Invention is the act of making something entirely new or of discovering an entirely new way of accomplishing something, and so often this is a result of trying many different approaches. For me, when one method doesn't work or achieve the results I need, I just try something else. Yet what will make an approach different from someone else's approach is the spark of creativity. To solve the problem, try applying a technique or a principle that, at first glance, doesn't seem to apply.
When I invent things, I know I'm trying to solve a problem. I'm exhausting all of the possible ways to solve it. I'm looking for an efficient way to make use of the information or progress that has been made so far. I'm finding a better way to do it. Or a way to do it at all.
Try Something Unlikely
In ancient Egypt, blacksmiths were good at forming swords other rudimentary tools by holding a piece of iron into a fire to make it malleable and beating it with a hammer. The hammer and anvil had been used for many years, having been invented in the iron age. But sometime around 1450 BCE in ancient Egypt during the reign of Twthomosis III somebody decided that a leather bag could serve as a bellows, and that the increase of forced air would make the fire hotter. Because of this, metal became more malleable, and could even be melted.
This is a clear example of using an unlikely object in common use for something else entirely. A leather bag, used for carrying things, becomes a bellows for metallurgy. Many inventions, in fact, require this kind of discovery.
To make these kinds of discoveries, we must learn about as many things as possible, but perhaps not in depth. Absorbing a little about plenty of subjects is food for invention. It helps you make connections between things that are, for all intents and purposes, not connected in the first place.
For instance: knowing about Voronoi diagrams helped me figure out how best to render fascinating patterns like those produced by raindrops on a windshield. My blog post on where ideas come from is helpful in understanding how to exercise your brain to make such connections.
Try Try Again
But even more discoveries happen a small bit at a time. And the light bulb is the perfect example. Most people associate Thomas Edison with the discovery of the light bulb. But really, he only participated in part of the invention: the part that made it practical.
In 1800, Humphry Davy, in Britain, discovered that applying electricity to a carbon filament could make it glow, demonstrating the electric arc. Some 77 years later, American Charles Francis Bush manufactured carbon arc lamps to illuminate Cleveland, keeping the filament in a glass bottle. Two years later, Thomas Alva Edison discovered that filaments in an oxygen-free bulb would still glow. Then he tried literally thousands of materials before settling upon carbonized bamboo for the filament. The new bulb could last 1200 hours. And it had a screw-in base! But it wasn't until 1911 when modern sintered ductile tungsten filaments were invented at General Electric, that their useful lifetime was increased substantially. Then, in 1913, Irving Langmuir started using inert (electrically nonconductive) gases like argon (instead of a vacuum) inside the bulb, which increased luminosity by a factor of two and also reduced bulb blackening. Nitrogen, xenon, argon, neon, and krypton are routinely used inside bulbs today. However, when mercury vapor is used, the gas itself is the conductor, producing blue-green electric arc.
Of course, light bulbs are being reinvented every few years now. Fluorescent bulbs are used in businesses largely because they are four to six times the efficiency of incandescent bulbs. Then there were compact fluorescent light (CFL) bulbs, sharing the same efficiency advantage, but in a compact light bulb form factor. And now light-emitting diode (LED) lighting. These new bulbs save about 80-90% of the energy (over incandescent bulbs) required to illuminate us. And they last about 25 times longer than CFL bulbs.
The future is going to be just as much about conserving energy as it is about producing it.
The main use for creativity in invention is simply so you can solve the hardest problems of all. These are the problems that don't have an apparent solution.
Two supreme examples of this kind of problem are computer vision and computer cognition. Teaching a computer to understand everyday objects like faces, kinds of clothing, the make and model of a car, and even something as simple as a tree is incredibly difficult. Humans do this very well, of course, and this belies its complexity. Teaching a computer to read and understand a book is also hard beyond comprehension. Small parts of this, like optical character recognition and a small amount of natural language processing have been accomplished. But for the computer to actually understand the subject matter and discuss it, or even better to learn from it, is practically impossible. People dedicate their lives to solving this problem.
A small example of the problem of computer cognition is what I once dreamt about: subject space. I envisioned a space where all concepts are related in different ways. Each concept is a node in the graph of subject space and arcs between the nodes relate them.
Here I show is-a relations as a green arrow between two objects. So the green arrow between FLEA and BUG represents the information that a flea is a kind of bug. Similarly meat, rice, and carrot are a kind of food. This is a subset relationship. Another kind of relationship has to do with ownership or possession. A cyan arrow from one object to another means that the source object can possess the destination. A dog has legs, and so does a bug. A has relation can have other information associated with it. For instance, a dog has 4 legs and a bug has 6 or 8 legs. Any relation, which generally is where the verbs live in this space, can have additional information associated with it, in the form of an adverb. For instance the eats relation can have quickly associated with it.
Action relations concern a direct or indirect object. These are shown in indigo. Legs walk on the floor. A human buys food, and a dog eats the food. A flea lives on the dog. In this way buys, walk-on, lives-on, and eats are relations. And by definition, those relations can have a timestamp associated with them. The sequence in which actions occur affect the semantics. Sometimes in a causal way.
Very complicated relations are two way arcs, like the dog-master relationship. There are other obvious relationships, like is-an-attribute-of, where appropriate adjectives may be associated with subjects. Even idiomatic expressions get their representations here. For instance hair of the dog is slang for an alcoholic drink.
Note that a human has legs but I didn't include an arc for that relationship. This shows that subject space is not planar. In fact, it is n-dimensional.
Such a graph is useful in understanding and parsing the grammar of text or spoken language. A sentence can then be encoded into a series of factual semantic concepts. For instance, if you know the man buys food, then you will have to determine what the food consists of. Based on this graph, it could be meat, carrot, or rice, or some combination of them.
Also, the relation eats really means can eat. When parsing text, the fact that a given dog is eating or has eaten food is yet to be discovered. Once discovered, this subject space graph helps the semantic understanding system codify the actions that occur.
Sometimes the solution, however complex, can come to you in a dream. And this shows a creatively-applied technique, graph theory, and how it is applied to a nearly impossible problem, computer understanding.
Trial and Error
It is quite remarkable when a discovery gets made by accident!
Physicist Henri Becquerel was looking for X-rays from naturally-fluorescent materials in 1986. He knew that phosphorus would collect energy by being exposed to sunlight. And he had a naturally-fluorescent material: uranium. But there was one main problem: it was winter and the days were all overcast.
So the put his materials together in a drawer, including a bit of uranium and a photographic plate, and waited for a day when the sun would come out. When that day came, he removed the materials from the drawer and soon found that the photographic plates were affected by the uranium without being first exposed to sunlight.
And radioactivity was discovered.
My point is that sometimes a discovery is the result of unintended consequences. As for me, I have invented a few effects by accidentally creating a bug in a program I wrote. This is part of the pleasure of working in graphics. In fact, the cool visual effect in my Mess and Creativity post was discovered as the result of a bug in a program that computed image directions.
Trials and Tribulations
One problem, the lofting problem, was an elusive problem to me for years. I spent a lot of time constructing better and faster Gaussian Blur algorithms over the years, and even learned of a few new ones from such people as Michael Herf and Ben Weiss. But it wasn't until late 2004 that Kok Chen suggested that I apply constraints to the blur. And an iterative algorithm to solve this problem was born. This is detailed in my Hard Problems post.