AI researcher Timnit Gebru resigns from Google • Hacker News Board

Re: AI researcher Timnit Gebru resigns from Google

#901

Post by thu2111 » Sat, Dec 05, 2020, 1:35 PM UTC

kthxbye123 wrote: post #896

thu2111 wrote: post #850

Earlier quoted context omitted.

Rude, maybe but not very. Dismissal, yes. The belief that AI should learn reality as it exists and not some ideologically purified version of it is a completely rational one, and dismissing the "work" of those so-called "ethicists" is no more problematic than dismissing the "work" of the Chinese censors who purify the Chinese internet. Indeed being dismissive of that work is to be praised, as it shouldn't exist at al…

Name the specific true facts about reality you think these models are capturing that are at risk of being 'purified' by AI ethicists. List them.

Well, read any AI bias paper. Facts about reality these models capture and which are considered problematic include stuff like "doctors are more often men and nurses are more often women" or "most people on the internet are white" or "most stuff on the internet is written in English". Basically the internet is mostly built by westerners and any facet of that gets classed as "bias" rather than being what it is: the truth.

Even trivial facts like in English "men chuckle, women giggle" has been used as examples of bias. That's not bias though, that's just how the English language works.

For example:

https://ai.googleblog.com/2020/04/a-scalable-approach-to-red...

Re: AI researcher Timnit Gebru resigns from Google

#902

Post by visarga » Sat, Dec 05, 2020, 4:51 PM UTC

jcheng wrote: post #777

LordFast wrote: post #736

Earlier quoted context omitted.

Me thinks it's a common misconception for strongly opinionated people to believe that companies want to hire them in order for them to "fix things and make the company a better place". In reality, what every company/team/management wants is someone who will give them some constructive criticism on small mistakes here and there that are easy to fix, but ultimately act as a supportive shield for the company against /ot…

That's pretty jaded IMHO. I would think most management wants positive change, not just to be shielded from criticism. But reasonable people can disagree about what the status quo is, how good or bad that status quo is, what counts as positive change, what timescale positive change needs to happen at, and what costs/tradeoffs they are willing to accept to accomplish positive change.

Probably Timnit and the management disagreed on "what timescale positive change needs to happen at, and what costs/tradeoffs they are willing to accept to accomplish positive change".

Re: AI researcher Timnit Gebru resigns from Google

#903

Post by visarga » Sat, Dec 05, 2020, 4:54 PM UTC

woko wrote: post #834

ethical-ai wrote: post #794

Earlier quoted context omitted.

What I really cannot understand is how someone who works on ethics can think that revealing names of anonymous reviewer is OK? Why would they need those names anyway? I'm also unable to think of any reason they might need the identities of the reviewers.

My guess would be that they took the issue as a personal offence, and they believe that someone higher in the power structure has some personal beef against them as a person, rather than against parts of their work. They want to confront that person.

Maybe they want a good old fashion public Twitter judgement for the reviewers. She has a crowd.

Re: AI researcher Timnit Gebru resigns from Google

#904

Post by visarga » Sat, Dec 05, 2020, 4:56 PM UTC

jpster wrote: post #843

sjs382 wrote: post #397

It's hard to make too much sense about what happened. Here's what I made of it, if anyone can point out any important mistakes/inaccuracies/details: 1. She and her co-authors submitted a paper for review 1 day before its (external, for publication?) due date. It's supposed to be submitted to Google 2 weeks prior, so they can review. 2. The authors submitted it externally before the review was returned. 3. Google deci…

> Timnit Gebru demanded access to all of the internal feedback, along with the author of the feedback. If Google wouldn't provide this access, she said that she intended to resign and would set a final day. Is the part that makes me cynical about Google’s motive. What sort of process calls for XFN feedback on your work but withholds that feedback and its sources from you? This is very odd.

Depends on how many followers the author has on Twitter and her history of Twitter use. After what I've seen, I think it was a good call.

Re: AI researcher Timnit Gebru resigns from Google

#905

Post by cauthon » Sat, Dec 05, 2020, 5:05 PM UTC

throwaway74453 wrote: post #900

cauthon wrote: post #899

Earlier quoted context omitted.

Having worked at three large academic institutions and been part of the submission process for about a dozen manuscripts, I have never seen or heard of a requirement for an internal administrative approval. IRB approval for study designs, yes, admin submission of grants, yes, but never an authorization that the results and manuscript could be submitted to a journal or conference

Are you serious? Every large collaboration in e.g. physics has those. You simple cannot create a splinter cell within the organization and publish whatever you want. You have to coordinate with everybody because the work is never truly yours. In case of e.g. CERN, literally thousands of people helped you in some way and they do not want you to potentially tarnish their reputation. Industry AI labs are like that as we…

It sounds like we’re confusing two things.

What you’re describing in this comment sounds like a consortium. These often an embargo on publication of independent manuscripts until the primary paper(s) have been published. However, once the data’s out, each investigator is free to publish whatever secondary analysis they like.

I understood the Google process to be an _administrative_ approval, based on the conclusions of the paper and unrelated to the quality of the science. This I have never heard of at an academic institution. The final call that a manuscript is ready for publication has always been made by the corresponding/senior authors, not admin.

Re: AI researcher Timnit Gebru resigns from Google

#906

Post by visarga » Sat, Dec 05, 2020, 5:07 PM UTC

IfOnlyYouKnew wrote: post #856

rubatuga wrote: post #318

Earlier quoted context omitted.

What Yann said about data bias is a fact. He wasn’t dismissing your concerns nor distracting from it. A lot of people tried to extrapolate what Yann meant, but from the way I see it, if more people learn about the cause of this racial bias, the more people will be inclined to use fairer and diverse datasets in their future ML research.

"It's December" is also a fact. He chose to mention a different one. Why, if not as an explanation / excuse for the model's behaviour? Every practitioner already knows that these models behave according to the data they are trained on. The general public has no use for this explanation because they aren't the one deciding on the training data. So the only practical purpose is to offer some superficial causal mechanis…

You're making a fallacy here, forcing the issue into black and white - if you mention first 'dataset bias', it must mean you 'offer some superficial causal mechanism that sounds like an explanation to that public'

Re: AI researcher Timnit Gebru resigns from Google

#907

Post by visarga » Sat, Dec 05, 2020, 5:39 PM UTC

je42 wrote: post #383

rubatuga wrote: post #318

Earlier quoted context omitted.

What Yann said about data bias is a fact. He wasn’t dismissing your concerns nor distracting from it. A lot of people tried to extrapolate what Yann meant, but from the way I see it, if more people learn about the cause of this racial bias, the more people will be inclined to use fairer and diverse datasets in their future ML research.

The point is that she has an expectation of Yann to go further than just stating a fact. I assume she thinks that Yann has enough intellectual capacity to improve situation of biased ML but doesnt execute. Further, my guess is that the reason why he doesnt execute is not transparent to her, hence frustrating.

> The point is that she has an expectation of Yann to go further than just stating a fact.

Yann was talking about a specific model, she was talking about ML & society on a grand scale. Can't we talk about specific things now? Why did he have to apologize for it?

My feeling is that Timnit is using people, she guilt trips, humiliates them online, makes a big fuss and it's all to whip up support in her crusades. She wants to win the public opinion by any means, so she needs famous opponents to pile on, such as Yann and Dean (who have universal admiration in the field). She's always playing the victim. I find this kind of behavior toxic, it degrades dialogue.

Re: AI researcher Timnit Gebru resigns from Google

#908

Post by visarga » Sat, Dec 05, 2020, 6:08 PM UTC

iandanforth wrote: post #324

contemporary343 wrote: post #292

It looks like the paper was critical of large language models: https://www.fastcompany.com/90582375/timnit-gebru-google-fir... The original article has been updated with Jeff Dean's response, while a member of Gebru's team (who still works at Google Brain) states that Dean's email is misleading and inaccurate: https://twitter.com/alexhanna/status/1334579764573691904 As I understand it: The paper had already been subm…

This. Failure to cite relevant research is a reason to reject a paper from a conference but that's not the job of an internal review. Research independence is fundamental and why professors are given tenure. Sure companies are companies and don't play by those rules, but you can't claim to be making a fair academic criticism if you're not adhering to academic norms. If the paper was bad it should have been rejected b…

Look, on paper I agree with you, they can't claim 'fair academic criticism' and censure papers without explanation. But in the real world, Google has an outsized contribution to AI. They are also a company, and need to pay for all the expenses. There are practical, day to day concerns that are essential for their existence. If they can't ensure their finances, the research program stops. So you can't really be absolutely impartial, just 99%, from time to time you make an exception to impartiality. It's like light speed, you can't reach it.

Re: AI researcher Timnit Gebru resigns from Google

#909

Post by srajan » Sat, Dec 05, 2020, 6:09 PM UTC

It sounds like Google was at odds with the research and the push by Timnit. And she gave them the easy exit with her ultimatum. It is shameful on Google to use a weak excuse as "not meeting protocol" to let go of a great researcher. Looks like the paper hit home raw. When explained, even a layman can see the bias these AI models will have. Looks like it is time for a better search engine to emerge and call out these monopolies.

Re: AI researcher Timnit Gebru resigns from Google

#910

Post by visarga » Sat, Dec 05, 2020, 6:17 PM UTC

YeGoblynQueenne wrote: post #408

contemporary343 wrote: post #292

It looks like the paper was critical of large language models: https://www.fastcompany.com/90582375/timnit-gebru-google-fir... The original article has been updated with Jeff Dean's response, while a member of Gebru's team (who still works at Google Brain) states that Dean's email is misleading and inaccurate: https://twitter.com/alexhanna/status/1334579764573691904 As I understand it: The paper had already been subm…

What I don't understand is why nobody discusses this from the point of view of academic freedom and research integrity. My boss says "retract your paper" and I retract my paper? In what dystopian sci-fi novel? That's just mad.

> In what dystopian sci-fi novel? That's just mad.

In this world, where there are bills to pay and you can't disadvantage yourself if you want to be ok. Having bills to pay is, I agree, just mad.