May you live in interesting times

The Oracle at Delphi

Joseph Schumpeter is credited with the formulation of the concept of creative destruction in a capatilist society, when "innovative entry by entrepenuers is the disruptive force that sustains economic growth, even as it destroys the value of established companies and laborers that previously enjoyed some degree of monopoly power derived from previous technological, organizational, regulatory, and economic paradigms." In a previous post I had noted that I would probably never again buy a Windows computer, but didn't look into the reasons which are making that increasingly likely. I'd like to rectify that by looking at some of the disruptive influences which are at work in the computer industry today.

It seems to me that the two largest factors at work are the explosion of smart phones and tablets as end-user computing devices, and the likely upcoming end to the period of Moore's Law.

The IBM mainframe is 50 years old this year. The mainframe, immortalized in movies from the period as being in the big room with the white coated computer operators, the blinking lights, and the spinning tape drives, was a huge, expensive, power hungry machine which centralized the power of computing in the IT department of large corporations. In the 1980's, innovation in the microcomputer space, culminating in the introduction of the IBM PC, facilitated the transfer of computing power to departments, and then individuals, first at work, then increasingly at home. These machines have continued to increase in capability and decrease in price, finally appearing to plateau out in the form of laptops with a minimum price point in the $250 range for new Windows-capable devices.

We are now seeing the next stage in this democratization of computing power. Smart phones and tablets, which a few years ago were expensive toys, have also benefited from increases in the capability of microelectronics to the point at which 7" tablets can now be had for $31.50. Admittedly these aren't very capable tablets compared to what else is available, but they are utterly amazing when compared to the landscape 10 years ago. This has the potential of putting what is really a rather powerful computer in the pocket of nearly every person on earth within the foreseeable future, with huge consequences in terms of availability of knowledge and access to education. This trend will only continue with the current emphasis on wearable technology, as initially seen in today's hot e-toy categories: Google Glass and smart watches.

This trend has been driven by three factors:

• use of inexpensive (relatively) low-power processors based on ARM technology rather than on the Intel x86 ecosystem
• rise of free open source software largely based on Linux and exemplified by Google's freely distributable Android OS, which powers around 80% of all portable devices
• the incredible spread of the internet and the phenomenon of social networking

These have both made possible and created a demand for a legion of always-on, easily portable and inexpensive personal devices. People, once exposed to this sort of power, simply will not continue to be satisfied with computers which are controlled by others, or even personally owned ones which they have to be at home, sitting up at a desk, to use. While there will always be some applications which favor workstations, such as photo or video editing, or heavy duty wordprocessing where a keyboard is a great advantage, almost all other usage can already be handled perfectly capably by hand-held phones or tablets.

The other major factor at work is the likely end to the march of progress driven by Moore's Law. IBM engineer Gordon Moore originally stated this "law" shortly after the introduction of the S/360 series mainframes. The law says, approximately, that our capability to put transistors on a silicon chip will double about every 12-18 months. This means that there should be steady and dramatic progress in the capability of the parts used to make computers. The law has been reflected in practice, so that, again approximately, the best possible computers are now hundreds of millions of times more powerful than what was available in 1962.

It is likely that this has led to a certain laziness on the part of software developers. After all, if next years computer will be twice as fast as this years, why put much effort into trying to make a program run faster on this years computer? Instead, we can put the same effort into adding extra features and attempt to use that to differentiate ourselves from our competitors. This viewpoint is not graven in stone. The Raspberry Pi, a $25/35 general purpose computer which was envisioned and originally created to get cheap computing power into schools, has been an incredible success with over 2.5 million units sold. This is at least partly because the Pi, which is based on a cpu chip that most techies think is horribly out of date, is managed by a non-profit institution. The Raspberry Pi foundation, in an effort to minimize cost as a primary goal, has focused on tuning and tweaking the software available for the Pi to make it as efficient as possible instead of just turning out a new model every year. Since developers and hobbyists are comfortable that the basic hardware will be around for more than a few months, they are much more willing to invest time and effort into optimizing their software. Concurrently, the low cost and huge user base are enticing many new people to experiment with the system, leading to a virtuous circle in which each new advance encourages more innovation and development.

If, as many believe, physical limitations mean that Moore's law cannot continue beyond one or two more generations, then it will become increasingly widespread to both innovate as a way of finding better solutions, and to write good code to make the most of the hardware we have. Both of these should tilt computer solutions toward using hardware longer and getting more value out of it, with the corollary benefit of reducing waste.

We are at or near an inflection point where we have ridden the old computing paradigm about as far as it will go, and need to embark on a new model. Exactly what that will look like 10 years from now is far from clear. It is a rule of paradigm changes that it is impossible to predict what a new paradigm will look like until after it's in place. But however this develops, it is pretty well guaranteed to be an exciting time.


It doesn't give me much help in trying to figure out what to buy next, tho.