We read the paper that forced Timnit Gebru out of Google • Hacker News Board

Re: We read the paper that forced Timnit Gebru out of Google

#501

Post by ThrowawayR2 » Sat, Dec 05, 2020, 4:26 PM UTC

getlawgdon wrote: post #345

mikaeluman wrote: post #219

The problem I find with this has much to do with character. I will not comment on the research (because I do not know the field) but rather comment on why character matters and why motivation is important. Having read her email and tweets; it is quite clear that she is as much a political activist for a far-left “woke” interpretation of the world that I view as of immediate threat to our way of living, democracy and…

"woke" has obviously become much more of a triggering dog-whistle for the reactionary, privileged right than a shibboleth for the left. your comments are not insightful to the issue but ironically they convey your character.

What I see is that a lot of mainstream liberals are waking up to how problematic and alienating that progressives are (even to the point of drawing criticism from former-President Obama https://www.nytimes.com/2019/10/31/us/politics/obama-woke-ca...) and are pushing back against their poor behavior.

If you're a liberal who is tired of "wokeness" (I am and a minority to boot), let people like the parent know that you're not on their side, both through your words and through your voting.

Re: We read the paper that forced Timnit Gebru out of Google

#502

Post by qsort » Sat, Dec 05, 2020, 4:27 PM UTC

free_rms wrote: post #493

qsort wrote: post #479

Earlier quoted context omitted.

> Yes, a linear regression is explainable. I don't agree that this is, in general, the case. Dense linear models trained on a lot of parameters aren't exactly simple to trace, and even the weights aren't as intuitive as they might seem.What about regularization? What about pre-processing? What about feature selection? There are a lot of examples of linear regressions where, for example, you can reverse the sign of th…

Ok, fair enough, the more sophistication you add, the less explainable it gets. That being said, you can look at magnitude and direction of various coefficients and conclude certain obvious things. Even the case where reversing the correlation by picking some features instead of others says something -- those features point different ways but one has a higher magnitude so it 'wins'. They're ultimately pretty simple s…

I agree that there is an intuitive concept of "explainability", which is why I said that a formal model for it would be interesting to see. I also agree that "the average explainability of a deep net is lower than the average explainability of a linear model" (whatever that sentence means), my point is that it's extremely hard to pinpoint what explainability means, because the usual characterization of "you can fit the model in your head" is extremely subjective and brittle.

I'm all for more research in this field. Heck, I'm all for more research in any field, but because our definition of "explainability" is so primitive and unrefined, arguing that we should build ethical or regulatory frameworks on top of it seems premature.

Re: We read the paper that forced Timnit Gebru out of Google

#503

Post by tinyhouse » Sat, Dec 05, 2020, 4:30 PM UTC

kenjackson wrote: post #487

tinyhouse wrote: post #377

Last comment about it since I spent too much time on this story. As a researcher in the field I feel bad people don't give Google more credit (I have no affiliation with Google). They created a research environment where researchers have freedom to work on their own interests and publish papers (they publish more than any other company). You don't find many environments like that outside of academia. I still remember…

This is a poor argument assuming discriminatory practices are happening there. This is like giving Weinstein a free pass because his movies are really good. Note, I’m not saying Google discriminates, but if they do then your statement doesn’t contribute to their defense.

Well, nothing in this story suggest they do. They might have an inconsistent review process, like pretty much most processes in these companies - from interviews to promotions. Let's see how many people follow her and leave Google Research due to their practices. My prediction is zero.

Re: We read the paper that forced Timnit Gebru out of Google

#504

Post by rayiner » Sat, Dec 05, 2020, 4:30 PM UTC

kenjackson wrote: post #475

nathanlied wrote: post #444

Earlier quoted context omitted.

Agreed; but this isn't just a Google problem. Seems to me like a lot of SF (and SF-inspired) "big tech" wants to be known for their "wokeness"[1], which leads to hires like Timnit and other "politically-outspoken" people, which in turn leads to situations like this, James Damore, and other individuals/situations that amount to workplace political activism. I am wholly uncomfortable with any discussion of politics in…

The problem here is that AI and race now intersect in non trivial ways. It’s like having privacy as political discourse when you are scanning emails for marketing purposes. There’s a line at which not having the discussion is equivalent to taking a political stance.

This is true. And the research raised an important point which was worth discussing. But it seems, based on the totality of the information, that Google received an ultimatum demanding that it take particular action with respect to the issue. That’s going beyond an ethics researcher’s role.

Re: We read the paper that forced Timnit Gebru out of Google

#505

Post by geofft » Sat, Dec 05, 2020, 4:31 PM UTC

thinkingemote wrote: post #483

geofft wrote: post #370

Earlier quoted context omitted.

Er, who is claiming " all research at Google is pointless and should stop until reform "? Presumably not the person who was trying to do and publish research at Google? (Are you confusing this with her email that said that participating in Google's feel-good diversity and inclusion work, specifically, was pointless and should stop, so that people could focus on their real jobs - such as research - instead of wasting…

A bit of a mixup I think. Good to clarify it. The Google walkout blog demands all research needs reform (but no stop) but the email did specify that specific field to stop.

Yeah, the email was clearly written amidst frustration and isn't the world's most polished piece of prose, but I think it's pretty clear that the thing it's saying you should do is stop writing DEI docs that meet OKRs and are aimed at an internal audience, not stop doing research:

> What I want to say is stop writing your documents because it doesn’t make a difference. The DEI OKRs that we don’t know where they come from (and are never met anyways), the random discussions, the “we need more mentorship” rather than “we need to stop the toxic environments that hinder us from progressing” the constant fighting and education at your cost, they don’t matter. [...]

> So if you would like to change things, I suggest focusing on leadership accountability and thinking through what types of pressures can also be applied from the outside. For instance, I believe that the Congressional Black Caucus is the entity that started forcing tech companies to report their diversity numbers. Writing more documents and saying things over and over again will tire you out but no one will listen.

(It seems to consistently use the word "paper" instead of "document" for research.)

And Jeff Dean's reply identifies that she was asking to stop the DEI work, not to stop research:

> I also feel badly that hundreds of you received an email just this week from Timnit telling you to stop work on critical DEI programs. Please don’t.

Re: We read the paper that forced Timnit Gebru out of Google

#506

Post by bigbubba » Sat, Dec 05, 2020, 4:34 PM UTC

wan23 wrote: post #395

angry_octet wrote: post #129

Earlier quoted context omitted.

You should read some other ethical AI research. This is par for the course. We're still suffering talks about killer robots.

We already have drones with cameras, drones with guns, and drones that are autonomous. How long until someone combines these?

A few decades ago? What is a fire-and-forget guided missile but an autonomous suicide drone? Some missiles, particularly anti-ship missiles, even use computer vision in the air to identify objects and pick targets. Mark 60 CAPTOR naval mines sit underwater listening to nearby ships. When a noise is classified as an enemy ship, the mine releases a guided torpedo to kill it. Of course autonomy in naval mines is nothing new; the most basic sort 'decide' to kill any object that bumps them too hard. The innovation is making them discriminate between enemy ships and everybody else.

Re: We read the paper that forced Timnit Gebru out of Google

#507

Post by lukeramsden » Sat, Dec 05, 2020, 4:39 PM UTC

UncleMeat wrote: post #462

derangedHorse wrote: post #460

Earlier quoted context omitted.

I've always found it funny how we treat companies like Google as imaginary people with no single individual bearing responsibility. To be fair, some people are aware of the people involved like Jeff Dean, but I look forward to a shift in the future where individuals within companies are as popular and held to the same standards as individuals within politics. Some of these corporate individuals already seem to have j…

> With respect to this particular case of Timnit Gebru, it sounds like she was already on her way to being let go. This is just wild speculation and part of a frustrating pattern where all sorts of questionable behaviors by authorities are justified by other unrelated properties of the victims.

> justified by other unrelated properties of the victims.

Is it unrelated though? Her behaviour in terms of communication, less what she says but more how she says it, is extremely related to this case.

However I do think I agree that it's speculation, at best.

Re: We read the paper that forced Timnit Gebru out of Google

#508

Post by CydeWeys » Sat, Dec 05, 2020, 4:40 PM UTC

sokoloff wrote: post #474

CydeWeys wrote: post #405

Earlier quoted context omitted.

It's really just academic and not germane to the larger overall discussion whether or not she got that one pay period's worth of money.

IMO, it’s germane. “I quit. My last day will be sometime in late Dec.” “We’ll pay you and recognize your employment through that date, but you are relieved of all duties effective immediately” is quite different from “Nope; your job ends today.”

How is that really a big difference? She's gone either way, because they wouldn't let her publish the paper. This way I guess she's more eligible for unemployment, but beyond that?

Re: We read the paper that forced Timnit Gebru out of Google

#509

Post by rayiner » Sat, Dec 05, 2020, 4:41 PM UTC

ALittleLight wrote: post #109

rayiner wrote: post #26

> An AI model taught to view racist language as normal is obviously bad. The researchers, though, point out a couple of more subtle problems. One is that shifts in language play an important role in social change; the MeToo and Black Lives Matter movements, for example, have tried to establish a new anti-sexist and anti-racist vocabulary. An AI model trained on vast swaths of the internet won’t be attuned to the nuan…

"What people say" is itself quite nuanced. Words that two black might say to one another may be casual but appropriate and yet extremely offensive if said by a white person, or, presumably, an AI, in a different context. What the AI should say is hard given that there is no one right answer to the question of what any individual should say. Different contexts change the equation completely. Seems like a nightmare to…

I agree. It’s a very hard problem that we need to thing about in context. Gebru’s research is important for raising the issue.

But it is a hard problem, and highly context dependent. It doesn’t seem to me like a proper subject for the sort of ultimatum that Gebru gave Google.

Re: We read the paper that forced Timnit Gebru out of Google

#510

Post by kenjackson » Sat, Dec 05, 2020, 4:42 PM UTC

tinyhouse wrote: post #503

kenjackson wrote: post #487

Earlier quoted context omitted.

This is a poor argument assuming discriminatory practices are happening there. This is like giving Weinstein a free pass because his movies are really good. Note, I’m not saying Google discriminates, but if they do then your statement doesn’t contribute to their defense.

Well, nothing in this story suggest they do. They might have an inconsistent review process, like pretty much most processes in these companies - from interviews to promotions. Let's see how many people follow her and leave Google Research due to their practices. My prediction is zero.

That’s the beauty of modern discrimination. You never come out and say you’re discriminating. Rather you just apply the rules selectively. An inconsistent process is the best best way to discriminate.

My prediction is you will see some people leave Google due to this. Likely those who worked closest to her. Apple is already extending interest.