Slashdot Mirror
Fabricating Nature and a Physical Turing Test
Nwe submitter arrow3D writes "A new startup in Norway is focused on design and fabrication at the level and quality of nature. Using pure mathematical volumes, rather than surfaces or voxels, they are developing a new generation of 3D modelling tools specifically aimed at high resolution 3D printing, to 'support the future of design and manufacturing.' Their software was recently used to create the multi-material Minotaur Helmet by Neri Oxman from MIT, as featured in Wired UK last month. An interesting thought (as recently illustrated in Dilbert) is the idea of a Physical Turing Test for synthetic objects and that both Turing Tests may require each other — i.e. only by designing and building at the resolution of nature can we achieve the intelligence of natural objects. Their software platform is still very much under development but they've started trying to
'save the world from polygons' with a KickStarter campaign that's live now."
The Slashdot Test: Any submission that includes references to Kickstarter and 3D printing is always posted to the front page.
AntiFA: An abbreviation for Anti First Amendment.
http://xkcd.com/505/
What about just using NURBS and procedural surface displacement as is common in the film industry..?
Nothing is enough for whom enough is too little - Confucius
CG artists and designers know very well the limitations and tediousness of modeling with polygons. Mesh models tend to have all kinds of problems such as cracks, holes and self-intersections. This is due to a disconnect between the real world being represented and the modeling software's attempts to represent real, volumetric, complex and “messy” objects by only surfaces.
The attack on polygons is rather unwarranted. True, surfaces are only able to visually represent an actual solid object, but then again for most visual media that's all you need them to do. Ever been on a movie set? The walls are thin wood supported by flimsy frames. Floors are painted on. Props and set pieces are often foam. Materials are cheap, lightweight, and easy to handle. There's no way any of that would work for an actual building, but again, it doesn't need to. It just needs to look like it could work.
Printing real world objects will need to account for much more than simply surfaces, much as a real structure requires more design and construction than a movie set. Developing procedurally generated materials and processes is an important step in making that happen. This goal of this project is to do just that.
In short: It's new media. New media requires new ways of working.
Shameless plug for my photos on Flickr
"Fabrication at the quality of nature" - A lot of marketing hype right there.
That aside, if you are not modeling something, and we are modeling still (it lives inside a computer dangit!), you have two choices:
1) parametric model (finite dimensional)
2) non-parametric model (infinite dimensional)
Infinite dimensional of course in the sense of as big of a sample sizes you use.
Their blog post disses polygon (a parametric model), but I bet that their product still inherently uses it. It also disses voxels,a volumetric sample point - i.e.non-parametric model. An interesting side point: voxels are not cuboids just like how pixels are not squares (http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.79.9093).
I think their product is essentially a mesh fixing/cleaning product and can save a lot of people a lot of headache (if it works). But that article is disingenuous and just a pile of hype that smells a lot like the "Unlimited Detail" farce a year or 2 back (http://www.eurogamer.net/articles/digitalfoundry-vs-unlimited-detail).
Another ad for a Kickstarter campaign. Yawn.
There many good "organic" modelers. Autodesk Mudbox is widely used by pros. Curved surface volumetric modellers go back a long way. I used one of the very first back in the 1980s, one based on deformable superellipsoids and running on a Symbolics LISP machine.
As for the "physical Turing test", if your demo reel doesn't show that you can pass that, it won't get you in the door at Pixar.
Ok, I didn't read the article, but that's certainly what it sounds like. POV-Ray is the ultimate nerd-raytracer: Programming language and retracer at the same time.
You know, 'cause they retain the simplicity of manipulation like a mesh, but the smoothness of a NURBS. Oh, and free source code from Pixar ;-)
I suppose the challenge in either method is the "blending" they're doing.
Eleventh Post!!
And the currently available fabricators build stuff in slices.
Is anyone else reminded of the Minbari when they look at that helmet? Would that make it a warrior cast?
When our name is on the back of your car, we're behind you all the way!
Constructive Solid Geometry
It's definitely marketing hype, but that's the kind of thing marketing is about.
There's kind of two processes being described in the product. Mesh repair is working with polygons and adding/modifying the polygons to produce a better mesh. This is intended to make better meshes for the second part, which is mesh mixing. Mesh mixing is functional, so can be sampled at any resolution.
Disclaimer: I work for Uformia, the company running the Kickstarter.
Seriously, how is this different than Constructive Solid Geometry? http://en.wikipedia.org/wiki/Constructive_solid_geometry
See AutoCAD, SolidWorks, MasterCAM, etc.
The analogy to the Turing Test doesn't make any sense.
The Turing Test was proposed as a way to tell if a human-made thing is intelligent, based on an inability to distinguish them from non-human-made things that are assumed to be intelligent, after you conceal all the factors that allow you to tell if the subjects were or weren't human-made.
The author is proposing the Turing Test is a way to tell if a human-made thing is human-made, based on an inability to distinguish them from non-human-made things that are assumed to be non-human-made, after you conceal all the factors that allow you to tell if the subjects were or weren't human-made.
You're trying to control for the same thing you're testing for.
Using pure mathematical volumes, rather than surfaces or voxels, they are developing a new generation of 3D modelling tools specifically aimed at high resolution 3D printing
So, identical to existing CAD packages such as Solidworlds, Pro/Engineer, Catia, FreeCAD, OpenSCAD etc? Yeah, that's a totally new generation right there, nobody has marketed products for solid modelling physical objects with the intention of producing them with some sort of additive/subtractive machining process before, no siree.
Hey! This is a profoundly awesome event. Every 3D printing company should be looking at this. Volumetric modeling of mixed materials is a dream come true for me. At last, no more leaky solid models dependent on mathematically unstable polygonal shells typically so full of holes as to not be usable directly by machines, limited in resolution with enormous file sizes for complex models.
Uformia wants to save the world from polygons, which reminds me about Euclideon (Unlimited Detail, Geoverse). There is another common point. By looking at the wiki page of "Digital materialization (DM)", we see that among the attributes of a DM system is "infinite - ability to operate at any scale and define infinite detail".
All this makes me wonder if there is a deeper relation, namely if an "unlimited detail" algoritm may use 3D data compressed (by an algorithm akin to fractal image compression) in a function representation (FREP) style, which already gives "3D who's in front information", then use a sorting algorithm exploiting the tree of the FREP to associate to each pixel from the screen a visible "3D atom". Explained in more detail here: Digital materialization, Euclideon and fractal image compression.