Listening to @goodfellow_ian talk about how far GANs have become. Things are moving so so fast. #ICLR2019
-
Show this thread
-
Cool horse-to-zebra conversion that also transforms the background. (It’s ML so it doesn’t known that it’s the horse we’re interested in, not the field—that’s often how some of what we call bias creeps in, too).
#ICLR2019pic.twitter.com/FZFplNObQK
2 replies 9 retweets 29 likesShow this thread -
I am so here for this topic, “Fast, Verifiable, and Private Execution of Neural Networks in Trusted Hardware.” Not that terrible benchmarks either. (Right now, works for inference, not training).
#ICLR2019pic.twitter.com/wzy6F8JyuX
3 replies 3 retweets 26 likesShow this thread -
So humans have a shape bias while CNNs have a texture bias when recognizing images! Fascinating.
#ICLR2019pic.twitter.com/zHMKC4wOIF
2 replies 5 retweets 26 likesShow this thread -
-
Replying to @Pinboard
I don't have the proper slide. Lemme find the paper.
1 reply 0 retweets 0 likes -
Here you go. https://arxiv.org/pdf/1811.12231.pdf … Fascinating, even when we think we get what ML does (identifying cats!), it's looking at something different. With visual stuff, we can also eyeball so nbd. But what about for hiring? Credit? You should tell all this to a Senate Committee.
1 reply 0 retweets 2 likes -
Also okay haha (kinda underslept and something I didn't know > joke I don't get).
1 reply 0 retweets 2 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.