https://www.reddit.com/r/MachineLearning/comments/hhonq4/d_yann_lecuns_last_substantial_post_on_twitter/fwdwpo7/
A comment from the reddit thread -
A paper demonstrating depixelation of faces was released, and once people figured out that it would (among many other mistakes) often depixelate pictures of black faces into ones with white features. ML twitter immediately starts talking about how "problematic" and unethical this research is, etc.
In the middle of this confused uproar, LeCun tweets that the model's biased behavior should be expected as it was trained on a majority-white dataset. (he later makes it clear he was speaking of that paper specifically)
He got some pushback immediately following that, but the focal point was Timnit Gebru's quote-tweeting him and saying she was sick and tired of his framing. She then makes it clear that if he doesn't reply he is ignoring her because of her race and gender. Despite her having made no claim of substance thus far, LeCun responds with a huge thread calmly explaining his reasoning.
But that was not good enough for her - her only response to his thread was that engaging with him is a waste of her time and that it's not her job to educate him.
LeCun himself has been clear in his desire to come to an understanding, and that he has no fundamental disagreement with her. Gebru has uttered nothing but abuse and accusations.
Following his thread others have joined the chorus, and he has been accused of gaslighting, mansplaining, tone policing, racism, sexism, and more besides. Throughout the week, well-known figures have been queueing up to tell everyone just how despicable his behaviour is, and how only people like Timnit Gebru can show us the way forward.
What happened? A bias researcher misunderstood one of LeCun's tweets and made it her mission to punish him. The entire community fell for it.
and from ycombinator -
| |||
| |||
|
and -
No comments:
Post a Comment