Jeff was generous enough to answer a few questions and help us understand where it all began.
Duann: Tell us a little about your background, your PhD, your current post doctoral research and how it applies to Endless Forms.
Jeff: My background is in evolving artificial intelligence. The process of evolution created all the amazing creatures in nature, so we harness that process by creating evolution in computers to evolve smarter robots. People have been evolving AI for a long time, but one thing that prevented the technique from working well is that it was not paired with another natural marvel: the process of development, wherein software (DNA) in a single cell can self-assemble into a jaguar, hawk, or human. My colleagues and I have found that combining evolution and development really increases the power of what evolution can produce. My dissertation—done at Michigan State University—shows that adding development to evolution produces digital brains with intricate, geometrically patterned neural connections that are much more intelligent than brains evolved with evolution alone. A next step was to investigate whether this developmental phase would also enable the evolution of complex 3D shapes, similar to how nature produces animals and plants. To do that we built EndlessForms.com.
Duann: What is the goal for EndlessForms.com, is it the software itself, is EndlessForms.com simply a way of proving the algorithm that happens to have an interesting visual outcome, or is it a way to give access to tools of creation for 3D printing or none of the above?
Jeff: There are three major goals for EndlessForms.com. The first is that it enables scientists to better understand how biological development works by seeing what sorts of objects are created with the particular abstraction of development we use. That furthers developmental biology and also the pursuit of evolved, artificially intelligent robot brains and bodies. The second is that EndlessForms.com allows non-technical users to create 3D objects with no knowledge of 3D modeling or CAD software. A final goal is education: only 39% of Americans believe in evolution. EndlessForms.com allows people to see the process of evolution create complex forms before their very eyes, providing a hands-on way to learn about evolution, developmental biology, and artificial intelligence.
Duann: Is the human/crowdsourced interaction actually important to the genetic equations underlying the software, the forms that the software creates or your research?
Jeff: The crowdsourcing component is critical. Evolution works best by exploring many things and keeping what works. Any individual may get frustrated on the site if nothing impressive evolves, but the community as a whole finds incredible objects over and over again. Importantly, other users can then further evolve these ideas, so each new discovery serves as a stepping stone to other amazing innovations. I have repeatedly seen on the site that a certain type of object is published that did not exist before, such as a bottle, mushroom, or a lamp. Soon thereafter, there is an explosion of different variations on that design that all descend from it. Such initial discoveries might be rare, or take a lot of work, but once one member of the community makes a discovery the rest of the community can build on it. In that way, the crowdsourcing on this site is similar to communities of scientists and engineers building upon each others ideas. Interestingly, the same thing occurs in nature: once certain innovations appear they can cause an explosion of new species that all use that innovation, which is called an "adaptive radiation." All of these ideas were realized by Ken Stanley (who also invented the developmental encoding we use), Jimmy Secretan, and the rest of the team that built Picbreeder.org, a site that uses the same technology to crowdsource the evolution of two-dimensional images.
Duann: How complex is the genome at this point, how many generations of evolution and how many objects have been published so far? There is currently 560 on Shapeways..
Jeff: Just today we surpassed having 1.5 million objects evaluated by humans, and we've only been live a few months! We've had over 100,000 generations occur and just under 1,000 objects have been published. People seem excited about the technology: we've had more than 25,000 visitors from 130 countries and all 50 US states.
Duann: You are in the same research labs at Cornell with other projects around 3D printing, FAB at Home, self assembling materials and more.. How does your work cross pollinate with other researchers, advisers, students and who has been involved with the project?
The Cornell Creative Machines Lab is a great incubator of ideas from many different disciplines. It is fantastic to be surrounded by so many nice and talented people. EndlessForms.com would not have come about without the people in the lab. Hod Lipson, the head of the lab, realized the great need for 3D design automation tools, and really encouraged this whole line of research. Once I had written the software to evolve 3D objects on my computer, the challenge was to create a website to allow the entire world to experiment with it. Fortunately, just at that time a brilliant scientist with web development experience named Jason Yosinski joined our lab at Cornell. We then recruited undergraduate Eugene Doan and master's student Sijie Liu, and the team of us rapidly built the site. Throughout the process Hod has supported the project and enabled us to properly resource it. Jon Hiller also provided a great foundation of code that we built upon. More generally, interacting with people from different disciplines and passions has been very fruitful. For example, we go to Maker Faire and I show how to automatically create 3D objects at one of the table and at the other end the Fab@Home team shows people how to 3D print them. It's a great mix of talented people with different backgrounds.
Duann: Many people see the need to use 3D CAD software as the barrier to entry to 3D printing, how do you see EndlessForms.com breaking down that barrier.
Jeff: We liken the 3D printing world to one where everyone has an iPod, but there are only a few songs to play. We have great 3D printing technology like that at Shapeways, but most people do not know how to make 3D objects to print. EndlessForms.com allows people with zero technical skills to create objects they find interesting and print them with the click of a button. There could not be a gentler introduction to the world of design and 3D printing.
Duann: The current iteration of EndlessForms.com is interesting as an academic project, inspiring as an artistic experiment and has the potential to be used to grow a product design, is there any plans to 'evolve' the project to include the option to influence the genome further by choosing which elements to grow or suppress, to constrain elements as 'hard points' while making others variable or to upload a 'seed' stl file which could then be used as a starting point?
The answers to all of these questions is yes! The thing most people want, and which we are working on, is to be able to upload a 3D design and further evolve it. In that way, EndlessForms can serve as a creativity-engine, suggesting design variants to you that you may never have thought of. You can get a sense of this now by evolving a common object, such as a bottle or a lamp: you suddenly find yourself looking at interesting versions of these designs that you've never seen on a store shelf, but which should be made because they are beautiful. We would also like to be able to add a physics simulator, so you can test whether an object is structurally sound or will stand up. Finally, we would like to add the ability to specify constraints and to evolve only subsections of a design. We've had architects make these requests because they want to evolve everything from interior designs to entire skyscrapers.
Duann: Where do you see the project heading in the long run, what are the possible applications you are most excited about?
Jeff: What truly excites me is to get these forms moving around as artificially intelligent robots. I see butterflies and bipeds published on the current site, and I want them to fly away or walk out of a 3D printer. The first step will be to allow evolution to deposit different materials that will function like muscle and bone in simulation and evolve creatures that can swim, run, and fly. The second step is to be able to print such 3D robots. Jeffrey Lipton, the Fab@Home project leader, is inventing ways to 3D print robots that will walk out of the printer. Once we have that capability, EndlessForms.com would serve as the perfect technology to design such robots.
Returning to static object design, EndlessForms demonstrates that there is a huge need to make 3D object creation easier. People want to be able to make 3D shapes, whether for the virtual or physical world. We need better tools to make it easy to go from idea to object. EndlessForms.com is one way to lower the barriers that stand in the way, but there are many others, and we'll be working on ways to make this process easier.
In the very near future, one big limit on the complexity of the designs at present is computational speed. As computers become faster, and as we develop the code to be more efficient, the designs will look better and better.
Duann: Is there anything else you would like to say about the project?
Jeff: The more people that use the site, the better the designs will get, so I encourage people to spread the word. Those readers that order EndlessForms pieces from Shapeways, please send pictures of them to us or post them to our Facebook page.
For more information on the site, including a video overview, pictures of objects printed via Shapeways in silver and bronze, and more detailed descriptions, see the press page.
A final thing to mention, since people want to use this technology with their existing 3D models, is that the ideas that drive EndlessForms.com can be used to automatically create variants of any pre-existing design. While that is not currently available on the site, how to do that is outlined in a scientific publication, which also describes all the details of how EndlessForms works. You can access that paper at my website jeffclune.com. The title of the paper is "Evolving three-dimensional objects with a generative encoding inspired by developmental biology."
Duann: Final question for bonus points: What 3D software (aside from Endless Forms) do you use and why?
Jeff: Ironically, I have no idea how to use 3D modeling software! I therefore speak from experience when I say that there are many people who want to play in this space, but cannot because they do not have the time to learn how to use CAD programs or Blender. It was a joy to hold the first EndlessForm object I created, because I otherwise would not have been able to make it.
Duann: Ok Jeff, Thanks again for your time.. I am really enjoying playing with the software and can't wait for more controls....