Photos looking at you: conversation with an artificial intelligence

Discussion in 'Member's Photo News' started by photoriot, May 21, 2016.

  1. I was getting overwhelmed by the different ways of organizing my pictures to show, and decided to build an artificial intelligence to sort it out for me, picking the next photo based on minimal guidance from me: similar, different, or random. Most of the effort has gone into the 'similar' option, assigning keywords to pictures and incorporating image analysis software. As the site associates pictures on similarity, you try to figure out what joins one to the next, so it is like you are analyzing an AI's dreams, or doing a photo crossword puzzle. To keep it fresh, the AI is running a live DNA dynamics simulation in the background, which your clicks exert forces on, and which influences picture selection, sort of like an artificial heart for the Tin Man of Oz, responding to your presence. (You can also watch the DNA at 1 frame per second.)
    Have you accumulated so many good photos that you can't decide how to show them? If enough people are interested, it may be worth commercializing.
  2. You need detailed business and marketing plans. Without these, you will go on floundering in the dark, as you appear to be doing now.
    To my mind, you have not really shared enough information in your post for us to engage with you. I have read what you wrote no less than three times, and I must admit I am still at somewhat of a loss to figure out exactly what you are doing, or want to do.
    You need to go back to your drawing board, or desk, or laptop, and redo your project summaries. Then share more with us. In other words, you must help us so we can try to help you.
  3. What info would you like?
    I guess a business plan could start with seeing whether people would want to put their photos on a site like this.
    If you can answer these two questions, you could be helpful I think. "Have you accumulated so many good photos that you can't decide how to show them?" "Having looked at, does it seem like an interface you would pay for as a solution?" Ok, three questions: "How much would you pay per picture?" I think you'd want a minimum of about 1000 pictures for the magic to happen.
    If the concept doesn't make sense to you, please try reading the material on the About page and let me know if it connects any better. It is pitched to people browsing the site.
  4. Another selling point is that the picture selection is fresh, so if you were to reset and explore your pictures again you'd see them in a different order. I've put a lot of work into this, because while developing and testing, I've looked at my photos more than I ever thought possible. My prototype just picked the next photo at random, but this mode soon wore me down. So I've spent almost a year developing strategies for associating the pics in meaningful ways, after mulling it over for about 10 years, while programming in a few different industries and getting ideas here and there. For example, the 'brain' part acts somewhat like an online ad auction (that chooses the ad for a web page), with photos competing to be shown instead of ads.
  5. I guess the fact that you are not getting a lot of feedback tells you something about the level of interest for your AI app.
    I checked it out and can honestly say that I would not at all be interested in it, esp. for a fee. I just don't see the added value.
  6. Thanks for the feedback Juergen. Maybe you'd need to see your own pictures through the lens of the site to see any value, if then. I guess a volunteer photographer would give me an opportunity to explore another user's experience, in case there is some general value to be gotten from the technology. Any takers?
  7. Retrying, png not liked.
  8. Here's a pretty graph, indicating some sort of hyperspace structure within a set of pictures and reminding me of clouds or smoke. The height of each point indicates the distance of match for the current picture with the Gist algorithm, as one choice follows another from left to right. The distance depends on the uniqueness of the photo within the limitations of the algorithm. You can imagine that above each point are the thousands of worse-ranked photos for that competition, like ranks of ghost soldiers clamoring to fight, each with a number assigned for that particular competition.
  9. I try to avoid click bait, and that's what all this sounds like to me. So I am, after reading what you wrote, entirely disinclined to go into your site and check out or read anything. It just doesn't grab me at all.
    My apologies if I am wrong about the click bait. However, this said, to me your postings are so vague, and you seem to be floundering about without really saying much of anything , that I had initially suggested you do business and marketing plans so you could then identify some sort of direction as to what you are doing, or trying to do, and then perhaps decide if your "project" has any validity or, to be blunt, any usefulness at all. Artificial intelligence? Are you using this term to try to get other photographers to post photos to your site?
    As well, I agree that it seems there is little or almost no interest in your site or what you are trying to do, from other members of this forum.
    At this point, I have said more than enough, and I prefer to just drop out entirely. Good luck to you with your project, whatever it is.
    Please consider all I have written entirely as "constructive criticism".
  10. Thanks for giving your opinion, JayDann. It reminds me of when I first told my family about email - no way were they going to get sucked into that crazy stuff, when the post office and fax machines and phone existed. If anyone ever adopts it, eventually you may find something in it.
  11. If I was advertising a patented product for protecting you against scams, THAT would be click bait :)
  12. Though in full disclosure, I do as much as I can with actual clicks to add variety to the picture selection.
  13. Looks like no takers. For the record, in case anyone else wants to pick up the torch, here is the rest of my fiendish plan, if I could only generate income: add more layers of artificial intelligence (two or three more people needed ideally, or at least one to listen to me) to build something that really feels alive, speaking in pictures the way the eyes of a dog speak, which you can build a relationship with over time. This would be a basic personality upon which might be added the more verbal aspects of AI, and which ultimately might become your trusted agent because you would know it intimately enough to know if someone was tampering with it. With that, either I could take over the world directly because everyone would trust me for having thought of it, or take over Disney Studios (by applying it to film editing, perhaps) and thereby the Trump AI and thus the world. It all seemed so easy until dashed my hopes! (Though to be fair, even the organizer of the local AI meetup, responding to a shorter description today, wrote "dude what are you talking about?")
  14. Here's a thumbnail description that you can send to your friends and family:
    The idea is that you pick from three options: similar, opposite, and random, and for similar and opposite an AI chooses the next picture for you. When it's working well, there is a brain teaser effect from figuring out what is similar (opposite is less interesting over time, but a good change of pace), and generally it encourages you to analyze and compare pictures. It's a complementary approach to organizing pictures by time and place, and more fluid than searching your pics by keywords that you have to think of and type in one by one.

    Ultimately I would like to develop the AI so that it feels like it's talking to you with pictures, so it and all the pictures it contains would become as immortal as humanity. In a way, it's a response to the fact that, without an easy interface, the past will slip away much faster when we're gone.
  15. Obviously, I could not resist returning to this post, to check the comments...
    Bill, you are obviously very keen on this. Well and good.
    I still suggest, back to the drawing board.
    Also still suggest, detailed business and marketing plans.
    The way to bo. Business does it. And your posts sound like pure business to me.
    Email? Mate, along with the mobile phone, the scourges of the late 20th century... Think Facebook! I rest my case.
    As stated in my earlier post, the best of luck to you with all this.
    Spouse Two said on reading all this, "you don't need to get involved with artificial intelligence. You were married to it for 10 years before you met me."
  16. Hmm.. I wonder if Spouse 3 could improve on that :)
  17. Now that I think of it, JayDann may be on to something - I might get more visitors if I say there are interesting ads (or please test my ad display), although people would be disappointed when they found there were none.. unless I'm using reverse psychology, which has always worked in the past..
  18. See 'combining pairs of pictures' in Philosophy of Photography for a discussion based on the rock photo and a matching one from the city.

Share This Page