If you ever learn about optimization, the second technique they teach you after “hill climbing” is something called “simulated annealing“. Forget trying to work through Wikipedia’s definition of it, here’s how I mentally visualize it:
Imagine you your job was to find the highest point on a random Frank Gehry designed building, say the Hotel Marques de Riscal. The problem is, you’re blind so you can only determine the height of one point at a time.
One way to do it would be to take the entire outside shell of the structure and turn it upside down and then drop a ball bearing in it. When the ball bearing stops moving, you declare that point to be probably the lowest point.
In the field of optimization, this is called hill climbing and it sure is fast but the problem with it is obvious: the ball bearing will likely drop into some rut on the side of the building before reaching the bottom (aka: a local minima).
Simulated annealing is like adding a giant paint shaker and imagining the ball bearing getting heavier over time. At first, the ball bearing is light as a feather and every random vibration is going to make it bounce around, sometimes to a lower point but also sometimes to a higher point. Over time, as the ball bearing gets heavier, it becomes progressively harder to jostle out of it’s rut but when it does move, it moves to a lower rut some distance away. Finally, the ball bearing becomes so heavy it’s virtually impossible to budge, at which point, you declare that position to be probably the lowest point. Simulated annealing can still get trapped in local minima but it usually does a hell of a lot better than hill climbing.
Now, imagine that the ball bearing is society and it’s searching for the solution to a problem like, say, attaching sheets of paper together. If you’re living in any time since the 1930′s, it’s overwhelmingly likely that you’ll be reaching for a “Gem” style paperclip.
What many people don’t realize is that the Gem paperclip wasn’t the only type of paperclip invented. In fact, the 1909 Websters Dictionary entry on paperclips referred to the “Konaclip” version which was considered the prototypical paperclip of the time.
From 1864 when the first paperclip patent was granted, till 1930 when the gem became dominant, there were literally dozens of designs for paperclips (which, thankfully, has been documented on the Paperclip section of the Early Office Museum). But since the 1930′s, technological progress in the paperclip arena has ground to a halt with the entirety of society standardizing on the Gem and every other style dying in obscurity (For a much more detailed look into the paperclip’s fascinating history, I refer you to the book The Evolution of Useful Things).
To explain this history, I refer back to the simulated annealing analogy. In the early days of paperclipdom, society wasn’t particularly attached to any one design so there was very little momentum in the system. Even a tiny vibration was enough to make another design viable. However, as time went on, certain styles of paperclips started to have legacy effects and it became a race between a few, select alternatives, most notably the Konaclip & the Gem. Finally, society had settled so firmly upon the Gem that it it would take the most extraordinary effort to have any other design supplant it which is why it remains the paperclip of choice to this day. In Simulated Annealing terms, the entropy was now so low that it became practically impossible to escape from a local minima.
Indeed, one of the characteristics of simulated annealing is that it goes through 3 distinct phases. A period of fluid, diverse shifting, a long period of stagnation punctuated by occasional radical shifts and then finally stability from which only incremental improvement is possible.
Looking at the pattern of historical developments of other technologies, it’s possible to spot these same shifts in between phases.
Before 1973, there was a diverse ecosystem of physical interaction paradigms for desktop computing. The most notable examples from that era were Ivar Sutherland’s Sketchpad using light pen interaction and Alan Kay’s tablet style Dynabook.
Then, in 1973, the Xerox Alto was released and virtually halted the progress of computer interfaces from that point forward. Nearly every desktop computer today is a recognizable descendant of the Xerox Alto.
The list of genuine innovations in physical UI that have been adopted since the Alto are as laughably few in number as they are trivial in scope. In rough order of importance: Speakers, microphones, webcams, the mouse wheel, wireless networking, higher resolution screens, a numpad, flatter screens & the Windows/Command button (this is not a sampling, it’s the complete list).
The list of potential innovations are as staggering as they are futile: Pen computing, Tabletop computing, Tablet computing, Augmented & Virtual Reality, Touch based interaction, Multitouch, MultiMouse, Bimanual Interaction, Single Display Groupware, 3D displays and the list goes on and on. In the last 37 years, billions of dollars have been poured into these alternative technologies in a vain attempt to supplant the Xerox Alto as the dominant paradigm with pretty much nothing to show from all that work except a bunch of pretty pilot projects.
And with each passing year, it’s become more and more difficult to foster a viable alternative paradigm because more and more gets invested into keeping Alto style computing firmly entrenched. Software is designed for a keyboard+mouse+screen, people have invested time learning the intricacies of a WIMP OS and thousands of companies have a vested interest in keeping the system entrenched.
The same pattern can be found in keyboard layouts with QWERTY, programming languages with C, web standards with HTML+CSS+JS, Office Productivity with Microsoft Office, Email with Eudora and so forth. In every instance, a period of rapid innovation was brought abruptly to a close with a dominant technology and, from that point forward, genuine changes in the status quo happen at best every decade and only after extraordinary effort *.
Almost every discussion of a new innovation focuses on the details of the innovation almost exclusively without considering the broader social context. Such discussion manages to miss the vital point that adoption of an innovation depends only a tiny bit on the actual innovation and almost completely on the current progress of society. Once a technology has reached a certain maturity point, whatever local minima that society has currently reached will establish itself be the dominant technology until the end of time regardless of what technological progress has occurred or has yet to occur. Like a game of musical chairs, it’s all about when the music stops. If you’ve managed to grab a seat at the table, then you’ll stay there forever, if you missed your spot, then it’s going to take extraordinary effort to regain it, no matter how much of an improvement you are over the status quo.
* This same model also explains why European & Japanese mobile & broadband technology is so far in advance of the US even thought the US were the pioneers of both technologies. Because the US started early, it’s industry matured around an earlier technological & social standard which caused it to fall behind the less mature, more technologically limber peers.