3 ms·
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
by malanj 13y ago
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 http://www.ece.rice.edu/~al4/lingamneni-cf12.pdf
There's a lot to gain by being less precise!
- stefantalpalaru 13y agoWe're used to inexact computing from floating point math.
- platz 13y agowell, inexact vs probabilistic do have some differences.
- sitkack 13y agoThe future of computation is probabilistically inexact numbers over an interval. http://en.wikipedia.org/wiki/Interval_arithmetic 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 http://web.media.mit.edu/~bates/Summary_files/BatesTalk.pdf
- malanj 13y agoHypothesis: 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.
- sitkack 13y agoTop of the line supercomputers consume ~10 megawatts of power, a human brain uses 20. Different problems yes, but the magnitude is staggering.
- MasterScrat 13y agoSee also: BlinkDB, a database system providing "Queries with Bounded Errors and Bounded Response Times on Very Large Data" http://blinkdb.org/ http://blinkdb.org/