This is an awful defense. White supremacists have explicit strategies for spreading hate online based on finding specific terms. See Fragmentation of Truth https://points.datasociety.net/the-fragmentation-of-truth-3c766ebb74cf … (h/t @ShriramKMurthi)
https://twitter.com/CapHaddock23/status/1284496678901383168 …
-
-
This Tweet is unavailable.Show this thread
-
The problem is that it's *so easy* for ML devs to miss training data bias. The Tiny Images dataset was used for 14 years before we realized that it contained racist and misogynistic categories (https://arxiv.org/pdf/2006.16923.pdf …).
Show this thread -
I just... don't understand the vehement defense of GPT-3 here. Are we not all in agreement that it's bad to say those things? And subsequently we don't want to train AI to say it as well?
Show this thread
End of conversation
New conversation -
-
-
That's hardly even the only kind of bad response. So much creativity!
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.
cognitive psychology. PhD