Thursday, June 4, 2015

Light chip

Recent research from the University of Utah seems to get closer to silicon photonics and faster light based processors, many orders better than conventional products. Splitting light has always been easy but Utah engineers have accomplished it with beam splitters of a mere 2.4 microns of size. This may propel us out of the ongoing rat race of packing silicon ever closer on conventional chips for less interesting performance improvements.

Moore’s law, held sacred by technologists and used by exponential curve plotting, singularity seeking intellectuals, for stupid predictions, may have done significant damage to the psyche of innovation in electronics. Humans, grand optimizers of their limited life horizons, always fall into the trap of incrementally improving what is available. They appear to be satisfied with metrics that double over long horizons – such as years and this is in stark contrast to their keen awareness of limited time. With less than a thousand months of life span, a metric that doubles every 18 months appears so much less interesting than one that explodes by a few orders of magnitude in the same horizon. More importantly, doubling computing power in 18 months with no perceptible impact on existing applications is a waste of time.

It is time to throw out processes that grow within the same orders of magnitude every year – they will certainly employ people, but they will not make any difference to humanity.