Help me understand this reaction @baykenney @dribnet. Can't it be *both* that ImageNet is a great contribution and is flawed? @trevorpaglen
@katecrawford @drfeifeihttps://twitter.com/baykenney/status/1107811767021903872 …
It is not about Imagenet, but about Internet itself. It is fundamentally flawed in how there are no checks and balances against hateful content and misogyny/racism It is allowed to thrive and multiply. Any scraping off from internet will have same issues
-
-
I don't really understand your point. This is clearly errors of human mechanical turk workers picking wrong synset definitions, what does it have to do with the internet?
-
This Tweet is unavailable.
- Show replies
New conversation -
-
-
Of course this is true, but the issue pointed out here seems more specific to ImageNet's labeling of collected images.
-
This is shallow art that only demonstrates cherry picking. It can equally label your favorite basketballers as surgeons & spacewalkers (accurate for their style of play!) Agreed it shows bias that exists in any internet data, including 10-year old Stanford student projectspic.twitter.com/xYCr7RIzUP
End of conversation
New conversation -
-
-
This was my result. It picked up a random person in the background and caption was really weirdpic.twitter.com/per2L9BLpX
-
That is really weird!
End of conversation
New conversation -
-
-
internet is fine. it’s the damn humans that are the problem.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
yes - and an irony here is that this very dataset of images loosely correlated with hateful content is exactly the resource you might want if you were building training set for automating detection of this hate speech online.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
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.