Full Version: Teaching LLMs how to solid model

Show messages: All  1  2-9

Great find ! Very many thanks Metin !

Best regards
. Ken

---

You're welcome! I'm curious how this will evolve.

---

One of the links describes what the letters mean: Large Language Models (LLMs)

---

The funny thing is : it's more speed to "talk" (write) for make something or it's more speed to make something as usual? ;)

---

I found the article and experimental process of evaluation interesting, thanks for sharing Metin. I was already somewhat familiar with OpenSCAD so the mechanics of using an LLM to drive it made some sense, but does seem at odds with the human joy of the creative process. Regardless of whether one likes it or not though, the revolutions of AI are lapping at many foundations, so it seems worthwhile to stay informed at the least. Patrick's thought about timing was a bit startling, but from what I've been watching, somewhat plausible.

"So when does text-to-CAD become a commonly used tool for mechanical engineers? With start-ups actively building products and the rapid improvement of frontier models, my guess would be something like 6-24 months."

Coincidentally I just finished reading Henry Kissinger's final book "Genesis: Artificial Intelligence, Hope, and the Human Spirit". I'd not read any of his work and would not have thought that a statesman in his 90s would be authoritative in this evolving field. Wrong I was, with co-authors Eric Schmidt and Craig Mundie, this is a very good read, not just about AI but about how some are looking at deploying it in high level contexts like governance, law, and economics. Kissinger's lifetime of statesmanship provides good grounding against what seems like highfalutin fantasy, and yet he is optimistic about its application to our world.

--Larry

---

Hi Larry,

Yes, the rapid rise of "AI" is something else. Right now LLMs are mainly still engines that string together statistical sequences, sometimes referred to as stochastic parroting.

I'm keeping track of the developments, but as you mentioned, the joy of creating is not the same when using AI. For example, I've tried a few of the image-to-3D LLMs, and I'm sometimes amazed by how correct the technology interprets depth in complex images, with an arm in front of another arm, for example. But in the end, it's the creative process that gives you satisfaction… Conquering initial hurdles, reaching a point where you realize it's going to work out (or not), surprising yourself and others, and the satisfaction of receiving praise for what you've made. All that is not applicable when you make use of AI.

I hope we'll reach a point where LLMs are helpful tools to aid creation, but do not take over the entire creative process.

---

Hi Metin,

Very well put! I think the appeal of the "shiny new thing" and perhaps "fast & cheap" will draw many toward this technology, but at a cost to our human spirits.

Okay, now we're sounding like a couple of sage old men proudly surveying their buggy whip mfg facility while Henry Ford is building a new plant nearby...

--Larry

---

Hahaha, yeah, I try not to judge too much. I did a lot of skeptical complaining in the initial phase of the AI hype, but I guess it's here to stay, so we might as well try to carefully embrace it, looking for ways to use it to our advantage.

---

Show messages: All  1  2-9