i forgot how depressing it was to read ML papers like it takes 16 phds and 6 pages of math & experimentation and hundreds of hours of compute to come out with "significant improvement over state of the art" and turns out it's like 1.2% on one benchmark
-
Show this thread
meanwhile in systems it's like two guys take a fat bong rip and come up with some shit like "bro what if we did queues.....but backwards" and boom suddenly everything runs 5x faster, ofc nothing is replicable and all the code is internal only but at least it's fun to read
1:24 PM - 14 Feb 2019
0 replies
0 retweets
22 likes
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.