4 ms·

Pardon my ignorance but what exactly was the issue with what LeCun said in that response? I feel like it's a pretty agreed upon paradigm that feeding an AI/ML a

Pardon my ignorance but what exactly was the issue with what LeCun said in that response? I feel like it's a pretty agreed upon paradigm that feeding an AI/ML algo with biased data for training will make the AI/ML algo biased. That seems unsurprising and like a perfectly valid statement so there must be something I'm missing and, at least from the article, I couldn't glean what the issue was since the response was just "I'm sick of this framing" without any kind of additional information. If I'm reading it correctly, the issue maybe was that he didn't go far enough and put too much emphasis on data set bias?

sillysaurusx 6y ago
I don't know; I had exactly the same questions. FWIW I work in ML, and I've said similar things as LeCun. The thing is, I would love to help (and listen) to those who have studied this extensively. But there wasn't any information to go on. The article makes it seem like they shouted him down while delivering no actionable advice.
- dkonofalski 6y ago
  Yeah... it's nearly impossible to figure out what's going on after the fact since the conversation is split amongst so many different Twitter threads. The best I can discern is that his response is dismissive of systemic issues that affect results and introduce biases but, to me, that feels like he's getting jumped on for not being general enough in his criticism.
  - Izkata 6y ago
    More specifically, IIRC the claim was that he's using the data as an excuse, instead of adjusting the data so the ML algorithm gets the desired un-biased result. My understanding was that it was mostly coming from activists that don't really understand how AI/ML works.
- read_if_gay_ 6y ago
  > The article makes it seem like they shouted him down while delivering no actionable advice. Worse. It makes it seem like they shouted him down while not even delivering any concrete criticism. If that's accurate you can't blame LeCun for leaving. And looking at your typical political "discussion" on Twitter, I wouldn't be surprised if it is.
stanfordkid 6y ago
This is the real problem with BLM -- they wanted a sound bite, he didn't give it to them so they attacked him.

sillysaurusx 6y ago

I don't know; I had exactly the same questions. FWIW I work in ML, and I've said similar things as LeCun. The thing is, I would love to help (and listen) to those who have studied this extensively. But there wasn't any information to go on. The article makes it seem like they shouted him down while delivering no actionable advice.

dkonofalski 6y ago
Yeah... it's nearly impossible to figure out what's going on after the fact since the conversation is split amongst so many different Twitter threads. The best I can discern is that his response is dismissive of systemic issues that affect results and introduce biases but, to me, that feels like he's getting jumped on for not being general enough in his criticism.
- Izkata 6y ago
  More specifically, IIRC the claim was that he's using the data as an excuse, instead of adjusting the data so the ML algorithm gets the desired un-biased result. My understanding was that it was mostly coming from activists that don't really understand how AI/ML works.
read_if_gay_ 6y ago
> The article makes it seem like they shouted him down while delivering no actionable advice. Worse. It makes it seem like they shouted him down while not even delivering any concrete criticism. If that's accurate you can't blame LeCun for leaving. And looking at your typical political "discussion" on Twitter, I wouldn't be surprised if it is.

stanfordkid 6y ago

This is the real problem with BLM -- they wanted a sound bite, he didn't give it to them so they attacked him.