Inspired by a previous post in this thread, I've been playing with AI generated images. It's been helpful in looking at different material combinations for a speaker base I'm designing. But it's also great fun to plug in different material finish combos and see what type of loudspeakers the AI creates. I find a number of them beautiful, even if not always "best practice" engineering in design. The AI creates a particularly lustrous sense of light and materials. A bunch of examples:
View attachment 279492
View attachment 279494
View attachment 279496
View attachment 279497
View attachment 279498
View attachment 279499
View attachment 279500
View attachment 279502
And I think this is, or could be, the promise of such rendering beyond a lot of the handwringing about its shortcomings and dangers. Presently it's a pretty cool way to inspire ideas, directions, you may not have considered.
Remember, these models are not conceived from a passionless HAL somewhere in the ether. They are LLM "large language models" that represent a synthesis of human thought and design from as deep into the web as the programmers spiders are allowed to roam. It gives one the ability to tap into an amalgam of human thought, to the extent that we are expressing it on the web today. That can be frightening. It can be vapid. But it does represent a significant part of who we are.
I compared it once to the way I approach a recipe for some dish I'd like to make. I'm 62. Pre internet, I used to take a book off the shelf and look for a recipe for, for instance, spinach pie. I was, with no further guidance, pretty much confined to the exact ingredients and measurements printed there if I expected to turn out any reasonable example of spinach pie.
Today is different. I go online and read a dozen or more recipes, easily and quickly, and then mentally assemble a picture of what is common, and optional or adjustable, to each recipe. This enables me to pretty quickly come up with something unique to my tastes and sensibilities that would otherwise have taken a lot of trial and error to come up with on my own.
With the advent if AI I see this process I just described being condensed a thousand fold.
What we should be most concerned about are the feedback loops. But that's in the future and kinda off topic at present. What results when the algorithm starts feasting on its own content? Flaws and all?