News for hackers. Stuff that matters.
Re:Redis new data structure: the HyperLogLog by malanj on Tuesday April 01, 2014 @05:47AM (#7507102)
The idea of inexact computing seems to be gaining a lot of traction over the past few years.

The HyperLogLog data structure reminded me of the research at Rice University on inexact processors: http://www.ece.rice.edu/~al4/lingamneni-cf12.pdf

There's a lot to gain by being less precise!

Re:Redis new data structure: the HyperLogLog by sitkack on Tuesday April 01, 2014 @04:06PM (#7511285)
The future of computation is probabilistically inexact numbers over an interval.

http://en.wikipedia.org/wiki/Interval_arithmetic

We currently have a huge over tolerancing problem. We don't need exact math most of the time.

http://web.media.mit.edu/~bates/Summary_files/BatesTalk.pdf

Re:Redis new data structure: the HyperLogLog by malanj on Wednesday April 02, 2014 @01:51AM (#7513808)
Hypothesis: inexact computations will be required to make strong AI work in the medium term.

You can't simulate a brain accurately (and you probably don't need to). Neurons aren't very accurate themselves, and it does the job fine.

Re:Redis new data structure: the HyperLogLog by sitkack on Wednesday April 02, 2014 @04:57PM (#7518948)
Top of the line supercomputers consume ~10 megawatts of power, a human brain uses 20. Different problems yes, but the magnitude is staggering.
Re:Redis new data structure: the HyperLogLog by MasterScrat on Tuesday April 01, 2014 @06:10PM (#7512093)
See also: BlinkDB, a database system providing "Queries with Bounded Errors and Bounded Response Times on Very Large Data"

http://blinkdb.org/

Re:Redis new data structure: the HyperLogLog by stefantalpalaru on Tuesday April 01, 2014 @08:05AM (#7507438)
We're used to inexact computing from floating point math.
Re:Redis new data structure: the HyperLogLog by platz on Tuesday April 01, 2014 @11:47AM (#7509142)
well, inexact vs probabilistic do have some differences.