What would be more useful to you - AI that generates 3D models from text, or a solid repository of Ready-to-Use spare parts?
RECODE.AM #22
Artificial intelligence has reached virtually every field of technology, including additive manufacturing. One of the areas where it is expected to have the greatest impact is the generation of 3D models based on textual descriptions, sketches, or images.
In theory, this sounds promising: the user provides a general idea of what they want to create, and the algorithm produces a ready-made model that can be immediately printed on a 3D printer.
This vision fits into a broader trend that promises automation and instant access to everything “at a click” - a 3D printer that fabricates objects we merely describe or type on a screen.
However, in practice, when we look critically at the current state of AI applications for 3D model generation, it becomes clear that this technology is still far from maturity - and quite frankly, overhyped.
At the same time, for more than a decade, there has been a well-established segment of structured, professional libraries containing ready-to-use spare parts - created, tested, and certified by engineers.
And today, it is still these well-developed repositories of 3D models that serve as truly practical tools for real work, while AI remains more of a curiosity than a functional solution.
AI in design for 3D printing
AI applications for generating 3D models typically rely on deep learning techniques and the processing of point clouds, meshes, or voxel-based spatial representations. The main issue, however, lies in the quality and consistency of the results.
Models created by AI often share the same flaws we see in AI-generated 2D images - missing details, distortions, incorrect proportions, and sometimes entirely unrealistic shapes.
While such errors might be forgivable in 2D graphics - like a hand with six fingers or a distorted face - in the world of 3D printing, every millimeter counts. Even a slight deviation in dimension, angle, or proportion can render a printed part completely useless.
Moreover, in professional contexts - especially in engineering, automotive, or medical applications - such errors can be downright dangerous.
One of the biggest problems with AI-generated 3D models is the lack of any possibility for certification or quality control.
Even if an AI-generated file looks correct on screen, it is very difficult to verify whether it meets mechanical strength, ergonomic, or safety standards.
In professional design, it’s not enough for something to merely “look right” in a visualization - what’s needed are technical specifications, material data, test results, and compliance with specific industry standards.
AI-generated models have no design history, no documentation, and no information about material properties or manufacturing constraints. Using them is a bit like downloading a random model from the internet and hoping it happens to fit - which, in the case of expensive or critical components, is not only irresponsible but downright foolish.
Spare parts database
Professional repositories of 3D models and spare parts offer an entirely different level of quality and predictability. Platforms such as TraceParts, GrabCAD, or 3D ContentCentral are good examples - they host hundreds of thousands of models provided directly by manufacturers of mechanical components.
In such cases, users can be confident that the downloaded model matches a real-world product and that its dimensions are consistent with the official technical specifications.
These repositories often also include CAD documentation, compliance certificates, and load capacity data.
It’s worth noting that many modern AI models for 3D object generation are trained precisely on data originating from these repositories. The irony is that artificial intelligence frequently relies on knowledge created by humans, attempting to reproduce or transform it without fully understanding the underlying technical context.
As a result, the generated models might be statistically correct - but not necessarily functional. Just as generative text can sound convincing while being factually inaccurate, a 3D model can “look right” yet fail to work in reality.
This, of course, does not mean that artificial intelligence has no value in 3D design. Its potential to accelerate the conceptual phase, generate inspiration, or automatically adjust shapes is enormous.
The real issue lies in how AI is often presented - as if it were already capable of replacing engineers, designers, and entire repositories of models. In truth, it remains an experimental tool that is only beginning to learn how to understand spatial reality in a practical, reliable way.
Until standards emerge that make it possible to certify AI-generated models, their use in production will remain limited to experimental or conceptual prototyping.
Therefore, if I had to choose today between AI and a well-organized repository of ready-made spare parts, the decision would be simple: technology can inspire - but only the quality of data allows us to make really useful things.
Why not have both? 😉