I am not big on utilizing large AI models such as "text2img," "img2img," and "vid2img" in direct architectural implementations beyond conceptualization. Yes, they are likely to replace the entire rendering process within a year or two. However, focusing solely on this aspect would essentially transform architects into rendering specialists (i mean, thats the current state anyway). Faster renderings do mean more time for ideation, but AI can expand our ability to perceive ideas and, in turn, enhance our current vocabulary—which I have many issues with, as my followers might know.
One of my greatest concerns over the past year and a half is how these big models have made great leaps in photorealism, which has come at a significant cost and has impacted the "creativity" of these outputs. This is something that both confuses and troubles me. I'm constantly thinking about quality, technicality, bias, and influence.
To adopt either a wholly optimistic approach or an entirely pessimistic one would be both superficial and lacking in understanding. No one knows how significantly these tools will influence what we do or in what direction. All we can do is observe, document, and critique.
The City - Wicked Emergence
Unconventional form-finding through non-architectural processes.
soundtrack: "Come here" by Kath Bloom.
#ai #architecture #formfinding #stablediffusion