Spatial intelligence in AI is on the cusp. After language, LLMs and then pixels, AI research is about to explode into the 3D world. This video is highly theoretical and granular. These CalTech scientists really get going on 3D applications of spatial intelligence at the 19 minute mark.
interesting explanation of using 2D pixels to create 3D models. they specifically state the intention of making AR and game environments more affordable.
the previous video is similar to other design engineering software but what I conclude is that any well-defined process with quantifiable results can be automated.
so what about us? for ID this might include Ergonomics to start with but much of human factors over time.
Also, Sustainability could be included perhaps performing an LCA from a BOM for example.
even if AI fulfills all these functions, they will be good at the basics but incomplete and need human experts to hammer out corner cases, and they will be separate functions needing human expertise to fully integrate into a final product.
My question is if AI can do all the grunt work how can an entry-level employee gain the expertise specific to that industry or product line?
Zhu definitely helps put things into perspective. Thereās also the part where we canāt actually predict how fast AI will become practical.
Hi everyone,
Iāve been following the discussions here, and there are so many insightful conversations around AIās role in supporting creative workflowsāI would love to pick your brains more on this topic! 
Also, apologies for the bit of solicitation here, but I thought this could be a great place to connect with like-minded folks who are passionate about AI and design.
Iām a Postdoctoral Researcher at the Human-Computer Interaction Institute at Carnegie Mellon University. Our research team, including Dr. Aniket Kittur and Dr. Nik Martelaro, is working on AI-based computational tools to support designers, engineers, and R&D scientists in product innovation.
We are particularly focused on enhancing creative processes by systematically exposing professionals to cross-domain inspirationsāmoving beyond serendipity to structured discovery.
For example, the Odon device, a groundbreaking solution for assisting difficult childbirths, was inspired by a parlor trick for removing a cork from a wine bottle. Similarly, NASA has drawn inspiration from origami for foldable solar panels, and the Wright brothers modeled aircraft control on bicycles. We want to scale this process and help innovators uncover transformative ideas more systematically.
Some of our past research includes tools like BioSpark, which enables analogical transfer from nature using AI, and Inkspire, which supports AI-assisted sketching for designers.
Weāre looking for designers, engineers, and product innovators to collaborate with us! As an advisory board member, you will:
If you are interested, drop me a message, and Iāll share more details on how to get involved. Also, if you know others who might be interested, feel free to spread the word!
Looking forward to connecting and hearing your thoughts!
Thanks,
Vikram
Typing to your CAD program seems better than hunting for seldom-used tools in drop-down menus.
Going by the example in that video, Iām not convinced.
That seemed to take 5 times longer to do.
hide / show / select / move
Having every element and layer naturally named helps immensely when tailoring for Large Language Model demos.
I compared Vizcom, Gemini and Adobe Firefly for making renders and iterating on concepts.
Firefly was garbage in my tests. Maybe there is some key to getting it to work, but I couldnāt get anything to work beyond the photoshop background erase and stuff like that.
Gemini is actually OK for renders, but it hallucinates features a bit. It loves to add USB-C to every product you ask it to render. Also, the text based interface makes it clumsy. For iterating, it was garbage. It just had no idea how to make another version of whatever I was trying to design.
Vizcom works really well for renders. I have to say, I think itās a Keyshot killer. I tried pasting a screenshot of Solidworks into and making a render and it was spot on after 1-3 interations. I also tried very crude sketches and it generally understands what my intention is. It also doesnāt add useless stuff like Gemini. The visual interface and ability to make multiple renders at the same time makes it really quick and easy to use. However, the iteration is pretty basic. I tried and tried and maybe I get one OK concept for every 10 variants I tried. It might be easier if I was doing color studies, but I didnāt try that, I was looking at forms and details.
