• The New Scientist (featured on cover). 2011. Darwin’s robots: A holistic, evolutionary approach means that robots could learn to design themselves.
• MSNBC.com. 2011. Intelligent design: Users power evolution in 3-D Web printing.
• Slashdot. 2011. Crowdsourcing speeds evolution of 3D printable objects.
• MIT Technology Review. 2011. 3-D design simplified: a new website could accelerate the adoption of 3-D printing.
• Communications of the ACM. 2011. ‘EndlessForms’ uses the Web to breed 3D printable objects.
• KurzweilAI.net. 2011. Now anyone can design and evolve 3-D printable objects interactively.
• ScienceDaily.com. 2011. No technical know-how needed: Endless Forms Web site helps users ‘breed’ 3-D printable objects.
• Hacker News (front page). 2011. Breed 3D printable objects, no technical know-how needed.
• Cornell Chronicle. 2011. 'Endless Forms' uses the Web to breed 3-D printable objects

On EndlessForms.com objects are evolved in the same way that plants and animals are bred. You pick the ones you like and they become the parents of the next generation of objects. As in biological evolution, the offspring look similar, but not identical, to their parents, allowing you to explore different designs. Under the hood, there is an evolutionary process in which the genomes of parents are mutated to produce new offspring objects. Each object is grown from its genome similar to how a single fertilized egg grows into a jaguar, hawk, or human. This grounding in developmental biology enables the evolution of complex, natural-looking forms. Once designed, the objects can be 3D printed (e.g. in silver, steel, or bronze). The practical purpose of the site is to allow people to create unique physical objects and see the power of evolution in action. The scientific purpose of the site is to explore what complex morphologies can be created with a computational implementation of developmental biology, which is a step toward creating robots whose bodies and brains rival the complexity, intelligence, and agility of natural organisms.

Clune J, Lipson H (2011) Evolving three-dimensional objects with a generative encoding inspired by developmental biology. Proceedings of the European Conference on Artificial Life. Pages 141-148. (pdf)

1. Evolutionary Algorithms

   This technology (a.k.a. genetic algorithms or evolutionary computation) simulates natural selection using a 'survival of the fittest' rule. The difference is that, instead of plants and animals competing, different versions of software are battling for their place in the next generation. A Darwinian process is set up so that the better programs will have lots of copies (versions similar or identical) in the next generation and less desirable software is eliminated. What makes some programs 'better' than others is determined by the person setting up the experiment (e.g., the ability to get out of mazes, drive a car without crashing, control a legged robot, etc). Over time the software gets better and better since mutations (random changes in the programs) and 'sex' (combining a portion of the code of one program with a portion of another) will occasionally produce a program that is a slight improvement over its parents. This slightly better software will thrive for a while until it too is replaced by the next slightly better software. Given enough generations, these small changes can add up to produce jaguars, whales, Olympians and poets.

   Natural selection, plus a lot of time, produced all of the "endless forms most beautiful and most wonderful" on this planet, to quote Darwin. In a computer world, because generations can happen in microseconds, we don't need millions of Earth years to pass before interesting things begin to happen. Evolutionary computation has frequently come up with better designs than any human engineer and has even received patents.




2. Generative Encodings

   Under the hood each organism is grown with concepts inspired by developmental biology: in effect, each object is grown in a manner similar to the way that a single fertilized cell grows into a jaguar, hawk, or human. Because we base the growth process on biological development, the resulting shapes are complex and have nice properties such as symmetry, repeated modules (e.g. segments), and nuanced variation. For more, check out EndlessForms.com/about_the_technology.




3. Crowdsourcing

   EndlessForms.com harnesses the creativity of the crowd by allowing anyone online to evolve objects and publish those they find interesting. Others can then further evolve, share, and rate anything that is published, allowing a collaborative exploration of the space of possible designs.




4. 3D Printing

   Any object on the site can be 3D printed with the click of a button and sent to the user. Designs can be printed in silver, steel, gold, bronze, ceramics, plastics, sandstone, and many other materials.




5. WebGL

   EndlessForms.com would not have been technologically possible as recently as a year ago because Internet browsers could not display three-dimensional shapes well. WebGL takes advantage of the graphics card in computers to quickly display complex shapes. Other recent innovations make 3D rendering possible without WebGL, but they are computationally slower and do not look as nice (compare EndlessForms.com in Firefox/Chrome vs. Internet Explorer/Safari).

While 3D printing technology is racing ahead, most people still do not know how to design their own 3D objects. EndlessForms.com enables them to design interesting objects without any technical knowledge, allowing people to create the objects in their life, such as jewelry, doorknobs, candlesticks, and sculptures. In the modern world, due to mass manufacturing, everything we buy is a clone of something that thousands of others have. Even if I pick what I think is an original, creative t-shirt and order it from an obscure internet website, I frequently encounter people all over the country wearing the same shirt! Humans love to create, but we don't have time to create and manufacture our own things. EndlessForms.com allows people to express their individuality. It's what companies like Nike that allow people to customize products dream of: instead of users choosing a few options to "customize" their shoes, here users have the freedom to design whatever they want, and it will be different from anything that's ever been created before.

Another application is to spark the creativity of designers. This paper describes how you can load a goal object and the system will try to evolve that object. Figure 8 shows that if you provide a seed concept (e.g. Figure 8a), the system automatically generates interesting variants on that concept (Figure 8b-i). A designer can thus load in this year's car or sunglasses model and see variations on that concept that may spark creative ideas for next year's design. The technology behind EndlessForms can thus serve as a creativity engine for designers, offering them ideas they may never have thought of.

To explore extent to which our abstract model of biological development captures the power of nature to make complex objects (e.g. animal morphologies).

An overarching goal of evolutionary computation is to create processes that will evolve robots as intelligent and agile as natural organisms. Right now both the bodies and behavior of robots are inflexible. Natural organisms, on the other hand, are smart, learn quickly, and can fly, swim, and run through a variety of complex environments. To synthetically evolve such robots, we need to understand how biology produces them. A key piece to that puzzle is how to create computer programs (genomes) that can grow into smart, agile robots. A major breakthrough happened recently when it was discovered that a certain mathematical abstraction of how organisms develop (called Compositional Pattern Producing Networks, or CPPNs) can capture the power of biological development to produce complex bodies and brains. Jeff Clune and colleagues recently published a paper showing that CPPNs can produce digital brains that share properties with natural brains, such as complex, regular patterns in neural wiring (see figures 15, 16, and 17 in this paper. Figure 16 specifically compares the new method (HyperNEAT, based on CPPNs) to the way things were done before (called FT-NEAT in the paper). This research demonstrated that evolving artificial intelligence with concepts from developmental biology results in much more natural looking digital brains. It further showed that if such brains are evolved to control running robots, the robots run faster and have more coordinated, graceful, and robust behaviors than when evolving without biological development. You can compare videos of robots evolved with biological development (HyperNEAT/CPPNs) to those without (FT-NEAT). Now that we know that CPPNs capture some of nature's ability to grow complex brains, the next step toward agile robotics was to see if the same abstraction of development (CPPNs) could also evolve complex bodies. The first phase of that plan was to explore what sort of three-dimensional shapes can be made with CPPNs, which is the ongoing experiment at EndlessForms.com. The second phase is to start to allow CPPNs to create similar shapes, but with muscles, bone, and brains, so shapes like those on EndlessForms can begin to move around in simulated (and eventually real) environments. The short answer to the question is thus that evolving complex shapes is a step on the path to synthetically evolving robots that rival natural animals in grace and intelligence.

Only 39% of Americans believe in evolution, in part because it is hard to see evolution happen before our eyes since the timescales are too slow. One reason we built EndlessForms.com is to allow people to see the power of evolution in action. Every shape on the website started from a randomly generated genome and was produced by random mutations plus selection, just as in nature. A difference, of course, is that humans are doing the selecting, much as in artificial selection (i.e. breeding plants and animals). We allow users to see the ancestral lineage of each shape, which helps teach the important evolutionary concept that complex things can arise from a series of small changes. Many teachers have told us they will use EndlessForms.com in their classrooms to provide an intuitive visual demonstration of evolution. In the first two months of the site being live we've had over 5,000 visitors from 70+ countries and 49 US states, which is a good sign that people are interested and having fun while being able to interact with evolution and see it occur in real time.

All of the following team members are in the Creative Machines Lab at Cornell University:

• Jeff Clune, Postdoctoral Fellow
• Jason Yosinski, Ph.D. student
• Eugene Doan, undergraduate student
• Hod Lipson, Associate Professor

The name is based on the last sentence of Darwin's On the Origin of Species: "... from so simple a beginning endless forms most beautiful and most wonderful have been, and are being, evolved."

On the Objet Connex 500 printer in our lab it takes about 20 minutes to print a single 5cm-tall object in plastic.

Jeff Clune. jeffclune@cornell.edu. +1.517.214.1060 (mobile)

Funding for this research was provided, in part, by the National Science Foundation, specifically an NSF Postdoctoral Research Fellowship in Biology to Jeff Clune (DBI-1003220), NSF CDI Grant ECCS 0941561, and NSF Creative-IT Grant IIS 0757478.