AI and Visual Truth

Any image generated optically with a camera lens might be acceptable in many different ways. It could be an entertainment or a crime scene. Evidence or an accident. Theater or documentation. What differentiates the “value” is not the photographer, nor the photographer’s intent. It is actually differentiated by the use that the viewer wants to make of it.

But the RELATIONSHIP between the scene and the viewer (aka meaning) is always only one of three relationships that can be important. The other two are the relationship between the photographer and the viewer (aka expression) and the relationship between the photographer and the image (aka intent). Entire industries and careers have been built and evolved based on cultural, social, or commercial agreements about those three relationships.

In all three relationships, the agreement could be “non-fiction images”. Some people will be interested in and motivated by that option. It will be tough times ahead for those who will not consider other possibilities. But the world of other options will not obsess about the word “photograph”. The word “image” is going to suffice (it already does). Image makers can help themselves by making their part in the expectations readily verifiable. Viewers will have to learn that distribution channels, not the medium, should be where they place their expectations.

All that said, most of the time, there is a practical reason why cameras would be preferrable to a pencil or a computer: creating a visual record.

Records exceed the notion of “documents” in that a record is always attached to a concern about something specific, not just a view of it.

There are several ways to create a visualization of some event or condition thought to be relevant. But most of them might not be reliable as a source of their real-world facts.

We go through this any time there is a measurable distance between the existence of what we’re concerned about and our direct witness of the concerning thing. The appearance of the image is asked to match the experience we would have if we were there.

Cameras, at their initial introduction, were instantly famous because of the relatively high fidelity their images offered — not only versus being at the scene, but versus any prior existing imaging method.

This is why anxiety over the trustworthiness of a photograph is 90% based in achievable optical resolution, and only 10% in why it was made.

But ever since the screen was invented, and at least since the paintings of Seurat (pointillism) and Lichtenstein (Ben-Day Dots), we have been able to generate images that at sufficient distance “resolve” to whatever they showed, because we had control over where any of their dots went — at least comparable in labor to weaving.

Having enough automation now to make dot rendering nearly effortless hasn’t changed much of anything. It’s just that the dots are now electronic — nearly microscopic and extremely easy to individually reposition any number of times by using instructions that run on computers like calculations.

So in effect, with our new instruments for generating an image, more emphasis is placed on those of us who “play” them to be able to prove that we are the decision makers and the instructors that programmed and approved the display.

The credibility of the imagemaker, per the maker’s expressed intent, is what is at stake.

New A.I. image generation is completely indifferent to anything other than its instructions. The more we are willing to accept an image maker’s explanation of why they made the image the way it looks, the less it matters to us that the maker has fabricated the image without recording being a part of the process.

The future value of photographic recording will be hugely diminished in situations where recording as a proxy for direct witnessing is not necessary — not required by our concern to verify something’s appearance.

But we can also understand when that requirement is not negotiable. The requirement to proxy witnessing is high any time our concern is represented by saying “IF X is true, THEN…”.

In that formula, which presents our key concern, a visualization of “X” is subject to any verification of its fidelity to something actual.

But if viewers don’t have a particular concern, then the value of the image is left to however much the viewer is interested in whatever way it provokes them.

The usual way to anticipate that is in terms of where the viewer chooses to go look, and why. For example:

In this picture, incentives, expectations, and inhibitors are seen related to each other. But their overlaps also suggest that what crops up in one area can mistakenly be evaluated in terms of another area. As a simple example, treating hearsay as if it is the result of research is easy to do if we are not openly cautioned against doing that.

See ongoing observations about AI in Visual Arts at www.artdotdot.com.