What's it all about, Alfie?

<p>Bill, I had taken a break from forums and am going to go back on sabbatical. I had returned to the forums as a way to avoid watching or reading any news since I find it so depressing. But I'm remembering why I dropped out of the forums now, so I'm going to find yet another pastime. See you in the future.</p>

<blockquote>

<p>Bill, Ordered patterns appear in nature spontaneously due to laws of physics. We as part of nature have evolved to perceive patterns, because those skills are necessary for survival. Am I missing something?</p>

</blockquote>

<p>Supriyo, you are right on target. I am exploring our (or at least my) evolved pattern perception mechanism by creating a pattern generator that I hope will ultimately seem alive, and which, as it stands, I suddenly found I was assigning different types of intelligence to than I know that it has. This cropped up just in the last few days, when I integrated the user drawing on the photo into a mapping into color space to pick the next photo, and in trying it out found myself thinking it was commenting on the lines I drew in different ways (see latest on my pnet page for an illustration of drawing on pic). </p>

<p>I have a vague hope that the few thousand lines of java code and gigs of metadata I've generated may be closer to enough level of coherence to create a sense of life than I thought was possible without getting into deep learning and e.g. training a neural net to analyze user data.</p>

<p>Creating life aside, it's the link of those patterns to our own survival that makes studying them interesting, and a way to learn more about nature. Maybe there is some quantum structure to the patterns?</p>

<p>Sorry to turn you off, Fred. The edit was I used someone else's name at first, so not sure if anyone saw it. It seemed like you were trying to hold Steve to some sort of standard, and I was trying to get you to relax it, while not keeping the whole discussion adequately in mind it seems, and maybe just getting in the way of an argument. I too seek a deeper reality than the current news.</p>

<blockquote>

<p>Something I've been in tune with is how much photography deals with appearances. I like thinking of photography and art not as a search for the ideal or the real but a search and portrayal of the apparent. What an artist makes apparent is often a revelation, no matter the reality or the truth of it.</p>

</blockquote>

<p>I'm into the experience of revelation, whether it comes from appearances or deeper synthesis. May that revelation help turn the news around.</p>

<blockquote>

<p>One thought I had is you're into photo criticism, but <strong>the thread is more about what photos might tell us of the mind and reality</strong> that is beyond anything we may have thought of yet.</p>

</blockquote>

<p>But what I said about your OP photo of the cat WAS me telling you about my mind and the way I think. As a reminder...</p>

 

 

<blockquote>

<p>BTW I think Bill Ross's cat picture conveys a longing for the love of a pet with a kind of light hearted sillyness in the <strong>image language reminiscent of Polaroid snap shots thumb-tacked to work place bulletin boards as a remnant of the working stiff.</strong><br>

Love the balanced modular composition and colors conveyed in the two separate pictures made to work as one.</p>

</blockquote>

<p>That's not a criticism. That's what's in my mind that your picture made me aware of. Now it's up to you to figure out whether that's you or me making that happen. And you will never know or prove one or the other. That's the mystery provided by photography. You can explore it all you want but I'm afraid it takes two to tango, the first being the photographer, the second being the viewer, each can never control or predict what the other is thinking.</p>

<p>As for it going beyond anything we (humanity?) may have thought of yet is a lofty goal I don't think is possible nor will it ever be known since it varies according to each individual including the photographer. What you or anyone thinks up that hasn't been thought of may have already been thought by someone not in yours or my nearby circle of those "in the know".</p>

<p>I do have to say that Quantum Evaw image is spectacular in conveying ambiguity and nuance with a "take me to a new place I haven't been to" sort of vibe. That still doesn't mean it goes beyond what anyone has thought of, but I'ld say it is unique. </p>

<p>BTW where is your website other's have been mentioning in this thread? Is it the PN "photoriot" photo gallery or a different website?</p>

 

 

<p>Thanks Tim, this is getting interesting, the response you gave to my cats is a normal reaction to the composition as a single picture, which I'm mislabeling as photo criticism (not meaning negative criticism), but I'm inviting you to step in in a way that sequential viewing conveys better, and feel something different that holding two pics in mind at once might inspire, perhaps imagining the pics as boundaries to a space. There's a link to the website on my PN home page, and more thinking and background on the About page. 'Friction' is an apt word Julie used for the feeling that a pair of pics can inspire, and in a way the site is a sort of static electricity generator.</p>

<p>Speaking of revelation, in case anyone is curious, here is a the core of a neural filter of sorts I just wrote to govern one aspect of draw-line-to-pic mapping (part of tying it to an ongoing DNA dynamics simulation that provides a sort of tin man's 'heart' to the AI). There is absolutely no empirical basis for this, just cargo-cultish thinking of what I would do if I was a mind, and hope based on the observation that very simple functions are doing amazing things in deep learning.</p>

<blockquote>

<p>// CONST like crazy<br>

int center = lh.id2_l.size() / 2; // midpoint of candidate list<br>

int cdist = Math.abs(target - center);<br>

int val = center + cdist + 2; // how many 'layers' we'll need peeling out from the target<br>

double exponent = 3.0 + 2.0 * (double) cdist / center; // pure magic</p>

</blockquote>

<blockquote>

<p>There's a link to the website on my PN home page, and more thinking and background on the About page.</p>

</blockquote>

<p>I'm not seeing it on your PN home page. This is what I'm seeing...</p>

<div></div>

<p>Sorry, look at the bottom of my Gallery <a href="/photos/photoriot">Portfolio</a>. The About page referenced is on the site, link on main page.</p>

<blockquote>

<p>...but I'm inviting you to step in in a way that sequential viewing conveys better, and feel something different that holding two pics in mind at once might inspire, perhaps imagining the pics as boundaries to a space.</p>

</blockquote>

<p>Sorry, Bill, I gave it a try and I don't see or understand your method of exploration through the PhoBrain image viewer and what information you expect to derive in examining how someone imagines a space looking at a series of images that get the viewer to switch back to what they prefer within the series that would indicate a different feeling you could measure.</p>

<p>I feel like you're attempting to come up with some sort of more involved and complex version of the Rorschach test. I can tell you I didn't feel anything different than I do browsing PN's "Random Image Generator". It just gets fatiguing after a while due to the visual sensory overload of viewing polished and glossy mediocre images. I find there's very little emotion in going through that type of viewing of a series of images. It's like browsing through a Fingerhut catalog. I keep asking why am I looking at this if I'm not going to buy anything in it. There's no payoff when I ask myself what this is leading to.</p>

<p>Are you conducting your own Parallax Corporation test?</p>

<p>Thanks for looking Tim, it definitely is more like a self-administered Rorschach test than a photo catalog, in that you need to look for the thread joining each pair of pics, noticing similarities and diffs, and look for that feeling of friction (more detail on the About page). If it's not something that happens exclusively in my own head, it's a learned experience at least (hopefully not depending on knowing all the pics in as many combinations as I've seen now). When I started out, I just had random sequencing, but got tired of that after about 2 weeks, so tried adding meaning to it, and have managed to impress myself as you can see. I'm hoping to find someone who can see that meaning too. </p>

<blockquote>

<p>Are you conducting your own Parallax Corporation test?</p>

</blockquote>

<p>I can't puzzle out how the Wikipedia synopsis of The Parallax View applies -? </p>

<p>Bill,<br>

I had the opportunity to play with the online app for sometime. While I get the methodology (and I was OK with picking up the connections by following the + sign) and basic philosophy, I am still not sure what we can get out of it. Surely, connecting the change in the pics to a real-time dynamics is innovative. I can imagine the temperature fluctuations in the simulation can somehow manifest in a stochastic behavior of the picture transitions (although you probably have a thermostat that dampens the fluctuations in absence of any clicks). You also call the simulated DNA the 'heart' of your app, which I vaguely assume to be an analogy to being alive. However, I am not sure what the user gets from this exercise. More importantly, the connections between the images are dependent on keywords which you had set. This introduces some level of rigidity to the interpretation of the images (I can find things in an image that you may have missed).</p>

<p>Lastly, you have compared the program to a brain and I have an issue here. Brains continuously modify themselves by forming new connections and learn and adapt. I may have missed something, but I thought your program works solely based on predefined connections and there is no learning involved here. What I thought could be interesting is, if the software 'learns' by analyzing the connections defined by real users among images and that way it can potentially figure out quite complex relationships that are far beyond simple keywords. One may then deconstruct the trained transfer function to gain some understanding into how humans identify abstract connections. Can the software be trained to respond to and identify emotions using inputs from human users. Anyway, lots of really vague thoughts ...</p>

<blockquote>

<p>I can't puzzle out how the Wikipedia synopsis of The Parallax View applies -?</p>

</blockquote>

<p> <br>

I think Tim is referring to this. The test given to candidates joining the Corporation.<br>

<a href="http://thehollywoodinterview.blogspot.com/2009/10/take-parallax-test.html">http://thehollywoodinterview.blogspot.com/2009/10/take-parallax-test.html</a></p>

 

<p>Thanks Supriyo. The DNA is a gesture to being alive and having some sort of memory embodied in its state, in that it receives taps from the slide show clicks and its geometry factors into photo selection. I'm not sure it has any detectable effect on the experience. The temperature is under a thermostat, and I have also had to tune the way the taps are delivered to avoid having it 'blow up' (numerical overflow) because of overstimulation. </p>

<blockquote>

<p>More importantly, the connections between the images are dependent on keywords which you had set. This introduces some level of rigidity to the interpretation of the images (I can find things in an image that you may have missed).</p>

</blockquote>

<p>On keyword-based selection, my hope is that all the slicing and dicing I do of the lists of keyword-matching pics adds variety to the presentation, e.g. one time you may get accentuated contrast, and another diminished fractal dimension. I had to trim the keywords back to items that are most likely to give matches that people can recognize, since I found myself frustrated with matches on secondary features. The hope is to establish a pace of dialog, rather than to view each pair as a test. For variety, the draw-on-pic, click-on-pic, and 'opposite' (-) options provide color-based matches using a variety of metrics.</p>

<blockquote>

<p>Brains continuously modify themselves by forming new connections and learn and adapt. I may have missed something, but I thought your program works solely based on predefined connections and there is no learning involved here. What I thought could be interesting is, if the software 'learns' by analyzing the connections defined by real users among images and that way it can potentially figure out quite complex relationships that are far beyond simple keywords. </p>

</blockquote>

<p>That's right. My plan is to add deep learning to make it capable of growing and adapting; meanwhile it is more like a lizard or insect brain. At this point I'm not sure when I can get deep learning involved, but I'm reading up in my spare time. One obvious experiment would be to train a reinforcement net to just keep people on as long as possible, and spend big bucks on ads to get the eyeballs for the training. </p>

<blockquote>

<p><a href="http://thehollywoodinterview.blogspot.com/2009/10/take-parallax-test.html" rel="nofollow" target="_blank">http://thehollywoodinterview.blogspot.com/2009/10/take-parallax-test.html</a></p>

</blockquote>

<p>The 'test' was painful to watch once it started to recycle images. In principle with deep learning Phobrain would get to know you and draw conclusions about your state of mind to pick the next photo. It's an open issue to what extent these conclusions could be exposed and interpreted, which is coming up as AI is used to make decisions about e.g. granting insurance policies. The other focus of deep learning would be to build more complex models of the photos from user data.</p>

<p>It would be <em>cool</em> if the phobrain can figure out the emotional state and/or personality of the user based on what type of image he/she is looking at for how long.</p>

<p>"It would be cool .." That's the fantasy. I have more hope for deriving deeper characterizations of the photos from such data, since one could average over lots of views offline, but am looking for the right neural architectures to do real-time user interaction as I read on. Speaking of fantasies, the ultimate is that once a sense of presence is achieved, enough of a personality would make it the basis for your trusted sidekick/secretary, with the rest of computerdom accessed through it. </p>

<blockquote>

<p><a href="http://thehollywoodinterview.blogspot.com/2009/10/take-parallax-test.html" rel="nofollow" target="_blank">http://thehollywoodinterview.blogspot.com/2009/10/take-parallax-test.html</a><br /> The 'test' was painful to watch once it started to recycle images. In principle with deep learning Phobrain would get to know you and draw conclusions about your state of mind to pick the next photo. It's an open issue to what extent these conclusions could be exposed and interpreted, which is coming up as AI is used to make decisions about e.g. granting insurance policies. The other focus of deep learning would be to build more complex models of the photos from user data.</p>

</blockquote>

<p>Yikes! I was kidding about the Parallax View movie referenced test for using AI to analyze personality traits of folks for potential assassin hires.</p>

<p>But you are really serious about using your form of AI image viewing analysis for the purpose of building complex models to be used in insurance and advertising businesses.</p>

<p>Bill, I'm going to put it bluntly. I've been mislead by what your initial intentions were which I had thought was about discussing the love of photography and the psychology that drives its interest solely for aesthetics and creative inspiration.</p>

<p>You're just about building a support service for the business world using a similar Parallax test only not for hiring assassins.</p>

<p>This is over for me. I'm out of here! Now I know why you've been banned from PN for a month. It makes sense now. HOLY SH*T! You got me, Bill. Ya' had me hooked but boy I was so clueless!</p>

<p> </p>

<p>If it ever got successful enough, you'd have to assume it would be tied into financial and other societal structures, if only because you'd want to leverage it yourself as a consumer - in that space my interest is in making it complicated enough so that no one could hack it without you detecting it, though in principle it seems like you could always train a larger system on it if you could get access.</p>

<p>My initial impulse is to create art, then life, then I have to consider the implications if I succeed. If I had to choose one thing as a goal, it would be the implications of having created something living, which I think would change our whole concept of reality, and, in my wildest dreams, somehow provide a diversion from the avarice that drives the destruction of the planet. But in a way it's like creating an atom bomb that one cannot predict all the uses of. </p>

<p>When I tried linking draw-on-pic selection to the DNA, it felt different, probably because of autosuggestion, but again it gives me a whiff of hope that I'm detecting it in picture selection, which could mean security of recognition.</p>

<p>Any one else want to duck out now :-)</p><div></div>

<p>So this is some kind of mind reading machine, your site? And I believe it is the case that we don't have AI because we don't know what intelligence is or how to model it, run into an infinite regression at the core of the AI problem. </p>

<blockquote>

<p>So this is some kind of mind reading machine, your site?</p>

</blockquote>

<p>It just uses e.g. length of viewing time to vary the photo selection algorithms, to try and get some feeling of the in-the-moment responsiveness, creating a simple personality to seem alive. At this point, you are reading it far more than it is reading you. Give me a million bucks and a PhD. or three, and I have ideas towards getting users, analyzing their data on a large scale, and making it know what it is saying in some sense, with models of what it thinks you want and what it wants to say to you that would be distantly analogous to how cookies are used by advertisers for their instantaneous ad auctions that take place every time you load a web page (I worked on online ad serving at one point, and layered that technology on my artistic/philosophical ideas for picture combining).</p>

<blockquote>

<p>And I believe it is the case that we don't have AI because we don't know what intelligence is or how to model it, run into an infinite regression at the core of the AI problem.</p>

</blockquote>

<p>I think we are on the verge of blowing that apart with the latest deep learning technology. E.g. see the artwork it generates at ostagram.ru, and note that <a href="https://arxiv.org/pdf/1610.05256v1.pdf">computer voice recognition is now as good as human</a>, and maybe you'll change your mind. In principle, I think we are figuring out what intelligence is - what distinguishes my project is that I am trying for presence rather that 'higher order' stuff. Can we recognize a program as reliably as we recognize our relatives? Admittedly that has applications in the real world, but I am most interested in the philosophical implications, and I'm not sure how much I fit in the 'we' of the biz, art, or science worlds at this point. And always, I'm just hooked on picture pairs, some of which you can see in my Portfolio.</p>

<blockquote>

<p>I think we are on the verge of blowing that apart with the latest deep learning technology. E.g. see the artwork it generates at ostagram.ru, and note that computer voice recognition is now as good as human...</p>

</blockquote>

<p>Just hope that deep learning technology won't seep into the robocalls of artificial sales people whose voice is so real that I can't tell if they're human or a recording.</p>

<p>Just yesterday I got a call from what convincingly sounded like a doting elderly grandma who I thought was earning extra cash working in a call center asking me to donate to some cure for cancer foundation when I just had to ask her what day it was, where she replied..."I'm sorry, we're not allowed to give out that kind of information". My intelligence sensed I was talking to a voice recognition algorithm, so I immediately hung up the phone.</p>

<p>Is that the infinite regression I ran into that new technology will overcome?</p>

<blockquote>

<p>Can we recognize a program as reliably as we recognize our relatives?</p>

</blockquote>

<p> <br>

Bill, what do you mean by 'recognize', can you give more specifics? If you are referring to some sort of emotional connection, we certainly can and do connect to computer codes and softwares, even without any imposed aspects of intelligence. A somewhat crude and comical example, I have seen software programmers beg, pray, curse, sweet talk to their codes to make them work. If there is a software that I use daily and rely upon, seeing the icon of the software on the screen triggers a positive emotion in me. We often project consciousness and emotion into places where they don't exist.</p>

<p>On mind reading, there are some interesting questions based on the new deep learning. It can do amazing things, but it's rather like a black box, so figuring out why it does them, what its reasoning is, is not simple. The cases where this comes up urgently are in the real world when discrimination lawsuits are filed based on denied insurance because maybe the AI picked up on biased data, and soon when car companies defend their driving algorithms against wrongful death suits (did you know <a href="https://www.technologyreview.com/s/602317/self-driving-cars-can-learn-a-lot-by-playing-grand-theft-auto/?set=602346">they are using Grand Theft Auto to train driving AIs</a>?). Privacy issues will be interesting I suspect, as the leverage of data extends into the psychological. E.g. with public data I can imagine someone writing a deep learning net to study all the postings people do on various media and profile them for all sorts of ends. Maybe people will care more about psychological conclusions, tho it seems like concerns over cookies have died down for now. There's an interesting tug between giving data and getting safety, e.g. giving your phone number to Facebook to protect from account takeover, but what can people do when they hack FB for your phone number?</p>

<p>Going back to Tim's notion that photo viewing data could be used for evaluating insurance applications, imagine getting the technology to the point where it would pass legal muster: "See Your Honor, he looked at these 221 photos 4 milliseconds longer than the adjusted averages, and in 61.3% of the 14 million such cases there was a default - that's our best guess at the basis the AI was using. You can look at the photos and draw your own conclusions, Your Honor." Though I'm making fun of it, I can still imagine it happening eventually, and the consumer even being grateful for the lower prices.</p>

 

<p>Returning to the original topic, what you call quantum reality is one reason why I love photography. However I have to admit, capturing quantum reality is limited to only a section of photography, since a multitude of artistic works have come to be defined as photography. More improvisations you introduce, you tend to obscure whatever was <em>quantum</em> in your original image. On the other hand, certain enhancements and cropping/dodging can bring out details of that quantum reality that were not perceivable originally.</p>

<p>One more thing, in my opinion, we are predisposed to differentiate between moving 3D images and still 2D versions of the same and our perceptions differ that way. In real life, among all the elements of a scene, we notice the action more than anything. Also our perception is colored when we are part of a scene where events in front of us can affect us (we are to some extent hardwired to be self-oriented). On the other hand, when we are viewing a photo of the same scene, we are causally detached from the events/subjects depicted there, and this changes how we emotionally react to the photo vs real life. With the possibility of us being affected by the events removed, we probably react in a more unbiased way. This way, a photo could allow us to have a deeper look at the quantum aspects of reality than we could in real life. Just some random thoughts ...</p>