coffee needs warming and then I have photographs to be made.

I think these all matter for everyone. I read @Zuiko-logist comment as an empathic agreement to what I said, focusing more on the harsh reality and possibly similar discussions held in the past about Photoshop etc...

 

[jsrockit]

I do not stress this at all. If these fools want to make cliched pictures with AI and try to pass them off as photography simply to feed their ego, then go for it. For those of us who love photography, we all know that using a camera and the post processing (whether analog or digital) is a big part of the fun. That said, I think AI created imagery will be its own genre and some artists will use it in unique ways that sets it apart from photography.

 

[ooze]

I think Hayao Miyazaki gave the perfect answer. All summarized in one sentence at 1:12 here:

 

[p.giannakis]

What camera and lens did the photographer state was used?

This wasn't within his replies. He just posted the same picture in colour saying that it is the moon and that light comes from a halogen searching. No mention about where this bridge is either.

 

[splitimageview]

Maybe he used one of these... 😆

 

[telenous]

^^ Suddenly, 'overengineered' takes on a whole new meaning

 

[nightlight]

Maybe he used one of these... 😆

That looks pretty - what is it?

 

[splitimageview]

For those AI photographers that prefer SLRs...

The model looks rather glassy-eyed, and that thumb?

 

[p.giannakis]

That looks pretty - what is it?

Nothing other than a picture created with artificial intelligence

 

[GMOG]

That looks pretty - what is it?

I think that is an AI generated image of a camera that does not exist!

 

[nightlight]

I see ! 🙂 perhaps AI is the cure to GAS!