• Photos are never a concrete representation of the reality. Photos are being pre-processed by image processor already and we also got Photoshop. One can even fake a film based photo if he knows what to do. The proliferation of image generation models and impainting models make the access easier but image manipulation tools always exist.

        •  Kache   ( @Kache@lemm.ee ) 
          link
          fedilink
          8
          edit-2
          4 months ago

          There’s the practical distinction between “everyone can do it with some dedicated intent” (so few actually bother) vs “everyone can do it on a whim”

          •  arglebargle   ( @westyvw@lemm.ee ) 
            link
            fedilink
            English
            24 months ago

            Admittedly a computer in everyone’s hand is new. But corel paint, for example, was 12 years old in 2003. People were basically making memes and creating scenes that never existed on a whim and for the lulz back then.

            And were much, much, better then these stupid examples!

      • I remember in the UK show Utopia from 2013, a government frames one of the characters for a school shooting by perfectly doctoring security footage to erase the actual hired shooter and replace them with a specific kid. And they do it all in a matter of hours. I remember thinking that tech was unrealistic, probably impossible. The best Hollywood VFX experts would need a week or more to make it that believable, and even they would need a ton of reference of both the kid and the lightning. Purely fantastical tech.

        And now, here we are…

    • Agree. Now I will say when faced with the decision between ethical issues vs profits big tech smashed that profits button without beating an eye, but the tools were always going to be made. It’s just too bad they didn’t stop for 5 seconds to think about how they could be mitigated.

  • @Gaywallet I have a couple thoughts on this:

    1. This seems like a way that device attestation could worm its way further into our devices. Right now Google is trying to watermark AI-generated photos as AI, but you could easily go the other way - if a photo hasn’t been manipulated, it’s signed with a key that is locked down to device attestation. What, your phone is rooted? That’s kinda suspicious - how am I supposed to know your photos are real?

    2. Short of that, though, I suspect that the most likely consequence of this is the videos will start being increasingly seen as necessary for true proof, since those are harder to fake - for now, at least. And of course, there will be a lot more misinformation on the internet, especially in the short term while awareness of this catches up.

  • Our basic assumptions about photos capturing reality

    Your assumption, not our. I thought its an old topic. Also what does reality even mean? What we see is the reality? Does this mean animals don’t see the reality, because they see it differently? I see this article is about AI, but even then we had image editing even before digital and computing ages. Images were photoshopped before computer in Darkrooms. But that’s not all. The entire process of capturing an image itself can’t capture the “reality”, only an interpretation (such as eyes do).

    Every photographer knows (or should) that photography is not always about capturing reality; its art. The framing of the image, meaning only a part can be seen, is an artistic choice. The shutter speed is an artistic choice. The depth of field of the lens, the zoom of the lens are artistic choices. What part of the image will get the focus?

    The cameras setting for ISO, meaning the sensitivity of light is also an artistic choice. The placement of light to get a specific effect or even a camera flash to get certain amount of light and maybe freeze a frame that would be impossible to see otherwise. And the choice of white balance affecting the coloring. Camera sensors can’t capture the entire spectrum of light or the complete dynamic range we can see with our eyes. That means there is compromise at this level too.

    I’m not done yet. The camera doesn’t capture an image as we know, such as JPEG. The raw format of a camera is not processed yet (but it has some processing builtin, just not too much). Meaning the image is very blurry and pixelated when converted to pixel format we can see. It has to be sharpened to make photos look as we know. Either automatically when saving as JPEG or later with a software that can handle RAW image formats of the camera. That is all an artistic choice.

  •  arglebargle   ( @westyvw@lemm.ee ) 
    link
    fedilink
    English
    54 months ago

    Those examples are so bad. You do not need AI to do any of those. That’s just cut and paste.

    Hell, we have had fake reality video overlays that are better than that on our phones for over a decade.

  • I don’t think many people would go so far as to consider photographs “the linchpin of social consensus” but even so, many of the assumptions we tend to make about them have always been false ones. I’ve seen things in real life that later featured on the front page of newspapers. The difference can be more stark than is often appreciated. Like any powerful medium of expression they can capture truth or they can obscure it.