Listen on the go, available on Apple Podcasts - apple.co/3MCK7hq, Spotify - spoti.fi/45eT5ci, RUclips Music - bit.ly/uhstudiopodcast or wherever else you may listen to podcasts. Search for UH Studio Architecture
🙌 Very relevant discussion! Definitely worth bringing this topic to the forefront of architectural theory, workflows and philosophies. It's here to stay, and it's our responsibility to integrate the technology thoughtfully into our craft.
You are absolutely right in context of using ai as iteration tool, or as a tool in a process. In its current stage. But it's not really a point of contingency. The question you are avoiding is who owns the end result. Aka who owns the data. Even if you will be able to iterate 1000 times faster. Does it really matter if both the result & your expertise would be fed to a machine to improve it? It's cool that manual labor & management jobs aren't going away due to their high specificity. But if you don't own your own IP, than the specificity of your white collar work will be going down with every specific piece of work you do as you produce more data that can be used to train the ai that will replace you. It's kind of a stretch to place your bets that your boss or clients would protect your data if a potential cost reduction will be too good. It's the same with 3d models nowadays. There is just not enough data to scrape & collect as with images due to it's size & complexity. But it's getting there. Also the fact that someone's made an overstatement, doesn't mean he is wrong. If you didn't got hurt doing something, doesn't mean the stuff you do is safe, you know.
That's a good question, but also no different than asking "Who owns the the IP of doric and ionic columns?" All architecture is based on historical references either to support or react against a specific references.
While the actual uses of Ai today can mimic architectural images and tasks, it can't ever complete an architectural workflow or even assist 100% in any task. That's because the training of these models are not done using architectural workflow as I puts and admisable outcomes as output. There for how would these tech can help us in something that he doesn't recognise. One some one trains a model in this way we could reopen the topic, but then that could be said about any other trait or job. The universal models today aren't more then hobby's
Listen on the go, available on Apple Podcasts - apple.co/3MCK7hq, Spotify - spoti.fi/45eT5ci, RUclips Music - bit.ly/uhstudiopodcast or wherever else you may listen to podcasts. Search for UH Studio Architecture
🙌 Very relevant discussion! Definitely worth bringing this topic to the forefront of architectural theory, workflows and philosophies. It's here to stay, and it's our responsibility to integrate the technology thoughtfully into our craft.
You are absolutely right in context of using ai as iteration tool, or as a tool in a process. In its current stage. But it's not really a point of contingency. The question you are avoiding is who owns the end result. Aka who owns the data.
Even if you will be able to iterate 1000 times faster. Does it really matter if both the result & your expertise would be fed to a machine to improve it?
It's cool that manual labor & management jobs aren't going away due to their high specificity.
But if you don't own your own IP, than the specificity of your white collar work will be going down with every specific piece of work you do as you produce more data that can be used to train the ai that will replace you.
It's kind of a stretch to place your bets that your boss or clients would protect your data if a potential cost reduction will be too good. It's the same with 3d models nowadays. There is just not enough data to scrape & collect as with images due to it's size & complexity. But it's getting there.
Also the fact that someone's made an overstatement, doesn't mean he is wrong. If you didn't got hurt doing something, doesn't mean the stuff you do is safe, you know.
That's a good question, but also no different than asking "Who owns the the IP of doric and ionic columns?" All architecture is based on historical references either to support or react against a specific references.
Awesome !
Thanks!
The value of online freelance services for architects has gone down dramatically. Meaning clients are paying less.
While the actual uses of Ai today can mimic architectural images and tasks, it can't ever complete an architectural workflow or even assist 100% in any task. That's because the training of these models are not done using architectural workflow as I puts and admisable outcomes as output. There for how would these tech can help us in something that he doesn't recognise. One some one trains a model in this way we could reopen the topic, but then that could be said about any other trait or job. The universal models today aren't more then hobby's
Take a look at lookx.ai