AI, Again

AI, AGAIN

by William Lulow

Note: This is an article I wrote a couple of weeks ago but, with all the hoopla about AI lately, I thought I would post it again, with a couple of additional comments and maybe one or two caveats.

I’d like to start with the notion that AI (Artificial Intelligence) doesn’t actually CREATE distinctly new or innovative photographs except perhaps for images obtained from drones or places that humans normally have a difficult time accessing. Because, ACCESS is a fundamental requirement of all photographs. The camera MUST BE THERE and hopefully with a human or at least directed by one. It has certainly shown that it can take human ideas and build on them and in that sense, create something new that is some kind of synthesis of two or more already created ideas, pictures, words or paragraphs. And, I have already experienced it making better sense of something I have written, sometimes! And, it certainly has proven that it can alter images, sometimes! (I have often had to go over elements that it has altered to get them just right)! But I still think it is incapable of bringing brand new, unheard of concepts to the fore. It can’t stand there with a camera and summon up a picture of something it can’t see. By this I mean it can use algorithms and instructions to formulate a “concept of a composition” (to use a now well known phrase), say. But it isn’t capable of human thought as we know it. A human can give it some parameters about a topic, image or even a concept and it can return something, probably grammatically correct if it’s the written word or some image of something about which it is asked. But, IT IS NOT HUMAN, by definition! It can only imagine what might be there. And that’s all well and good for what it can do. Humans might then be able to fashion this into some recognizable art form, but it hasn’t created any original art. It can’t muster a new thought or idea. And, it probably can’t write a novel like “War & Peace” on its own, with no human assistance or reliance on anything in the history of the planet. It has a limited database which seems to be limited by previous human input, although it seems to be able to invent something within its scope of learned knowledge. And that’s the point isn’t it? AI seems to thrive on history, what went before! It also seems to be able to partner with other machines to solve some mathematical problems. But none of this I would consider original thought in the human sense. Since it is not actually human, I find it difficult to imagine a robot searching for a place to stand to get a view of some object in order to realize an image of that object with a specific purpose in mind. I am reminded here of  Archimedes famous statement, “Give me where to stand and I will move the Earth,” knowing full well that such a place doesn’t and cannot exist! It seems to be able to complete tasks it is given, however. 

As an aside, I have seen pictures and videos of what is being referred to as, HUMANOIDS standing with a camera and presumably taking a picture. But, how would it know where to aim the camera, or what to view using a robotic “eye” unless it was given some parameters? Most humans need some parameters to produce great images. I have also seen these things (they are not humans), taking orders at a lunch counter. But there, a human would be giving it an order. AI might be able to be programmed to know what a “hamburger” is. And it might be able to recite a “menu” of items or foods. It also might be able to cook something from a pre-programmed menu. But the program would have to be written by a human, initially. I’m not sure I can imagine a robot then being able to do something about a burger that was too well-done, say. 

Computers will do whatever humans tell them to do, nothing more and nothing less. So, if some computer output is faulty or doesn’t produce a desired result, first check to make sure that the input was correct. The old expression “garbage in, garbage out,” is an apt analogy here. 

Certainly AI has produced many things already and it’s just in its infancy. But, generally speaking, as an artist, I don’t like it. 

I am in the business of creating images that please people. Images they can react to and that can show something, sell something, convince someone of something truthful (hopefully), and above all, document history, people, families and the human condition. I have been doing this almost my whole life! 

There are those who would differ. They might say that a thought is only a mathematical construct consisting of certain parameters. I don’t think so. It is one way of looking at the world, but people have thoughts every day that often don’t conform to strictly mathematical concepts. 

In the world of photographs, we have been able, with the help of computer applications, and before that, other means,  to create images that don’t really exist in the natural world or universe, but those are constructs based on things that were already created by nature. I’ve used this picture before. It was made at Jackson Hole, WY, but the clouds were from Switzerland:

There are those who would argue that photographs must be real to be believed. But that would eliminate all kinds of constructs from the world of photography. My view is more akin to if it’s made with a camera on film or digitally, it’s an original photograph no matter what is done to alter it with a computer or in a darkroom after it has been recorded. Ansel Adams’ constructs from the darkroom were seldom what his camera actually saw. They were images he probably had in his mind AS HE WORKED HIS DARKROOM MAGIC on them, in an effort to enhance and fashion them into his own, particular ideas! 

The thing about photography when it was first introduced in the early 1820s, was that it was an accurate, visual representation of reality. We got to see EXACTLY what Abraham Lincoln looked like, rather than a painter’s interpretation. My guess is that portrait painters before 1800 made fairly accurate likenesses of what people looked like. That, after all, was their stock-in-trade! And they were pretty good at it. So we have come to expect photographs to be much more true to life than painting, at least. 

With that said, I have been making computer generated solarizations, or at least what Photoshop calls solarizations of certain landscapes as well as people. I used to make them in the darkroom by exposing a print to a brief flash of light as it was in the developing tray. The light would cause the light-sensitive crystals in the photographic paper’s emulsion to tend to migrate to the areas with the greatest exposure (the whites), thereby causing a kind of reversal in the developing process making a positive print look like a negative or a combination of positive and negative. It always fascinated me. Now, Photoshop has come up with a way to do this electronically. The beauty of it, to my mind, is that as you are using this filter, you never know quite what it will achieve. Here is one example. This is a view New York City at night that I subjected to several filters in Photoshop in addition to solarizing it. You could almost never come up with such an image using traditional color film. But, it is a work created first from an actual photograph, then from an idea about using the tools a computer application can provide. :

This is an image of some dancers in Braga, Portugal. Same technique with a different subject. But you can see here that I was seeing how this idea could apply to using people as well as landscapes:

So there is nothing artistically wrong with creating images that might not exist in nature, but I don’t know that a machine could create something like this if it wasn’t directed by a human first. 

The takeaway here is that AI is useful when and where it is appropriate. But because it is “artificial” it will never replace original ideas and concepts put forth by a human. 

Additional comments:

There are some real concerns expressed lately about AI and it’s ability to discern for itself, actions it could take to disable human control. I first heard of something like this with the film “2001: A Space Odyssey,” by Stanley Kubrick. In that film a super computer named “Hal” malfunctions and decides not to do what its human handlers ask it to. It’s interesting that its called a “malfunction” because those happenings can usually be controlled by the humans that created it. 

But what if AI enabled machines can teach each other to subvert the uses for which they were intended? What if AI went “rogue”? These are the kinds of circumstances that many scientists are wondering about today. Some are even contemplating that AI enabled machines can replace humans altogether, at some point. They are already creating online “chats” with customers for a great many businesses. The intent of this article is to question AI’s ability when it comes to making portraits and other images necessary to the world of commercial photography. 

At this point in time, I can’t conceive of a robot that can stand in front of a scene, say, decide that the lighting, actual scenery, and point-of-view would be sufficient to produce an award-winning image. The robot would have to be able to analyze everything it sees, determine what use it would want to put it to, figure out where to place itself, then take the picture and determine what, if any, changes it would make to the image to make it truly a “work of art.” As I mentioned, it’s relatively easy these days, to conceive that a machine could make an image. But a human would have to program it as to where to stand and what kind of image was wanted or needed. I can imagine a machine programmed to do such things, but not to come up with the original thought. I can even imagine machines communicating with each other to discuss various procedures originally introduced by humans, but not original thought. 

So, again, ARTIFICIAL INTELLIGENCE  is just that: artificial. Humans have to realize this and be prepared for its limitations.