←back to thread

AI engineers claim new algorithm reduces AI power consumption by 95%

(www.tomshardware.com)
352 points ferriswil | 2 comments | 19 Oct 24 18:03 UTC | HN request time: 0s | source
Show context
remexre ◴[19 Oct 24 18:48 UTC] No.41889747[source]▶
Isn't this just taking advantage of "log(x) + log(y) = log(xy)"? The IEEE754 floating-point representation stores floats as sign, mantissa, and exponent -- ignore the first two (you quantitized anyway, right?), and the exponent is just an integer storing log() of the float.
replies(2): >>41889800 #>>41890236 #
convolvatron ◴[19 Oct 24 18:55 UTC] No.41889800[source]▶
yes. and the next question is 'ok, how do we add'
replies(2): >>41889877 #>>41889991 #
1. dietr1ch ◴[19 Oct 24 19:06 UTC] No.41889877[source]▶
I guess that if the bulk of the computation goes into the multiplications, you can work in the log-space and simply sum, and when the time comes to actually do a sum on the original space you can go back and sum.
replies(1): >>41890126 #
2. a-loup-e ◴[19 Oct 24 19:35 UTC] No.41890126[source]▶
Not sure how well that would work if you're often adding bias after every layer