One of the great insights I gained at the Supernova Conference I attended today was this: "Do not underestimate the significance of Exponential Curves". The Exponential Curve in question here was presented by Dr Greg Papadopoulos, CTO of Sun Microsystems.
This chart wonderfully demonstrates that the cost of the same unit of computing (as measured in MIPS) has fallen about 1 million times from 1980 to 2005. This means that in just a few years, feats of computing previously unimaginable will be possible that might lead to a future breakthrough with a potential impact greater than the sum of previous innovations. Personally, this implies that seemingly amazing outcomes like near perfect machine translation or major improvements in Natual Language Processing (NLP) are more likely to be achieved in the near future (sooner than many of us expect). Combine that with similarly huge improvements in bandwidth (again using one of the charts from Papadopoulos' presentation)
today's applications of video will seem somewhat primitive compared to potential future applications like countless high definition, on demand or live video channels available on a multiplicity of consumer devices. In fact, this potential is illustrated by the recent stats showing that YouTube is responsible for 10% of all internet traffic and 20% of all HTTP traffic.
This plays well into the hands of current start-ups like Powerset that can benefit tremendously from exponential increases of computing power at the same cost level for the next few years. It will be equally interesting to observe if Amazon's daring initiative with its EC2 and S3 Services can level the playing field for emerging start-ups given the tremendous computing infrastructure lead built by web giants like Google.
