Thought sparked by a recent exchange w/ @Ted_Underwood: one reason, I think for the success and popularity of image generators over AI text generation is that we have a higher tolerance for what one might (with analogy to the visual arts) term abstract expressionism in pictures./
With the images generators, that which is not strictly photorealistic (not their strong suit) often appears merely stylized: “surreal,” or “like a Monet.” Or else just cool and weird./
By contrast, with text generators, most of us (barring those of us who read LANGUAGE poetry in graduate school) have a relatively low tolerance for prose that deviates from discursive norms. We want normative grammars and normative higher level structures./
Put another way the imprint of the algorithm on a machine generated image brings us unknown vistas, takes us to where the wild things are. By contrast, with prose, the mark of the algorithm is just a fail, a reminder of how far the tech needs to go before it can achieve mimicry./
There’s a useful analogy here in sensory perception: it’s a well known phenomenon that most of us have a much higher tolerance for visual distortion than audio distortion. We’ll watch a grainy television station (remember those?) but we can’t abide a record skipping./
Anyway, just some more musings on the new paragone between word and image.
ps. This is also the reason why canonical human-authored #elit never really found a popular audience outside the academy. It’s not the links (Twine proved that), it’s the high modernist aesthetic it styled itself upon.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
