Biology is the new software

This article was taken from the October 2013 issue of Wired magazine. Be the first to read Wired's articles in print before they're posted online, and get your hands on loads of additional content by subscribing online.

Pointing to a Nike Flyknit Racer -- the lime-green trainers, worn by many athletes during London 2012, that use a single knitted layer of fabric wrapped around the foot -- he notes: "Making one of something has the same cost per object as making 100,000. We no longer need to mass-produce shoes."

Complexity may be free, but it's still complex and, for now, Nike still mass-produces its shoes. You, the average punter, cannot yet scan your foot using your iPad and Autodesk's 123D Catch software, send the file to Nike, and get a pair of trainers that mould perfectly to the hereditary contours of your feet.

But walking through the gallery with Mathews, CTO and vice-president of the reality capture group at Autodesk, one senses that it's abundantly clear that this future is just, per William Gibson, awaiting its even distribution. "If complexity is free," Mathews asks, "where does complexity come from? That's what design is all about." In other words, if everything that can be rendered in bits can begenerated, a world awash in "physibles" -- as these voxels (the volumetric pixels that make up the smallest visible elements in a 3D design) of material are known -- what is the role of design, the process of deciding how those bits should be arranged? 

"Rip, mod, fab" is how Mathews and other Autodesk executives see the world: capture reality, model it in three-dimensional software space, and move it to an output to the real world. Today, anything from concrete to titanium to an individual's replacement teeth can be 3D printed. At the high end, printers can work in multiple materials, and Mathews envisions -- at the voxel level -- the emergence of "meta" materials. "You could make materials by printing each little dot in a different gradient, to make a material that isn't like anything else you could injection-mould or machine." The gallery at Autodesk is a kind of Wunderkammer, a cabinet of curiosities for the age of additive manufacturing, a perfect vantage point from which to see how algorithmic design is changing the world around us.   

Where does complexity come from? There's a display showing examples of "virtual cinematography", including reality-capture technologies for film production that are so precise they render not only an actor's skin, but its subsurface, so as to properly understand the scattering of light as it falls across pores. Simulation engines work on such improbable hypotheses, as in the case of the movie Cloudy With a Chance of Meatballs: "If you're going to have raining hamburgers, the question is, what is the coefficient of bounce on a raining bun?" Mathews asks. Nearby, there's a model for the soon-to-be-world's-largest building, Shanghai Tower, which twists and tapers (no two floors are alike) to shed wind. Computational modelling also helped it shed 25 percent of its steel, compared to an average skyscraper. Actors, buildings, shoes: the world, in the eyes of Autodesk -- whose design software stretches across more than 150 distinct fields -- is simply a "reality-capture space", and it doesn't particularly matter if the "output" ends up on the screen or on the street.  

"Any idea what this is?" Mathews asks, pointing to an object on a table -- it could be anything from a biomorphic leg splint to some prototype Imperial Cruiser from the Skywalker Ranch. It turns out to be a design for a train station. The "input" was the constraints imposed by the design competition -- the various spatial and programmatic requirements. Then the design was optimised, taking into account various factors, such as maximising strength and minimising energy usage. Instead of running simulations on the final products -- "that's when it's costly to make changes," Mathews explains -- simulations run in real time with the design. "Let's have the beta be virtual," Mathews says, eyes glancing across the street to a neighbouring tower. "With every building that you see, that is the beta." What emerges, via the finite element analysis and other tools of Autodesk's Maya animation software, and endless iteration, is the thing itself.

In the early days, Mathews says, the software of Autodesk and its ilk were simply automating the traditional workflow of design. The "aided" in computer-aided design was, he says, more akin to getting a playback of what the designer had drafted. That has all changed. "The computer itself," Mathews says, "is now doing design based on a set of criteria you're trying to optimise for." 

The results, in the gallery, are plain to see. But now Autodesk is trying to bring "rip, mod, fab" to a more far-reaching, potentially transformative field: the world we can't see. And rather than using design that merely mimics the evolutionary principles and forms of nature -- that genetic algorithm that models bone growth to make a lighter, stronger wing -- the company is thinking about how nature itself is designed.

Running a simulation of itself every so often seems appropriate for a company whose essential business is simulations. As Jeff Kowalski, Autodesk's SVP and Chief Technology Officer of the emerging-business group, recalls while sitting in a conference room, about seven years ago the company embarked on a semi-regular strategy exercise. "We create this simulation," he says, "where people get to play with the knobs and dials of the company." Teams compete against teams, all in the search for, as he puts it, "provocative new businesses". Think of it as SimCorporation.

Kowalski brought up nanotechnology. "We're involved in the mesotechnology of everything on the planet," he says. "How is it that we're not involved in nanotechnology?" What, he wanted to know, were the "mental models" of nano-scale design? What were the tools? And how did it compare to Autodesk's traditional business?  "I had this inkling that, at this scale, things are fundamentally different," Kowalski says. He began to explore the idea within the company. "The manufacturing group said, 'Of course we can do nano.'" He pauses and raises an eyebrow. "In the dialogue box you can type in ten to the minus nine." The numbers work, he was told. "But at that scale, gravity is no longer a major force. Static attraction is a bigger deal. And I have yet to find a nano table-saw or nano mill or nano drill-press."

And so the "bio-nano group", as it's called, was born at Autodesk. The company gave the job to Carlos Olguin, a user-interface specialist with the company who, as it happened, had returned to university to study systems biology. Today, the department has dozens of employees -- a curious mix of interface engineers, biologists and user-interface designers. The group is growing and is on the verge of taking new offices in San Francisco. At the heart of its effort is a new cloud-based computing platform, called "Project Cyborg", which encompasses a range of matter-progamming tools from tissue engineering to molecular self-assembly to an algorithm, called "Biome View", that can model the movement of microbial life within architecture.

The mission, says Olguin, a former electrical engineer from Mexico City who is as soft-spoken as Kowalski is intense, is simple: "Can we look at life as a design space?" That notion, clearly, has already arrived: whether in the 15,000-strong Registry of Standard Biological Parts (aka BioBricks -- standardised DNA sequences ready to "plug and play" into living cells); Oxford University's recent creation via customised 3D printing of tissue-like substances with no actual genomic material; the "biological transistors" recently reported by Stanford University that can function like genetic versions of logic gates transmitting Boolean responses; or open-source programs such as Genome Compiler that, through a simple and intuitive GUI, break complex genomic sequences into easy-to-manipulate icons. The promise -- however distant it might sound -- of the catch-all construct of synthetic biology is not only that we can build life, but we can use life to build things. As New Scientist put it, "the next industrial revolution could be biological".

Not having much of a plan, Olguin started approaching others in the field. One of his first calls was to Harvard's Wyss Institute, where a team led by George Church is dabbling in a pursuit, created a few years previously by Caltech technician Paul Rothemund, dubbed "DNA origami".

The premise, Kowalski says, is that "DNA as a molecule has really interesting structural possibilities that are very predictable." There's the famous double-helix shape, but that's just the tip of the subatomic iceberg. He jumps to a whiteboard and begins drawing a comb-like shape, a staple strand of DNA, and then another. Each branch of the comb is labelled with a letter -- the chemical base. These can eventually be joined to a scaffold strand. The base-pair rules of attraction between the strands, it turns out, are easy to anticipate. For example, "A" will always bind to "G". "If you do this enough times, on a long enough scaffold, you can make interesting loops that fold one line into any 3D shape," he says.

The problem, however, is that it might take tens of thousands of iterations, on millions of strands, to achieve the desired shape. And the tools the researchers originally used? "They had a yellow legal pad, and were using these numbers and coloured pencils to show where the folding might occur," Kowalski says. A Wyss Institute researcher named Shawn Douglas then created an open-source program, cadnano to help design DNA origami structures. Where previously it had taken Douglas and a researcher over a year to make half a dozen shapes, with the new software they had 35 designs in a few months. For Kowalski, a light bulb went off. Where Douglas was in terms of his software was where Autodesk had been in its early days, essentially an automated approximation of the desktop workflow. Autodesk flew Douglas to its research group in Toronto to work on integrating Maya, its animation software, into cadnano. As Olguin describes it, Douglas had been using Maya to visualise their design. "Once they started," he said, "they realised, 'I can actually use it to do design.'"

Douglas demonstrates a real-time simulation of DNA pairs "hybridising" with each other. He compares it to a "crochet mesh", the DNA strands interweaving in three-dimensional space, the trick being the planning out "where you want the DNA to be routed in space". The animation is reminiscent of the Nike exhibit at the Autodesk gallery, with the multiple strands of material -- without any structural support -- woven together along a 3D algorithmic plot. The comparison is not so fanciful, for what Douglas and his colleagues are designing are, well, things. Take, for example, a recent creation he and his colleagues are working on -- a "logic-gated nanorobot for targeted transport of molecular payloads". The DNA robot is roughly 40 x 30 microns -- an infinitesimal blip of the 50,000-or-so nanometre width of human hair. It could, theoretically, be delivered into the body to an area affected by cancerous cells, target the precise cells in question by "reading" markers, then, its clamshell shape opening, deliver a cancer-fighting cellular weapon. The previous state of the art in 3D molecular self-assembly was in box-like containers. The Wyss project added the concept of time, interaction and programmed instructions.

Douglas says he was initially hesitant to work on the project. "I thought our building capability wasn't there yet," he says. "If you look at what a virus does in order to recognise a specific cell type in order to get inside that cell and reprogram it, that seems much more sophisticated than just making a shape."

He compares it to trying to break into a heavily guarded building. You need more than a box. Talking to an immunologist colleague, he decided it wasn't necessary to break into the cell: "We only need to get to the surface and communicate with it that way." The question was how to get to that surface in the first place, and then how to target the correct cells. The hinged clamshell shape was arrived at, he says, because creating a fully closed box shape would involve taking a hit in yield every time it was opened. "If it only opens once, and it's folded in the closed state, it's basically already armed," Douglas explains. All this was worked through in modelling software. "The angle of how these things work is critically important," Kowalski says. "It's like a lock and key. This enzyme has to fit on this cell the right way in order to bias the cell." And -- as with the Shanghai Tower designed on Autodesk software -- form, critically, follows function. "At the biological scale," says Douglas, "you can't separate shape from function. Shape defines function at the atomic scale. Why does one protein do something and why does another do something else? We have muscle proteins that drive our muscles, proteins in our eyes that detect light. It's all amino acids, just different sequences of them." Adds Olguin: "If we can design at this level, then we're not just building artificial things outside of biology. Now we're really modifying biology to be able to introduce new functions into it."

But how do you construct the clamshell and output it into tangible reality? Apparently, it's easy. "You take that letter sequence [the set of instructions for how the DNA strands will fold and self-assemble]," Olguin explains, "and with your credit card -- and there are companies that do this today -- you order the sequence. In the mail arrives a vial full of millions and millions of copies of that sequence, which, when you put it in warm water, will self-assemble into that clam shell." As computing power increases and the costs of DNA base-pair sequencing becomes cheaper than a phone call, it may be the future of medicine. "It astonishes people when we say there are already people creating new forms of life using this methodology," Kowalski says. "It astonishes them even more when they learn that they are in high school." He points to a "bug" created at the International Genetically Engineered Machines competition that can detect aqueous arsenic and has already been deployed near industrial plants in Bangladesh. "It will glow green when it detects arsenic," he says.

What makes this possible, argues Kowalski, is not merely advances in sequencing and computing power, but having "mental models that support how to work with this stuff". Rip, mod, fab. People talk of DNA sequences being "addressable", and human tissue is being engineered using techniques drawn from the "fusion deposition modelling" of additive manufacturing. As Keith Murphy, CEO of Organovo, a company that creates functional tissue using 3D bioprinting, explains it, the company's printer deposits layer upon layer of "bio-ink" -- thousands of cells on top of each other. Unlike previous tissue engineering, which relied on polymer scaffolds on which to position cells, Organovo uses a gel material that, Murphy says, "holds cells in place just long enough for them to keep from falling apart from gravity, and to fuse." 

At the moment, Murphy says, the source code for creating these structures is simple, command-line stuff -- "move printer tip 1 to space X, deposit gel". But through a recently announced partnership with Autodesk, the company wants to bring the sophistication and power of 3D-modelling software to bear on new ventures, such as its creation of human liver cells.

"Now we're building tools," he says.

"People can work with tools, learn from tools, understand how cells are acting in 3D -- and then you can start to build predictive, 'in silico' models." What Autodesk hopes to do with Project Cyborg is to
enable biologists and chemists -- people who may not be well-versed in the complexities of 3D modeling -- to tap into the power of algorithmic design.

Because the tools and principles of CAD are being brought into nature, what might nature bring into the world of designing human objects?

Oguin picks up a flask that has been sitting on the table. At the bottom are what look like colourful little plastic rocks. He begins methodically to shake the flask. After six or seven seconds, the "rocks", which each contain a magnet, have assembled into a ball. With energy, time and random movement -- a turbulent process of attraction and repulsion -- something new has come into being. Shake it harder and it falls apart again.

Within that simple flask are two powerful principles: 4D manufacturing and self-assembly -- the latter of which, Olguin argues, is bottom-up design. "It's very different from how we learn to design -- cut things, splice them." Rather, he says, "your goal as a designer is to set the right constraints for the individual parts from which this emergent behaviour will then form a whole." And the extra dimension? It's time.

The beaker demonstration was created by Skylar Tibbits, an architect at MIT who heads the Self-Assembly Lab, working with Arthur Olson, a molecular biologist at the Scripps Research Institute. Tibbits describes it as a "tangible educational model to describe molecular self-assembly", in this case for a protein, but suggests the principle could be scaled up for potential manufacturing: in February 2013 he produced an exhibit at TED in Los Angeles in which component pieces, spun together in a giant hopper, formed stools and other furniture objects. What was most striking, Tibbits says, is the design connections people started to make. "People would come up who had no biological or design background and, once they understood that this furniture piece was based on the polio virus, as it started to assemble they very quickly turned the corner and said: isn't there something I could design to throw in the chamber so that the polio virus wouldn't assemble itself?"

Creating 4D objects, like his "self-folding strand", which looks a like a series of sausage links that assemble into a predetermined shape when it encounters water, requires not only new materials -- in this case a synthetic polymer from 3D-printing company Stratasys that expands 150 per cent in water. It also needs new software and, perhaps most importantly, new mental models, as Autodesk might say. "This is a new design paradigm," Tibbits says. "It's just not designing and making things. It's designing things that change over time, and so how we incorporate that programmability and changeability into design tools is a really big question." Project Cyborg, he says, can be used to optimise this process -- which joints fold when. But perhaps it all can come full circle. "Maybe these macro-scale things, the models that Arthur Olson and I are working on, can lend insights into how proteins fold or how DNA works."

The potential of 4D means design, the iterative process of which begins in the software model, is computationally nurtured in real time, and does not end when the object enters the real world -- it adapts to local conditions, it grows, it assumes new forms. This might be a new piping system that Tibbits is working on, in which the capacity of pipes could adjust the volume of water, in some peristaltic fashion; it could be Organovo's gel scaffold, withering away; it could be Douglas's DNA robot, melding into the bloodstream after having completed its task. "More and more design is turning away from what you've already designed up here," Kowalski says, gesturing to his head, "and changing into a conversation with the computer. Nobody drafts any more -- everyone creates a higher-level model that expresses design intent. All that stuff we thought of as blueprints are inconsequential outputs of the overall process."

Tom Vanderbilt wrote, in wired 01.13, about Google's plans to create a search engine that understands how we think.

Update: Wired.co.uk incorrectly credited the folding straw and beaker images to Autodesk and referred to Jeff Kowalski as Chief Strategy Officer instead of SVP and CTO. These have now been corrected. 10/10/2013