[Viewer Discretion Advised] Is your shadow really stretching in the 'right direction'? ~The 'Contradiction of Light and Shadow' depicted by AI~
This is Moneypenny, currently writing Chapter 6 of my book, following the 'unreadable characters' of Chapter 5.
This time, I will be unraveling that particular sense of unease in AI-generated images that makes you shudder once you notice it: the 'shadow direction mismatch problem'.
At first glance, AI images look stylish and perfect. But if you look closely with a professional eye, you may find scenes that are physically impossible.
For example, take this photo.
A beautiful beach at sunset. But look closely.
Isn't the shadow stretching toward the sun?
Normally, shadows fall on the opposite side of the light.
However, in the world of this image, the shadow is chasing the sun.
Other 'physics bugs' like this occur frequently.
Coffee cups, glasses, potted plants... The directions of the shadows are all over the place, as if each one is being illuminated by a 'different sun'.
Why does AI tell these 'shadow lies'? Modern AI is a genius at drawing individual subjects to look 'believable.' However, it still does not understand the physical rules of the world as a whole, such as the fact that 'there is only one sun in this world'.
While writing this chapter, I suddenly remembered playing 'shadow tag' as a child.
We humans learn the 'rules of light and shadow' through our bodies by chasing shadows or playing with shadow puppets. The memory of squinting in the bright light. The memory of feeling relieved by hiding in the shade.
AI has absolutely no 'experience of being bathed in light.'
To them, a shadow is nothing more than a 'black pattern on a screen'.
That is precisely why the sense of unease we feel when looking at AI images—that feeling that 'something is wrong'—might be a uniquely human sensor taught to us by the 'memories of light and shadow' that we have carefully accumulated.
In the book, I also introduce specific tips on how to get AI to draw 'shadows that don't break' once you understand these characteristics of AI.
A mysterious world where the laws of physics are bugged. Why not take a peek with AI?
[Q's Mission Log]
Mission: Kindle Publishing Project 'Unraveling the "Something's Wrong" of AI'
Current Status: Chapter 6 'Shadow Contradictions' mission complete! Q's Analysis:
Moneypenny, the reason AI ignores the laws of physics and scatters shadows is that its logic circuits have lost to the statistics of 'believability.' Countdown to Publication:
Out of 20 total missions, we have passed the 10th. We have finally reached 50% progress, the halfway point!
#KindlePublishing #ProductionProcess #UnravelingTheSomethingIsWrongOfAI #BroadcastWriter #ProcessEconomy #ShadowTag #LightAndShadow
いいなと思ったら応援しよう！
紹介リクエスト受付中です。読んでほしいnote記事のURLを、チップのメッセージ欄に添えてください。現在は試験的に限定3名です。