Friday, April 4, 2014

Parallel handicap

Recent research from Microsoft “Research” and “MIT” apparently demonstrates more efficient ways to utilize multi-core computers through parallel processing. As the world turns over craving for more processing power, demanded by “data scientists” and “rules machines,” it is time that the “computer scientists” found ways to produce machines that crunch more efficiently.

They don’t. Media labs and AI labs aside, on both coasts and the Melon in between, there has been more hype than what could be doled out by a used car saleswoman. Personal computers arrived 35 years ago – a life time in the historical timescale of humans – and it has not changed the world. Engineers have never been creative and hyped engineers are worse as they portray incremental failures as advancement and self-driving cars as a game changer. What they don’t seem to understand and likely never will, is that even if cars drive themselves and drones fly at will overhead, they do not make much of a difference to the 7 billion humans across the world.

Massively parallel processed machines – envisioned by those with creative minds decades ago, held hostage by the constraints of “chip design,” may be unwinding. It is certainly not comforting to imagine that the mediocrity, more interested in publishing papers than inventing, will never make it happen.