Imagine: “Hey Cici…Create a 10"x14” rectangle with 2" radiused corners. Within the rectangle, lay out three quarter-sized concave pockets equally spaced along the meridian of the rectangle. Texture the flat portions within the rectangle with a 1/10" diagonal hatch."
And it just does the design?
“Hey Cici - put a Mayan calendar into a 10” Radiused circle, carved with a 30-degree Vee Bit - center the whole thing within another circle with a 1.5" border."
The request: Utilize AI models to create design elements within the CC windows…replacing the need to drag/stretch/select/etc. and just type what you’re thinking. Allow for iterative prompts to modify the current state of the design - and continue to allow manual tools to work.
I don’t see how the usefulness outweighs the development costs here. Adding “Provide output in SVG format” to the AI prompt lets you import the result pretty much painlessly.
Iterating the design in CC doesn’t make much sense. The actual image can be iterated just as easily in the AI program, and if you have toolpaths set, I don’t see how the machine could seemlessly change the design and keep the toolpaths sane for a design of any actual complexity.
Personally, I would prefer to avoid adding any AI features. Without getting into the many problems with generative AI, it’s an unnecessary addition and it would make sense to dedicate development resources for other, much more helpful features.
The good news is you can design however you want. AI has some good features but also has some glitches. Drawing up a design yourself has some good points and some pitfalls as well. So design however you want and enjoy the ride.
Not trying to stir up the AI vs. Non-AI controversy - it’s all a question of speed. Plus, the engines are built to be embedded…there’s less development work than it might appear to be. They’ve even put AI into Photoshop - and, despite being pretty good at Photoshop, I find myself using it more and more.
It’s a question of when, not if. The reality is that AI can be a very powerful tool. Not the panacea, but a very useful time-saver.
It’s already available to a very useable extent and it’s free. (I pay for ai so I am not sure if free ai will work…your results may vary) Ask your ai to write the code for OpenSCAD, a free cad program that writes from code. Copy/Paste, F6, download STL. Chat GPT isn’t great at complex CAD but it makes awesome boxes and trays. I have Alibre, Plasticity, Aspire, and Blender but for simple things I do that first choice. So fast, so simple. My wife designed her own cosmetic tray on paper with a ruler and dimensions. I shot it with my phone and uploaded it with instructions for writing the code. It 100% worked and lives on her counter now. The hardest part was drawing it on paper.
I absolutely get that you can use an AI to generate something you can import into CC - that’s not the same as being able to use it within CC - where you could make an adjustment, simulate it, modify it, simulate it again, etc…without having to jump back and forth between multiple programs, import files, etc.
There are always work-arounds - but they are not always efficient for what you’re trying to do. If the thrill is in the use of the tools (as it rightfully is for a lot of people), then sure, we can cobble together a bunch of tools and have a great time on that adventure — but if the goal is to quickly create designs that can cut and be displayed / sold - then the motivations change a bit.
I have read several articles about the way people use AI. A lot of people use AI like google searches. The more in depth your query is the better your result. It takes some practice to get better results. However if you have privacy concerns about sharing pictures and information remember that everything you share with AI goes into its database. What you put into AI is no longer exclusively yours and may be used or show up in other people’s query. If you on the internet everyone is snooping on you anyway. There is currently a commercial that has giant eyeballs looking over people’s shoulder while using their computer.
We are becoming a surveillance society. Flock cameras on the road, your neighbors ring door bells, cameras all over private peoperty. Televisions that report what you are watching. Cookies sending data to a lot of 3rd parties you dont even know about. Smile for the camera.
We’ve been a surveillance society for many, many years…we just know more about it now — and it’s easier to do, since we’re all creating online trails all the time and technology allows faster processing of huge quantities of data. But, at some point, your comfort comes from realizing that you’re just not all that interesting or important to people (unless they have something they want to sell you) - and that you have NEVER been actually “entitled” to privacy.
It was funny to me, that people worried about the gov’t tracking them based on something injected in vaccines, when those same people carried cell phones, used cable TV, used Credit Cards, had subscriptions to specialty magazines, etc. etc. etc. Now AI is the new thing to worry about.
I say, embrace it and use it for your own benefit. That’s why I’d like to see it embedded into CC - where we can quickly complete a design, based on iterative natural language prompts that do not require translation based on @WillAdams-level knowledge of the software (unless you really need it - or want to have it).
Anyone with coding skills can write an API to get this accomplished, and I am sure Carbide3d could do that. It’s just a matter of buying compute. There are only about 10 ± ai’s capable of doing this and yes…it could be wrapped into a CAD program. But the issue for a small company is not that. It’s how do I charge for it. How do I keep it competitive. So what we’re asking for is simply a wrapper…like Meshy for example…that doesn’t actually compute but is sold as a wrapper for major ai. Because the costs of compute are astronomical and (without looking at their balance sheet) Carbide can’t do it. We are going to see a shakeout of all these little ai companies and I hope my favorite cad companies aren’t falling for the glitter when there are 3rd party apps that have already taken the bait.
At this point, I think pretty much everything which could be said on this has been said.
Folks who agree w/ the suggestion may like the top post and the powers-that-be will be able to evaluate based on that expression of interest.
Folks who have examples of projects which they have designed with the help of a Large Language Model and made w/ their machines may post to the Gallery:
Folks who want to learn how to use the software are free to ask, either here:
Smile for the camera.