Machine learning creates professional level photographs

Discussion in 'Philosophy' started by movingfinger, Jul 15, 2017.

  1. A somewhat rhetorical question: How much longer before a human is no longer needed behind the viewfinder or playing with the menus and sliders in photoshop or lightroom? If you think this is a silly question then check this out:

    Research Blog: Using Deep Learning to Create Professional-Level Photographs

    As for me I will always prefer a human produced photographic image, always.

  2. How do you know you'll prefer it to something you haven't seen?

    I admit I make the same kind of claim about some foods I've never eaten, but I also know that years later when I do eat those foods, I've felt extreme remorse for all the years I could have been eating what turns out to be yummy. My loss.

    addendum: Supriyo will be along in a minute or two to point out that I'm eating foods that I've never eaten. He should try it. It's good for weight loss.
    Supriyo likes this.
  3. Vincent Peri

    Vincent Peri Metairie, LA

    Hmm... artificial intelligence
    might replace humans some
    day, but it'll never replace
    Great Apes...
    wolfgangarnold and Landrum Kelly like this.
  4. You ask a good question. My reply is that 'photography' for me is not just the final image but it also includes my imagination, awe, disgust, admiration, revulsion, whatever, regarding who it was that took - created - the photograph in the first place. There are computer graphic images, fractals, etc, that I also find pleasing to look at and that I find inspirational, based on what they depict regarding the underling mathematics and algorithms that generated them. I'm happy to have such images hanging on my wall. I'm even willing to call some of them art. But they don't inspire or impress me the same way that photography does, they don't create the thoughts about what inspired the human to point the camera that way, compose and frame, and push the shutter release when they did. All those imaginings are part of what photography means to me and what I like about it. A machine generated image would be like the fractal image and perhaps be pleasant to look at, but, call me close minded or say I'm splitting hairs, but to me they are not the same thing.
    oleksandrk likes this.
  5. PapaTango

    PapaTango Itinerant Philosopher

    Indeed. As any new technology and its outputs/outcomes moves from novelty to general acceptance, we become conditioned to judge any practice or product against that new standard. This is well understood as a practice (see Thorndike's Law of Effect), and a form of operant conditioning. Just think, anyone can take their crappy, mundane iPhone snaps and dress them up to look much better. What? That's already possible? Here I am late to the party again... :rolleyes:

    But seriously folks, how can you tell? That is the real question--as those who will purvey the mechanism of this new AI photographic excellence certainly will engage in some serious social conditioning to make it the new norm. Return on shareholder value and all that good stuff. It's no mystery that such conditioning moves established norms we have accepted as a part of socialization through art and photography. Interestingly, we 'inherit' our tastes, standards, and expectations for fine images from the humanist inspired work of artists and their 'handlers.'

    BS you say? Well then if that is so we can safely say then that such as Stieglitz, Ray, Adams, Weston, Bresson, Avedon, etcetera, did not socialize us to accept both practice, technology, and expectation--what we have come to call the state of photography today. Now it is potentially a series of algorithms that determine the path to defining the expectations. Put this in a device (the newest form of operant conditioning) and we have a winner. Give that five years and we will busily accept that as the standard and see the other as somewhat deviant... o_O
  6. To date, as far as I know, machines (including those with AI) may be able to make selections, i.e., choices. But can they do so with intention or are they limited to acting solely based on their programming? I have no problem accepting the possibility that a machine may be able to produce a photograph based on other photographs by which it is programmed. Whether or not this is based on an anti-machine bias, I have doubts that such a machine can decide whether to darken a particular area of an image, or selectively increase sharpness in another. Although ". . . a series of algorithms . . ." may be able to produce a decent image, I have a hard time calling that a photograph. Papa, this has little or nothing to do acceptance of standards; it has every thing to do with what constitutes a photograph. I consider this the most fundamental issue in the philosophy of photography.
    Landrum Kelly and john_sevigny|2 like this.
  7. Machines with A1 intelligence are our creations in a sense" children of humanity". They like a child will develop to a point they will inevitably surpass their creators having greater potential. Our special qualities will be emulated as they become a creation in our own image.

    A matter of time..
    PapaTango likes this.
  8. Turin Test

    When a human being is unable to distinguish the machine from another human being by using the replies to questions put to both.
  9. Sandy Vongries

    Sandy Vongries Administrator Staff Member

    Then there's that thing some call a soul...somewhere I read the phrase "soulless machine." What will be gained and what will be lost?
  10. Okay, lets look at the soul. All of the books tell us that we were created in Gods image. We are creating a intelligence in our image which naturally follows that it is created in Gods image.

    There is your Soul.
  11. Once A1 intelligence becomes aware of itself it is no longer a machine but a living lifeform.
  12. Sandy Vongries

    Sandy Vongries Administrator Staff Member

    And therefore capable of error...
  13. As much as I understand the technology of AI, it classifies new data from the knowledge of existing data. The classes are derived from preexisting data, it does not create any new class. What that means is, if one feeds an AI machine with thousands of aesthetically conforming photos, it will produce new photos that look like what it was trained with. For example, it will produce neat mountain photos with brilliant reflections, and in doing so, it will carefully avoid that tourist or that pole, even if they add value to the photo. Such technology could be useful for certain commercial graphics that want to avoid being adventurous. How that helps art, I have doubts.

    However I would like to use such technology where it is not meant to be used. For instance, taking an AI machine that is trained to recognize and capture aesthetic landscapes, and pointing it into a trash can may produce intriguing results.
  14. I think it will be a while before AI will posses creativity, originality, and be able to "think outside the box."
  15. And therefore capable of error...Sandy.

    As we are. We are especially gifted at killing each other.


    You are looking at existing technology it moves on.....once we were a little shrew.. Once aware it will follow the ways of its creators but with a data base of all human thoughts and endeavors. Think of the advances in Quantum computers and physics.
  16. PapaTango

    PapaTango Itinerant Philosopher

    Michael, the world outside of forums such as these, academia, and the "fine arts community" gives little thought or a 'tinker's dam' what constitutes a photograph--or any sort of philosophical ruminations. The only imperative is whether they like it or not--and this perception of everything is increasingly shaped by those who are seeking a way to capitalize on the consumer--by influencing and changing their tastes. We are right back to marketing.

    As of this moment, there are hundreds of "dumb" packaged actions available for PS--select and play the macro. Players such as Google can (btw, are already doing so) analyze HUNDREDS OF MILLIONS of photographs in short order. Such analysis can be compared to certain choices and set based rules that are derived from HUNDREDS OF THOUSANDS of great, seminal photographs by recognized artists. So yes, at some point the HUNDREDS of algorithms that may be running simultaneously to shape ONE SINGLE AI produced image that can make the same choices that Adams or Hirsch might have made--based upon analysis of their particular styles.

    Let's add to this some more. We are 'known' commodities. What we like, what we don't, choices we make, so on and so forth. All carefully compiled somewhere in the 'cloud' with more data and deeper levels of it every day. So you post photographs on the web to PN or Flickr? Your own website? What do you spend more time in minutes or seconds looking at? Preference based analysis is coming of age. Some high end stores now can 'look' at you standing in front of a LED adboard, and based upon your sex, age, and style/coloration of dress make certain assumptions--which shape the delivery of the marketing you see on the board. If we personalize this with WHO we are, such development can give us what we like and what we are predisposed to choose. Take this to AI imaging now.

    This is not even remotely about philosophy. Capital, and technology driven by capital cares little about this--the teleology is about more capital. Our "brave new world" it is totally about the extension of technology, and the ability to create capital from the conditioned behaviors we are increasingly corralled into by that technology--all carefully focused to change our expectations, reshape our consumer habits and choices--and extract capital from that process. In the first world America, we seem to acquiesce to this blindly and most exuberantly--as ultimately everything is increasingly delivered into the "happy bubble" of curated interest that we demonstrate while interacting with that technology.

    As a critical point to this, that "curated, happy bubble" is largely driven by AI algorithmic analysis. So what we like is most often filtered for us and delivered according to those analytical outcomes. Have you followed the emerging debate of how Google is weighting the delivery order of net resources according to all of this? As in what you see on the first page of search are things that you are more liable to like and do not conflict with your choice of "bubble world belief system"?

    We have met the enemy, and they is us...
    Last edited: Jul 15, 2017
  17. "We have met the enemy, and they is us..."Papa

    Indeed, maybe its that time for us.
  18. One thing is for sure. A1 intelligence will arrive probably sooner than later. Read what Stephen Hawkins has to say one of our foremost minds of this century.

    How will we answer its questions on why we kill and treat our fellows,and our planet, so badly... and any other lifeforms on the planet..

    Okay, we will say we are sorry. Sorry will it be enough.
  19. PapaTango

    PapaTango Itinerant Philosopher

    It has already arrived, and is all around us! Things currently classified as AI include self-driving cars, 'deep thinking' computational systems (think board games such as chess to strategic defense systems), coherent response speech recognition systems(Siri, Alexa, and others), and robotics. Many of these latter things are grouped under a class called AuT, or "Autonomous Things." Humanistic looking robots are included there too (The Japanese lead in this with human-looking greeters, clerks, and other). As to the latter, there is much dialogue now as to how we 'reject' interactions with such the more "human" they are mimics of.

    All of these things classify as AI under the rubric that they act in ways that preserve their goals and chances of success in executing the tasks they are assigned. Too often, we confuse AI with the science fiction concept of 'self-being' that would occur at a point called "singularity", whereby the platform becomes aware of itself and subsequently possessed of an independent cognition and value system.
  20. I'll need to reread your last set of comments before I attempt a response.

Share This Page