Researchers at Cornell University have launched EndlessForms, a website that lets users create sculptures virtually and render them in physical form. The site demonstrates a technology that designers could use to create new products and accelerate the broader adoption of 3-D printing.
People can use EndlessForms without any prior 3-D design experience. The user begins by choosing an object from a randomly generated gallery. The site creates a new gallery of variants of the chosen object, and the user selects one of the variants. The process repeats, gradually refining the design into the shape the user desires. Users can share this shape with other users and, if they wish, send the object to a 3-D printing service to render it in a variety of materials, including plastic, silver, and gold-plated steel. A five-to-seven-centimeter plastic model typically costs less than $10.
Jeff Clune, a postdoctoral fellow at Cornell and the project lead, believes this approach is a critical advance over earlier attempts to produce objects and shapes through digital mutation and selection. Those attempts produced things that “just don’t look that natural,” says Clune. “So we went and stole the secrets that biological evolution took millions of years to discover.” According to Clune, generating objects in this way automatically gives them useful properties such as symmetry, and when these objects are printed in 3-D, they usually turn out to be structurally sound.
The rules EndlessForms uses to generate objects and their variants resemble those of developmental biology—the study of how DNA instructions unfold to create an entire living organism. “Embryos create patterns in the form of chemical gradients,” says Clune. Chemical gradients—changes in the concentrations of particular molecules—control which parts of an embryonic organism’s genome are expressed. “There might be a simple linear gradient, like head to tail, or there might be might be a repeating gradient, like that which governs the creation of segments in a caterpillar,” says Clune. Combining a head-to-tail, front-to-back, and repeating gradient, for example, creates the basic body plan of vertebrates.
Rather than simulating chemical gradients, EndlessForms stores a “genome” for each object that describes a collection of simple mathematical functions analogous to these gradients, so that, for example, a sine function takes the place of a repeating chemical gradient. The website produces variations on objects by mutating this mathematical DNA. By summing the contributions from all the functions described by the genome, and checking whether the result is above a threshold value, EndlessForms determines whether blocks of 3-D space called voxels are either filled or left empty in an object.
Currently, EndlessForms can generate only a limited range of novelty shapes. To contain computational costs, the system uses a relatively small number of voxels for each object. And because each object is initially randomly generated, a user with a specific shape in mind may need to do quite a bit of selecting and evolving to produce something close to that shape. However, the Cornell group and collaborators plan to develop the system to enable a user to input a 3-D scan of an existing object and then evolve variants from that. This would allow, for example, a designer to scan in a pair of sunglasses and use the system to evolve new styles.
The technology is “very impressive,” says Neri Oxman, director of the MIT Media Lab’s Mediated Matter research group. She believes the user-friendliness of the evolutionary approach could help drive the broader adoption of 3-D printing technologies, similar to how easy-to-use image editors fueled the growth of digital photography and graphic manipulation. “People could scan their own toothbrushes or other objects and evolve various designs of such items for members of their family,” says Oxman, noting that this could ultimately have an impact on design similar to the impact that blogs and social media have had on journalism, opening the field to the general public.
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
The Biggest Questions: What is death?
New neuroscience is challenging our understanding of the dying process—bringing opportunities for the living.
Driving companywide efficiencies with AI
Advanced AI and ML capabilities revolutionize how administrative and operations tasks are done.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.