Why would other employees be knowledgeable in the employment relationship between someone and a company? Unless the employees are in the direct management chain, or HR or something, aren't they just parroting what Timnit is claiming?
Full text:
Just wanted to say that I was stunned by what just happened to Timnit Gebru, who was in my team until a few days ago (I'm sure you've all read what happened already). I have always been and will remain a strong supporter of her scientific work to make sure that AI becomes a positive force for the minoritized, as well as her generosity and tireless actions to lift the voices of the silenced ones. She taught me a lot and still is. I stand by you, Timnit. I also stand by the rest of my team who, like me, was stunned and is trying to make sense of all this. In particular the Ethical AI team, but also the broader Brain Research team.
Also I read the highlights of the paper linked here and it seems fine and reasonable. These are reasonable questions to ask. Sure we can mitigate the cost of training these models with green energy or finding new optimization./compression techniques etc... but that doesn't mean that with ever evolving data volumes we will ever fully mitigate that. Bitcoin for all its merit consumes a lot of energy. Is that more than USD or something else?
Another thing to consider is that AI right now has a rich get richer winner take all dynamic. We have titans battling it out for NLP supremacy. Deep mind just revolutionized protein folding which has direct implications into biomedical research and technology. Small academic labs are at an extreme disadvantage unless they collaborate with big tech. Interestingly physics research evolved in a similar way with large particle colliders and telescopes started up. So you can make a direct analogy to that, that the basic research can only be done with million dollar investments. That's an ethical question. GPT-3 trained on misinformation would make pizza gate look tame in comparison. So there are huge social implications.
Even real datasets collected from society and social data will have implicit and explicit biases in them. There are huge ethical concerns with data, data collection and with the interpretations AI models will make.
Your credit score, your social credit score, a company that refuses to hire you because a chat bot has labeled you subversive.
The TSA/DHS blanket banned people from flying and gave no feedback or transparency into why. It took members of Congress to get some people redressed.
Google needs a massive clear-out. People aren't doing their jobs anymore there, they've become so obsessed with fictional narratives about non-existent racism and sexism.
My point wasn't even about the truthiness of the claims of either side, but rather that if party A claims X, and then parties B and C just repeat the claims without being able to add anything - that's not news. That's just a person having allies who believe their claims.
If Microsoft chimed in and said Timnit actually resigned, no one would write an article on that (besides for how bizarre of a statement that would be by another company), because no one expects Microsoft to be apprised of what happens between Google and its employees.
I bet Jeffrey Dahmer didn’t have a history of complaining about cannibalism either, but that doesn’t necessarily make him the good guy.
1. Did in fact go through and pass the internal approval process that he implied it didn't.
2. Was then subjected to a second, intransparent "review" that is not customary practice at Google.
3. The results of that "review" were presented to the author as a decree (which also is not customary practice).
4. And when the author subsequently protested that she would not work under such rules (as well she should) that was taken as a pretext to let her go — without even involving her direct manager.
Under those circumstances, Dean's response is misleading and evasive, a practice often characterized as "gaslighting" or, in the term that the grandparent comment ridiculed, a "micro-aggression".
It's not about whether you're a "white dude", it's what rules you set for "white dudes" vs. "non-white non-dudes".
Did Jeff Dean think Timnit was a white dude and so let her publish, but when he found out she was a brown woman, he said he cannot stand for this?
These 4 points were all taken from the blog post I referenced, which I found well documented and thus credible:
https://googlewalkout.medium.com/setting-the-record-straight...
> Why would they publish multiple of her papers on basically the same exact subject, but then decide they need to stop her at all costs for this paper?
You tell me. If they were truly on "basically the same exact subject", there can't really be a scientific explanation for some to be publishable and others not, can there?
Or, your theories are flawed in some manner and don't actually capture the truth. I'm asking you how previous behavior would fit into the theories that you yourself believe.