AI

steve_shubin

A mini rant on AI — LOL

When I think of AI and CAD, I don’t really get all that excited

I could see AI basically acting as a replacement for the mouse. In other words, something that uses voice commands instead of mouse movement and clicks

I guess if some one were doing basic construction such as repetitive apartment units that are not all that fancy, that AI could probably lend a helping hand.

And maybe for certain engineering, you could punch in all the specs for soil conditions, and once you design the structure, it’ll help you with what type of footers you’re gonna need and what type of steel structure is going to be strong enough

But I do not see AI replacing designers and engineers such as John McClary or architects in the same manner that a lot of the writers from Hollywood fear being replaced

With regards to architecture,  even though nowadays, code is more consolidated across the country, I imagine there are still a lot of local jurisdictions that have their own take on the code as to what year they are adopting. And many jurisdictions likely don’t adopt a particular years code lock stock and barrel. Local jurisdictions have numerous variations on a particular year.

Then there’s the matter of ergonomics as to a particular clients size and aesthetics — the particular wants, desires, needs and idiosyncrasies of that client, as to exactly how they want their structure built

I already have some pretty fancy architectural CAD stuff to where once I lay out the basic walls in a 2D manner, it’s one command, and I’ve got all the framing in 3D in one click — the headers, trimmers, cripples, sills, king studs, bottom and top plates. So there’s already a lot of automation so to speak

But if there’s much custom design work being involved, I just think that I could probably mouse click my way through the program just as fast.

Take for example one very small aspect of an architectural CAD design

A door

The specialized architectural CAD program that I use has 23 different dialog boxes just to set up a door — in which the first dialog has 16 different pull down menus or fields for numeric input, where all that could give you I don’t know how many different choices.

When you start to do the math, I may have in the tens of thousands of different options on how to set up a door. Probably more like in the hundreds of thousands of different options or variations on door design

so I’m curious how AI is going to help me

And remember — a door is one very small consideration when it comes to designing a building.

When you put it all together, I probably have to choose from millions of different things 

But I have to know how to quickly hone in on what I want to go with.

Hard for me to believe that AI is going to make it a whole lot faster

A lot of times it’s not the CAD program that slows you down, but rather, it’s figuring out exactly what it is you want and how it’s all gonna mess together. It’s the solving of the overall design and the details — that can chew up the clock

I think there’s always gonna be the need for thinkers. Human souls and minds that know how to create beyond the box. And I think Onshape is the type of program made for that type of person

MichaelPascoe

AI is a tool like calculators and spreadsheets. Use them properly, and they have the potential to speed up your productivity. Similarly to calcs and sheets, you have to know what you are doing, they won't work unless they are guided in the right direction.

Midjourney, for example, is an AI tool designed for art. Use it within it's scope with proper guides and it can save quite a bit of time. It would have taken me hours make this 3D printed shoe concept art by modeling in Onshape then rendering it with render studio. Instead, I had Midjourney do it in 30 seconds:

There are AI tools for art, writing, graphic design, etc.. It is only a matter of time before a decent one is introduced for CAD. Since there are so many types of CAD, it may be a while before specific CAD genres (like solid modeling) are helped by AI. Though Onshape has all of the right ingredients for it: Cloud based, API, custom apps, custom features.

That being said, the human brain is a fancy organic computer. Based on our current level of AI, it will probably be 30 more years before we will publicly see human replicable AI. With this general info, I would assume that privately, it already exists.