Clever framing. AI regulation becomes urgent when the government might have trouble keeping information from its own citizens.
But, China might beat us to “win AI”.*
*[讽刺]
1 Like
A couple fascinating demos, thanks. as Ryan pointed out before discussing spaceships, anything done on a computer will be up for grabs by AI. But it comes back to rules vs complexity as to when. We do a lot of stuff away from our laptops that will survive but isn’t really why we got into design originally. think design research, usability testing, etc.
The other argument is that human nature is too unpredictable for AI to grasp. Understanding and translating aspirations, desires, fears and irrational decisions into physical things with enough value to pay for is what we’re best at. I think it’ll be a long time before AI will be smart enough to do that (and really if it gets that smart it’s got more important problems to solve)
So I’m still an optimist, we have probably the most time to evolve the profession. Fuzzy, non-linear and aesthetic centric (with the caveat that VBL implementations go away) will still be human when most of the rest of the business and engineering process aren’t. 10-20 years perhaps more if gov’t regulation kicks in
1 Like
A few more updates. Fusion360 and Claude showing their skills with ID (furniture, electronics). Good design remains in the details. For now. The parametrics in Fusion are indeed impressive. I wonder about editing them however. These examples are light on showing iterative improvements. Something I’m currently working on in my work.
The parametric model in Fusion looks impressive but I wonder if the AI has essentially downloaded an existing model from grabCAD (or a similar website).
Although the output in the other example is currently unusable due to inaccuracy it suggests future potential.
I work in construction and if I could get AI to reliably model up everything from the architects’ intent dwgs and those from other contractors it would save a large chunk of my time. However from the example shown it looks like I’d spend more time checking and correcting the AI generated model than modelling it myself.
AI currently appears to work for generating novelty output but nothing that can used to manufacture/construct anything useful. The danger is that it encourages people without the required knowledge of the work being carried out to trust the AI output which could have disastrous consequences. Just think of all the badly cobbled together AI T&Cs but in physical form.
4 Likes
The last two sentences you posted have been my experience (design consulting) and it’s exhausting. So many people who have no idea what they’re doing are confident in their “generations” and insistent that it can be made.
I’ve been referring to it as Dunning-Kruger Overdrive
4 Likes