Text to Image

[Boydstun]

There is an AI group on Facebook at which, a few days ago, I made the following post. In the first Comment to the following, I'll show a response it got from another member of the group.

"It has been my experience that when humans make films containing settings of nature or industry from scenes described in literature, they do not measure up to pictures in my mind that I have when reading the text. Pretty sure that will be the same when AI makes scenes from word descriptions, however well. I suspect the images in my mind from text are of only flashing, partial determinacy and tied to feelings from past experience. Such episodes could hardly be expected to be attained from a 3rd person view, I notice."

[Boydstun]

Here is the response:

Quote

Your words above strike me as an oddly original description of a commonly experienced phenomenon: The thing depicted visually by some other somehow fails to meet the standard of what our personal imagination had already dreamed up. Everyone has experienced this, yet what I find so remarkable in your explanation is how you specifically chose to describe the reason someone else's version of the imagery can be expected to fall short of what we imagine. You associate the creation of personal mental images with personal feelings and experiences. Up to this time, I had always assumed that some other person's visual representation of a given image is rejected by us for no other reason than that it simply didn't agree with our mental image. But now, with the help of your thoughts, I see that there may be a bit more to it than this. You seem to be suggesting that when we dream up a mental image, perhaps based on some text description, we aren't just creating a visual, but more importantly, are creating a UNIQUELY PERSONAL visual, meaning that in creating this visual, we are connecting with personally meaningful past feelings and experiences, and that this process is, at least in part, HOW we construct the mental image in the first place. I never before considered the role of past personal feelings and experiences in determining our current choice of mental imagery. What your thoughts on this help me see is that it then makes perfect sense to suppose that no one ELSE'S depiction of a given text content can completely satisfy us. That this other person got our personal vision WRONG, then, maybe isn't the main thing. That they could not have possibly got it RIGHT, on the other hand, since they don't share our personal feelings and experiences, seems to be the more important factor in why the images they present fail to match our personal versions. They fail to match them not by chance, merely, but because their version can't be directly based upon our unique feelings and experiences.

 

[StrictlyLogical]

This reminds me of much, and brings up a thought or perhaps a sentiment...

a certain asymmetry...

 

Although "We cannot know things-in-themselves" is flawed

it is a certainty that

"things-in-themselves cannot know We... only We do." is true.

Edited October 25, 2023 by StrictlyLogical

[tadmjones]

Yesterday I heard mention of a phenomenon called aphantasia , the self reported lack of 'innner' imagery. Oddly it seems "I have it", odder still a few weeks ago I was discussing this with my wife! But I thought the fact that what I was trying to explain to my wife and her reporting back of her phenomenal experience of imagination visually, in comparison to what I was describing didn't 'line up' was due to semantic misalignment.

I did not realize that when most people report 'seeing' 'things' in their mind's eye or imagination that they were not speaking metaphorically. I always assumed 'visual imagery' was a collectively agreed upon ambiguous concept to describe an 'inner' understanding or cognition(?) of an imagined 'thing', and not a 'quasi-actual' visual image.

If I were prompted to close my eyes and imagine say a pink elephant, I only 'see' the dark behind the lids, there is nothing 'there' that appears anything akin to visually apprehending a pink elephant, but the 'idea' of a pink elephant is present or experienced, perhaps better described as almost a state of awareness of being predisposed to 'accepting' a nonheretofore 'appearance' of a 'visual experience' of a pink elephant .. actually it is rather hard to describe, especially because I never thought I would have to describe this aspect of experience , going off the assumption that all youse all was just speaking metaphorically!

I do dream 'visually' and I sometimes mistake the experience of having read a novel as having watched a film of the story, but i don't seem to be able to bring up visual imagery 'on demand' , so maybe on the spectrum as it were, lol.

So , yeah AI imagery ain't never gonna live up to my expectation, but only because the bar is too low !

 

 

Edited October 25, 2023 by tadmjones

[StrictlyLogical]

1 hour ago, tadmjones said:

Yesterday I heard mention of a phenomenon called aphantasia , the self reported lack of 'innner' imagery. Oddly it seems "I have it", odder still a few weeks ago I was discussing this with my wife! But I thought the fact that what I was trying to explain to my wife and her reporting back of her phenomenal experience of imagination visually, in comparison to what I was describing didn't 'line up' was due to semantic misalignment.

I did not realize that when most people report 'seeing' 'things' in their mind's eye or imagination that they were not speaking metaphorically. I always assumed 'visual imagery' was a collectively agreed upon ambiguous concept to describe an 'inner' understanding or cognition(?) of an imagined 'thing', and not a 'quasi-actual' visual image.

If I were prompted to close my eyes and imagine say a pink elephant, I only 'see' the dark behind the lids, there is nothing 'there' that appears anything akin to visually apprehending a pink elephant, but the 'idea' of a pink elephant is present or experienced, perhaps better described as almost a state of awareness of being predisposed to 'accepting' a nonheretofore 'appearance' of a 'visual experience' of a pink elephant .. actually it is rather hard to describe, especially because I never thought I would have to describe this aspect of experience , going off the assumption that all youse all was just speaking metaphorically!

I do dream 'visually' and I sometimes mistake the experience of having read a novel as having watched a film of the story, but i don't seem to be able to bring up visual imagery 'on demand' , so maybe on the spectrum as it were, lol.

So , yeah AI imagery ain't never gonna live up to my expectation, but only because the bar is too low !

 

 

Just a second… how would you visualize a spatial problem?  For example imagine placing furniture so that it fits a room but also imagining it in place to determine if there is flow and if it will work functionally long term?  Do you not visualize it i.e. see it in your mind’s eye?

If someone described “An isosceles triangle pointing straight up, its horizontal base longer than and resting on a square, a smaller vertically oriented rectangle resting in the square at its base, a small circle inside and to one side of the rectangle” do you see anything in your mind’s eye or would you literally have to draw it first following this description as if they were a set of instructions?

[tadmjones]

SL

With the furniture example and going from actual past experience and thinking how that process worked , I'd say I do it more by 'dead reckoning'? on the spot?, I'll estimate the length of a sofa , usually with a cue like how many floor boards equate to its length and spot out the equivalent length in a different position in the room for fit.

Thinking about geometric question seems more like an 'inner' visualization, but I'd rather describe coming to understanding the particular arrangement of shape in not necessarily 'tactile' but more an intuitive feeling of what that shape 'would be', but definitely not with words associated with visual imaging.

 I suspect some of the distinctions or differences in relating the phenomenon may be semantic, and perhaps there is a limiting aspect in the recall of the process.

 

[Boydstun]

Tad, you're correct, I've heard, to use "all youse all" for an audience of three or more.

You're sounding normal to me, just with above-average introspection to report.

Be that as it may, I thought I'd add that I have certain flash-sketch images of scenes in memory that were set when I first read the scene in literature. The final scene of Hank Rearden in his office with the farewell salute from the mills and Galt and Dagny etc, etc, in the abandoned rail tunnel. I don't seem to have any particulars of what their faces look like other than being man's or woman's and being White, but such images and movements in the scenes as I do conjure do not change upon re-reading the book. And I doubt they'd get displaced by any filmmaking of those scenes.

At least one dream researcher Hobson proposed that the stories we make in our dreams are handicapped attempts to make sense of images that are being randomly presented to the dreamer. I've noticed that Freud was right in one thing at least about the phenomenology of dreams, for what it is worth, and that is that frequently objects or actions or problems in dreams are residue of real waking experiences of the preceding day. I think sometimes the day residue can be from experience of television or film images. I still recall a dream from before 1980 during which I became awake. It was just of a nude blonde woman, details of her body indefinite, but she was vertical and as without support and amidst wind-billowing satins and shears, all of it in bright slightly golden light accompanied by my feeling of the greatest preciousness and attractive beauty. (I don't recall if I was having one of those involuntary you-know-whats.) I entered it into a poem I wrote a few years ago called "Dream to Sleep". By her face and hair, I knew well enough that she was not any particular person I'd known at all. Indeed, I'd say she mostly matched ads from 1960's television for Breck.

I have a curiosity, Tad. Imagine having a glove on your left hand. In your mind, take it off and turn it inside out. Pull it onto your right hand. Does it seem like it should fit the right hand? Does it seem a verdict is not reached by this mere imagining? It was in the early 1990's I think that psychologists and neuroscientists found the sequence of brain activities that support human abilities to turn objects over in the mind. I don't know if there has been similar research on inversions such as in the glove transformation. (I've tried it with a real glove, and it fits. That might be useful someday.)

Edited October 25, 2023 by Boydstun

[StrictlyLogical]

5 hours ago, tadmjones said:

SL

With the furniture example and going from actual past experience and thinking how that process worked , I'd say I do it more by 'dead reckoning'? on the spot?, I'll estimate the length of a sofa , usually with a cue like how many floor boards equate to its length and spot out the equivalent length in a different position in the room for fit.

Thinking about geometric question seems more like an 'inner' visualization, but I'd rather describe coming to understanding the particular arrangement of shape in not necessarily 'tactile' but more an intuitive feeling of what that shape 'would be', but definitely not with words associated with visual imaging.

 I suspect some of the distinctions or differences in relating the phenomenon may be semantic, and perhaps there is a limiting aspect in the recall of the process.

 

OK... so... DID you imagine a drawing of a house, such as what a child might draw?  If not.. you are mystery to me, but at least my understanding of your words (which I can't quite fully believe) would be validated.

If you had to draw it before "seeing" it, you win.  If you did visualize a house... I'd say you and your wife are not so different.

[tadmjones]

Through the years I’ve cut up my fair share of graph paper for modeling to get a bird’s eye view of furniture arrangements. 
 

I am intimately familiar with our most used room floors , being addicted to the idea that bucket and rag is the only true and acceptable cleaning process. I know where every chip scratch and blemish is, but if go into the garage , I know I know the particulars , but the idea of ‘seeing’ any concrete element ‘from memory’ in a anyway approaching a literal visual experience is somewhat alien. 
 

It could be that I am exaggerating the sense of a literal apprehension of a visual image that is generally experienced by you (and all youse else).

The phenomenon I’m discussing isn’t as much surprising to me , as is the scenario of stumbling on it, lol. An odd comment lead to the topic in conversation , an idea I never really considered discussing and then hearing it mentioned as an offhand comment in a podcast, is more shocking lol.