Speed of technological progress and social effects

iDNES.cz - Å est Å¡kůdců, kteří zaruÄ?eně zniÄ?í váš poÄ?ítaÄ?
Pokud nenastane zvrat v použitých technologiích, za 15 let budou procesory a grafické Ä?ipy vyzařovat na centimetr Ä?vereÄ?ný stejné teplo jako sluneÄ?ní povrch.

Now here's an interesting throwaway line in an article about protecting hardware: "Unless there is a change in the technologies used, in 15 years processors and graphics chips will emit as much heat per square centimeter as the surface of the sun." This refers to the importance of heat in computing which, while keeping much of the progress, in check remains almost completely obscured from public technological consciousness (how many people have heard about a water cooled computer plugged into somebody's refrigerator).

But this conceals an even broader point. The much talked about Moore's law (originally formulated for the increase in transistor density now often simply [mis]stated as predicting the doubling of computing power every 18 months) has so far withstood many predictions of doom due to supposed physical limits (the effect of 'natural-selection-like' processes might be a useful area to investigate). In effect, the computing power of many of the components found in a computer has risen astronomically. However, technological progress has left the psycho-social (for lack of a better term) aspects of computing behind. Speech recognition is nowhere near enough to some of the outlandish claims made for it in the recent past (although http://www.podscope.com can do more than I ever thought possible) and semantic text processing relies more on complex stochastic processes than real analysis. It is surprising how bad even the most moder text-to-speech generation sounds and any pruposeful text generation still remains in the 'gimmick' stage. Now, I remember my computational phonetics teacher saying more than 10 years ago that no new major advances in speech recognition would be made irrespective of the speed of computing unless the theoretical model changed. Until very recently I believed he was right but I'm beginning to wonder whether he underestimated the raw power of pure 'natural-selection' driven engineering (I owe this concept to Steve Jones in Almost like a whale). It would certainly be worth investigating what has changed and how to allow for Podscope and what the limits of it are.

But there is also an interesting point to be made from the perspective of the user. I'm always reminded of Microsoft's OLE (Object Linking and Embedding - now called something else) which claimed to revolutionize the integration of different styles of documents but which took almost 10 years of increases in computing power to approach anything near usability. Now people embed spreadsheets in documents without even noticing but it was not always so.

The whole point is the limits of technological progress as a predictor of social progress. It is certainly a facilitator of it (as the telephone, railways and other technologies for communication - which includes roads) but it may take a tremendous amount of it to induce even a modest change or just a tiny nudge to result in a whole sale transformation. (I'd better stop writing before I get into complexity theory but I will not stop thinking about this. Here's an article that is making the same assumptions about complexity in a different context.)