Design with Nature as Role Model

This entry is adapted from a presentation I did at the University of Iceland today, hence all the decorations.

Using nature as a role model in design is one of my biggest interests. By this notion I’m talking about how we can study nature and use its solutions, designs and methods when making our own designs and technologies, a practice often referred to as biomimickry.

The subject might sound strange and distant to some of you, so let’s start with a few examples:

1. Velcro (or touch-fasteners): Velcro, these strips of fabric that stick so tightly to each other, were invented by a Swiss inventor, named George de Mestral. The story says he was walking his dog one day and when he came home noticed how cockleburs where sticking to his pants and his dog’s coat. He examined the cockleburs under a microscope and noticed the hook-like structure on the burs. This led George to the invention of the well known velcro, an everyday object that we find ourselves using many times a day.
2. Camouflage: Camouflage patterns, the essential decoration on every piece of a sport hunter’s equipment and the pattern that helps conceal armed forces in combat, is based on camouflage patterns found in nature, used for the very same purpose – hiding prey from predators and vice versa.
3. Swimwear: At the 2000 Olympics in Sydney, 28 out of 33 gold medalists wore Speedo’s Fastskin allovers. These suits improve swimmer’s speed by as much as 7.5% by reducing the drag in the water, as compared to the old-fashioned human flesh. The design is based on the features of sharks’ skin, that uses tiny V-shaped ridges (called dermal denticles) to reduce turbulence around their body and therefore the drag in the water.
4. Artificial Intelligence: Artificial Intelligence is certainly one field where we are seeking to imitate nature, in that case human intelligence, with human technology. More on that later.
5. Genetic algorithms: For certain kind of problems it has been found useful to use genetic methods to develop problems. This is not to say that you could breed Microsoft Word and Excel and expect a useful outcome. It is rather a methodology that can be used to attack certain kind of hard problems where the solution is not straight forward. Examples include control systems for computer controlled brakes, various search methods and pattern recognition. The methods are largely comparable to those of nature’s evolution. You start of with a (randomly generated) set of variables or modules that define a possible solution to the problem – call these sets genes. You can then make a new set of individuals using two methods: mutation and / or breeding. This results in a new set. This set is then put up to a test. In nature this is the natural selection part, whereas in genetic methods this is a test to see how well each individual solves the task at hand. The “fittest” individuals from this generation are then used as the initial set for another round of the same process. The process is then repeated, improving the solution with every generation and – if all goes well – resulting in a solution to the problem at hand.

A methodology on the rise

Looking to nature in this way is a methodology that is on the rise. Ever since we were young children we have been told that the strings in a spider’s web are incredibly strong. Scientists are now using its design as well as silk to create artificial materials with similar properties; the design of seashells is being used as an alternative to plastic in some cases; and the healing mechanism of a rhinoceros’ horn is a target that people are eagerly trying to mimic to make self healing materials.

In software, various methods of biomimickry are being studied. This is especially true as stated before when solving hard problems where traditional methods have not worked properly. Examples include virus protection and spam filters. Neural networks as an approach in AI are also an obvious example.

One of the more interesting accounts of biomimickry is the use of genetic methods in design, in the following case form design. The company Affinova has a design methodology they call IDEA. In this example the task at hand is to create a new bottle for drinks. Firstly, Affinova brings in human designers that come up with various innovative bottle designs. Each of these bottle designs is then broken into its basic elements, the cap, the neck, the label, the body and the base. These elements are then “bred” or mixed to make genuinely new types of bottles. At this stage the different designs are evaluated by a focus group representing the target group for the product. The designs that they like survive and are used to make the next generation of bottles, in very much the same way as for the genetic algorithms above. The results of the process are designs that can be radically different from anything that the human designers visualized in the beginning, but have been proven to be successful by a group that represents the target consumers.

I recommend Affinova’s interactive tour, for further explanations.

In a similar way, the almost 10 year old experiments of Karl Sims are really interesting. In these experiments, Sims made a simple virtual world. Basic physics, such as gravity and water pressure apply in this world, but the only objects that exist are boxes. Joints can link two boxes together and every joint can be moved within a limited range of movement. A set of linked blocks represent an individual in the world. Using once again similar genetic methods as explained above, the behavior (movement pattern of the joints) and the features (how blocks are linked) of these individuals are evolved to meet a criteria, in the simplest example the criteria is just to be able to move. Once again, amazing results – and what is more, with this simple model, locomotion similar to methods found in real life creatures, emerged. Check out the video (9Mb).

Are these the optimal solutions?

But are the solutions found in nature, necessarily the optimal solutions to the problems at hand? The answer is no. Nature’s designs are as good as they need to be, no better. All extra improvements are expensive in terms of time and often energy and there is no selection pressure to do any better than what ensures the survival of the group.

This is why we find the extremes in nature’s designs in environments like the savannas of Africa. Here we have a lot of resources, fierce competition and predators lurking in every shadow. This is where we find the fastest creatures on Earth, the best camouflages, the sharpest claws and the best defense mechanisms.

The other extreme is found in places like remote islands where there is plenty of food, little competition and nobody to eat you. Under these circumstances you find something like the extinct dodo bird, a big fat bird that used to live in Mauritius and proved to be easy prey for hungry sailors and hunters; the large lizards on the Galapagos Islands that have to stay still during the night in order to preserve body heat (they are cold blooded like other reptiles) making them easy prey, if there were any predators to eat them; and our very own Icelandic Geirfugl (Great Auk), a large flightless bird of the family Alcidae (think puffins) of which the last pair was killed in 1844 on an island south of Iceland.

That those designs are “bad” can certainly be disputed, but they obviously lack the grace, adoptability and extremes that the savanna animals display.

Even natures “good” designs are often clumsy. Stephen Jay Gould mentions the panda’s thumb as an example. Panda bears are close relatives of common bears such as the grizzly bear. The panda however has adopted a very special diet, consisting entirely of bamboo leaves. To eat the leaves it certainly helps to have a hand to hold the leaves and the bamboo tree, but bears have no thumb. Instead of evolving one of the five fingers of the bear pod to become a thumb similar to ours, the panda’s thumb is made with a mutation of one of the bones in the wrist, making it an inflexible stubby thumb, far from the more optimal design of our thumb.

Looking at nature’s solutions, one can quickly see that the designs are not optimal; it’s obviously possible to do better. Take man as an example. Why don’t we have even bigger brains? Why are we not equipped for running faster? Why don’t we have eyes in the back of the head? All of these are features that would certainly improve the design (aesthetics aside). The answer is partially that it is not needed. We are quite capable of survival in our current environment (even too so, asking the Great Auk). The other part of the answer is that nature simply hasn’t had the time to test all the possibilities. Evolution is a slow process and the possibilities that have been tested by breeding and mutations are only a tiny fraction of the possibilities.

“Good” designs that nature finds can therefore last for a long time, almost unchanged. Sharks and crocodiles are two such examples. Sharks have been more or less unchanged since the Silur period (for some 420 million years) and crocodiles since Triassic (about 220 million years ago). This means that crocodiles existed in almost the same form as today long before most of the dinosaur species even emerged.

Artificial Intelligence vs. Artificial Flight

In an article by Kenneth M. Ford and Patrick J. Haynes (in Understanding Artificial Intelligence), they interestingly compare artificial intelligence with “artificial flight”.

They point out that in early attempts at mechanical flight; the pioneers believed that imitating birds as much as possible would hold the solution. Early planes therefore often have bird like features, like a beak that we now know as an obvious fact that have nothing at all to do with flying. This sentence could be found in the magazine English Mechanic as late as 1900, only three years before the Wright brothers managed their first flight: “The true flying machine will be to all intents and purposes an artificial bird.”

It was however exactly the release of this notion that led to the solutions to mechanic flight. Sure the Wright brothers used gulls’ wings as an inspiration for their wings, but there is no wing-flapping (a common attempt in early mechanic flight), the plane is driven by a propeller and the means used for stabilizing the aircraft in flight are fundamentally different from those that birds use.

It should also be kept in mind that planes serve a very different purpose to that of birds. Planes are for carrying passengers and baggage, they don’t need to stay put in mid air, dive for prey from high altitudes or be able to land on a branch. Also, the materials available, especially to these pioneers, were inflexible materials such as wood and steel, not the flexible and durable organic materials that make bird like flight possible.

As a side note, Ford and Haynes mention the fact that some people denied to call mechanic flight “flight” well into the 20th century. Flight was what birds do, and mechanic flight was “artificial flight” at best. It is funny to compare this to artificial intelligence and the fact that AI is constantly claimed to be a failure even though it is increasingly solving tasks that would certainly have been called “intelligent” 50 or 100 years ago.


From the above, I draw the conclusion that what we should seek in nature is inspiration and exceptional methodology, rather than expecting to find the exact solutions. Nature’s solutions are not perfect, but its methodologies allow us to develop and optimize our own solutions.

Nature is good at what it needs to be good at. Intelligence makes highly efficient “operating systems” for the hardware (bodies) it controls. Nature far exceeds human technology in efficient use of energy, and in building structures such as bones, various strings and building materials e.g. of shells.

Human technology on the other hand exceeds nature e.g. in vision (microscopes and space telescopes); computational power; amount and accuracy of data storage and retrieval; speed (of vehicles vs. animals); and certain properties of chemicals.

The methodologies we can utilize are mainly two:

1. The evolutionary processes as explained in the examples of genetic algorithms, Karl Sims’ creatures and the bottle design.
2. Emergent properties of complex systems such as neural networks, fetus development, ant colonies, etc. Systems where complex and unpredictable behavior emerges from large systems of interconnected, simple, predictable modules.

Further study and experiments with these methodologies, will allow us to apply them to specific problems and design technologies that far exceed our current solutions and often nature itself. We are able to work with these methods on a totally different timescale than the one nature has to work with and using our intelligence we can guide the processes intuitively in the directions we want, whereas nature has no designer and hence no goal to work towards.

One comment

Comments are closed.