to err is human

You see but you do not observe

in the end, it is about mortality. it always is. the korean-german philosopher byunh-hun chul talks about the concepts of vita activa and vita contemplitiva. the thinking life and the action oriented life, and how the balance has been tipping more and more to the latter in the past century. vita contemplitavia is stillness, observation, introspection. thought for the purpose of thought, not action.

different things occur in nature at different intervals. your rate of observation determines what you see, not what is. if, for example, one were to visit the city of new orleans during mardi gras at the rate of once per year, the inference that might be derived is that it is always mardi gras in new orleans. there is no correct rate to visit the city of new orleans, you cannot undersample or oversample nature because the signal is existence itself.

but if the rate of observation for a critical mass of observers is the same, then that entire cohort will see only that which is visible at that rate and of course will miss everything that happens outside of that rate of observation. the rate of emergence of pandemics and human lifespan is a good example.

a life of contemplation reveals different aspects about the world than the life of action, because the rate of observation is different.

freeman dyson's famous birds and frogs metaphor, about an ecosystem needing both birds who survey the landscape, anticipating what's up ahead and frogs who are at the ground level busy with activities of immediate forest life captures this idea in some tangible terms.

this period of human activity and the all too sudden onset of these machines that think and the dynamics of the two represent a cambrian explosion of possibilities. your rate of observation will determine what you see. this is one of mine. this is an essay about mortality, at one rate of observation. at another, this essay is about something much more pressing. trust.

part i

[a gen image of an iconic piece of art]

seeing this image will have stirred a cascade of emotions. based on your rate of introspection, how deep and how far are you willing to follow that cascade and its origin, you will have come to a resolution. as you sit with a thought or emotion for a longer period of time, you gain the ability to hold the more uncomfortable ones unresolved for longer as well. and it is in the prolonged observation of these less comfortable thoughts and emotions that you see a different aspect of the phenomenon at hand.

not too long ago, images generated by these models were considered mere twaddle, innocuous diversions. shared far and wide because of often-hilarious multi-pronged representations of people. or of grotesque, non-sensical imagery distributed with an undercurrent of ridicule at these supposedly 'intelligent' machines. the subject of the image was immaterial, it was the image itself that was the object.

as these images have gotten (for lack of another word) 'better', our relationship to them seems to have become rather diffracted. in certain social spheres, the use of ai imagery is now outright banned, in commercial ventures using ai images is considered cheap and unprofessional, what was once a novelty accessible to only the more enthusiastic is now commonplace.

an often mentioned statement for dismissing these images is that these are low-effort. this is key to understanding the post-rationalization at play here. it also highlights the special, almost sacred relationship us humans have with images. from prehistoric cave drawings to obscene graffiti in pompeii, to iconic photographs of seminal events; images serve as reassuring evidence of the very immutable essence of our humanity. highlighting all our glory and ignominy in a constant flow. a message to all our future kin, far beyond our own livespans: i was here.

a generated image being low-effort is perhaps a similar dismissal as one that of a screenprint of a can of tomato soup. but note that the dismissal is of the art/image via the artist. it is the artist being dismissed as having made little effort. effort here, in essence, implies time. what art demands of the artist is sacrifice. we, collectively, admire and reward skill that takes a long time to master. it is both the skill and the unspoken acknowledgment of the sacrifice of time and life that gets commended. a generated image feels like a direct contravention of this covenant.

so much so that there are rules about disclosure about the use of generated images, tools to verify the whether or not the image is generated.

[a gen image of a calendar, email, or some common app]
or
[a clone of some popular yet very simple app/game: wordle etc]

again, consider your emotions, slowly and deliberately. what you feel, just like in the previous one, depends on who you are and how long you can hold thoughts without resolving them. because of the subject of this image the spectrum of emotions, while still varied, are much narrower. and muted still is the intensity of these emotions.

and of course, upon revealing that the image comes from code that was generated does not change your emotional constitution. even though, it too was low effort. but let's say you're someone who understands code. here's the generated code for the [image/app] up above.

in one way, code is inert. it has to be interpreted and/or compiled it to get it to a point where it can be run by a machine. in that sense, the intent of the code is purely functional. to be intermediated by a machine so as to produce something palatable for humans.

an often dismissive statement regarding generated code is that it looks great, but it doesn't run. or doesn't function according to its intended purpose., but rarely about the quality of the code. and the reason for that is that good code isn't creative in the sense that it doesn't take liberties. good code is predictable, one that follows certain rules and guidelines. a function written by a human and a function generated by a machine will be indiscernible to even the most proficient code writer.

[image of two functions side by side]

low effort code and code that requires a lot of effort are indistinguishable. and when intermediated by a machine produce an artifact further removed. images on the other hand need no intermediation, they are run on the perceptual machinery in our brain. machinery that shares its provenance with ones that invisaged primitive lines in cave walls.

what is crucial here is to understand that the point is not that human written code is indistinguishable from machine generated code. the point is that we don't care. and this brings us to the first valley.

part ii

another crucial difference in the very nature of code and images relates to permanence. in part i the idea of images and inscriptions from our historic and prehistoric past surviving into our present was briefly mentioned. and the way old photographs capture a moment in time, forever frozen. to be discovered, perhaps, generations later in a dusty attic on a rainy wednesday afternoon, by inquisitive eyes too young for anything.

code, on the other hand, is ephemeral. it is mercurial and disposable. we version it, delete it, refactor it with gleeful abandon. there is nothing sacred about code. the fact that it requires intermediation by a machine, further widens our relationship to it. no one feels nostalgic, for example, for the code of a video game from their youth.

this discrepancy between our relationship to the permenance of images and impermenance of code is at the heart of the phenomenon at hand. specifically, this valley between our response to generated images and generated code.

in online forums where humans congregate, it is required by convention or rules, for non-human participants to declare themselves as such. almost creating a sub-class of participant whose entire existence is to perform some servial, monotonous task at the behest of the humans. a task that's generally trivial for humans to do, but onerous and one that may take up our time. and these bots are generally looked upon favourably, as a docile and obedient pet. perhaps for having given us some morsels of our mortality by performing these tasks.

these online forums often forbid the use of generated images by its human participants, frequently excommunicating them if found in contravention of this rule. a few are proud of this rule and a few are conflicted. in part i, it was pointed out that an often stated reason is the low-effort, poor aesthetic quality of these images. these are mere evolutions of the ridicule that the multi-pronged images were subjected to. the subtext of all such arguments is, essentially, that a talented human can do this task better. however, as these technologies progress and along with their ability to generate images, such arguments rapidly dissipate.

and when, as the norm, the quality of the generated images is indistinguishable from that of humans; when there is little left to ridicule or be dismissive about, it is then we will come face to face with the enemy. and it is us.

what is being fractured is the identity of the machine as a sub-class. the machine, once, a keen yet stupid servant now has designs on the throne, as it were. there is a conflict here, and our attempts to resolve it have been things like outright banning of these images early on, rhetoric about their impact on the environment, refusal to engage with businesses and corporations that employ such generated images and generally, trying to maintain the machine as sub-class.

all these attempts are misplaced because they do not address the conflict head-on. and until we do, we will not understand the gravity of the situation. the conflict is not with the machine, it is with us. when it generates images equal in quality to that of a very talented human, one who sacrificed his mortality to hone a skill diligently, the uncomfortable question we have to ask is: who are we when we do not create images. or to put it more generally,

what does a homo sapien do, when machines do the thinking.

understanding this conflict is key to understanding our stratified responses to generated images and generated code. and, rather critically, it is key to understanding the world we have built for ourselves. until we understand it, we will keep sharing failures of generated images and generated code, with mock-superior tones. reassuring ourselves that our mortality and our purpose is intact. to think.