Google staff rally behind fired AI researcher
bbc.com
bbc.com
It's becoming clear a rising number of her colleagues are in support of what she was seeking to publish. A reviewer on reddit mentioned the authors have plenty of time to make revisions if there were issues with the content. Not to mention, supposedly the internal reviewing processes are usually limited to the extent of sensitive disclosure and rarely the contents of the literature. Yet, emails from Google leadership seems to imply retraction was needed, providing very little reason.
The opaqueness of the internal feedback and expectations are highly suspicious and I was suprised to see no one bothered to question it. People have been bringing up the 2 weeks, yet internal Brain researchers have refuted that have never seen a policy like this excercised. It does appear as an attempt to suppress an important paper that was critical of another colleagues work (BERT) and the employer's intended research goals. It's similar to the kind of pushback Greenwald most recently received - yet the reactions here have been distinctly flipsidded - in support towards Google executives no less and unsympathetic to the researcher.
It's never good enough.
[1] https://ftw.usatoday.com/2018/05/nfl-49ers-colin-kaepernick-...
I’ve read some of her research (that’s on arxiv) and it’s... not on the level that “Google AI researcher” implies. It’s good that they hire from the Humanities, but we tend to smush all academics as if they all had equal rigor and value.
Maybe she’s a great activist. Or a great cook. Being it that I’m not into activism, I can’t judge that.
----
Edit: I have to hijack my sudden spike in karma (and, I assume, visibility) to "amplify" some important research milestones in the Humanities that impact the problem of "AI ethics" taken seriously.
[General]
* Baudrillard - "Forget Foucault"
* Delanda - "Philosophy and simulation"
[Science and technology issues]
* Latour - "The Pasteurization of France"
* Latour - "Reassembling the social"
* Caillon - "Acting in an uncertain world"
[Computers and society]
* Various authors - "Software Studies: a lexicon", available here https://monoskop.org/images/a/a1/Fuller_Matthew_ed_Software_...
Gebru has a BS and MS in electrical engineering, and a PhD in computer science, specializing in artificial intelligence; all from Stanford. After receiving her MS, she worked at Apple's hardware engineering group. After her PhD, she worked Microsoft's AI research lab.
Maybe the problem is precisely that they didn't actually hire from the Humanities? Although the cynic in me thinks they just hired someone with a "SJW" discourse rather than find themselves criticized for their market strategy -- whether their use of their monopoly power is ethical.
As it stands it really sounds like they got someone to offer Yann LeCun as a scapegoat for AI products (Amazon Rekognition is infamous); these services then serve as scapegoats for police departments, which...
I honestly don't know how to take this comment.
Are you comparing her work to cultural theorists like Baudrillard, etc.? Because a brief read through her papers and they seem to be dealing with common-sense ethical concepts, not French social critique.
Otherwise, this comment is just odd - no real specific criticism of her body of work except to suggest that it is less good than what you'd expect from a "Google AI researcher".
Gebru is a successful "scholar" as far as the credentials game goes; I don't presume to have anything to teach her.
Edit: Are we sure calling something "common-sense ethics" is a common-sense proposition? Are we comfortable deriving this Ought from that Is?
Again: back up your assertions or they are worthless. I'm reading her papers and have also read many of the authors you are mentioning, there is very little in common.
> Are we sure calling something "common-sense ethics" is a common-sense proposition? Are we comfortable deriving this Ought from that Is?
So your critique now is that her papers are steeped in a moral realist perspective? It seemed you were going for the dogwhistle-y "diversity hire" rhetoric, but I'd welcome any actual critique on the merits.
I'm not proposing a critique of her work. What I've found on arXiv reads like standard fare for its genre. Hotelling famously joined an early seminar on linear programming with Dantzig only to interject "... but the world is nonlinear!" That's not what I'm out to do -- not with this particular researcher, nor with its proxies on twitter and HN.
The scenario I'm looking at is one where the polarities are not left versus right or woke versus straight, but Humanities versus STEM. If people are to reject her kind of work -- and I'll be glad if they do, even if I'm not really willing to press them in that rejection with a proper "critique" -- then they could walk away burned by theory and postmodernism and all such crap. I'm proposing an entirely different path, or a space of different paths even.
Conjecturing what could be the real reason, person A conjecturing why a person B does what he does (hates our freedom, for example) and why one should not take B's own explanation of his own actions -- these to me are an avoidable form of discourse. Nothing good comes of it other than an obligate outrage.
Welcome to engineers playing organizational politics.
As someone I saw post on Twitter relate to this... the best way to discriminate is to have rules, but apply them discriminately. So you can always point to the rule, but it really only applies to certain people.
And HN falls for it every single time.
Oh yeah, we totally didn't fire this Googler for the unionization push, it was because he looked at one of his bosses' publicly available calendar.
By definition the default position was "leave the company"
I agree with you that it was a "firing", we can debate over whether it was justified, but it wasn't a resignation.
Isn't that technically resigning if those things aren't given to her?
This was the email she sent, as a manager at the company.
Here's the full paragraph that that is from:
> Imagine this: You’ve sent a paper for feedback to 30+ researchers, you’re awaiting feedback from PR & Policy who you gave a heads up before you even wrote the work saying “we’re thinking of doing this”, working on a revision plan figuring out how to address different feedback from people, haven’t heard from PR & Policy besides them asking you for updates (in 2 months). A week before you go out on vacation, you see a meeting pop up at 4:30pm PST on your calendar (this popped up at around 2pm). No one would tell you what the meeting was about in advance. Then in that meeting your manager’s manager tells you “it has been decided” that you need to retract this paper by next week, Nov. 27, the week when almost everyone would be out (and a date which has nothing to do with the conference process). You are not worth having any conversations about this, since you are not someone whose humanity (let alone expertise recognized by journalists, governments, scientists, civic organizations such as the electronic frontiers foundation etc) is acknowledged or valued in this company.
If a manager within the company wants to publicly express her frustration with leadership, then this is an alright message to send — leaving aside the very implicative final sentence.
It still seems reasonable to fire anyone who sends such a message to their colleagues rather than working through the office's politics more quietly, but it also seems reasonable for staff to protest and rally behind that person in support of them.
Can you unpack this a bit? Several employers, past and present have encouraged employees to communicate and attempt to resolve conflicts (presumably, maybe ostensibly, giving them opportunities to work it out as adults and professionals) amongst employees first as the first actionable steps. In fact, current employer has this as a point of policy for interpersonal conflicts barring severe enough circumstances where a reasonable person would say "yes this requires intervention from on high".
That said, why should it be any different in a situation such as this? I feel there may be context to your suggestion that isn't fully stated here-as to why it's reasonable to fire someone over this. Maybe there are implied parameters of just how egregious a situation has to be in your opinion that I didn't intimately tune into?
To me the final sentence reads as a continuation of the hypothetical the author is asking the reader to consider, as a metaphor for the treatment she experienced, not a direct j'accuse to the reader/recipient.
I completely agree with you, but what happened here was that this manager sent this email to a large group of employees, not the specific employees she had the issue with (which would be fine!).
This email wasn't an attempt at conflict resolution, it was sharing how upset she was, at her manager, with a bunch of coworkers.
For the last sentence: in the event where one wants to complain to a large group of coworkers, I think it's pertinent to stick to the basic, unarguable facts of what occurred. Suggesting that you've been dehumanized by the company is adding fuel to a fire that you should instead be aiming to resolve privately, if you want to continue to work at the company.
or
Is it perhaps in fact more reasonable to fire them using this and other behavioral or disciplinary incidents? Of which-I must say, in terms this Googler, none of us are privy to
It'd be acceptable if it's a relevant party, like your skip or your manager's skip, even if it would anger your direct manager. This, however, is entirely different. It's venting your frustration to a large group of coworkers. It takes away from productivity as a whole, and is socially inappropriate.
I don't know about Google's management policies, but at most companies I've worked at, senior employees and managers are held to an even higher standard, and this would make firing a more likely choice than some type of correction.
Regardless of the hypotheticals, from what I've read, she offered an ultimatum, and they declined her ultimatum, which resulted in her being fired:
> Dr Gebru had emailed her management laying out some key conditions for removing her name from the paper, and if they were not met, she would "work on a last date" for her employment. According to Dr Gebru, Google replied: "We respect your decision to leave Google... and we are accepting your resignation.
I absolutely do not believe it is, barring mitigating circumstances or policies that neither you nor I are privileged to access.
In most cases, this is called "venting". I have NO doubts my subordinates probably talk about me in DM or have gripes about some of my issues as their leader. But I know and have seen my direct reports be very opinionated with me in 1:1, they are very outspoken with me. One person had a moment of raw emotion, not because of something I had done but the organization at large.
And I encourage all of it because it means I can trust them to be honest with me when I need that candor, and in return, because I meet people where they are, give them room for expression and authenticity, they know they can trust me not to bring retribution upon them for being opinionated, even amongst themselves. I might not always be able to move the needle in the direction they wish, but they know I will always make sure their opinions have a voice at the leadership table-even if it means I am the one who has to take the hit for it while we continue trying to meet our mission as Ops Engineers. That's Leadership. Capital L.
Among larger groups? So what? No really, so what? Can I tell you that I am a founding and sitting member of an org wide DEI initiative and we actively encourage affinitized groups of employees to form resource groups, kvetch among themselves and present areas of contention to us to enact positive change at large? That we have two diverse members of the executive team on this initiative? That it has the full support of our board of directors and even our group of investors?
Would you believe me if I told you we had five people accept offers at the company who specifically said the reason, not one reason, the reason they accepted the offer was learning that this was something our organization has in place?
Now:
This is not most cases-granted, but I am regressing my interpretation to the mean intentionally because reacting to outliers in any direction as a default response gives room for extraneous and unnecessary harm. I default to maximizing good and giving people a voice, not looking for a reason to exclude them from making the business better for customers and employees with equal vigor.
Microsoft last year had a similar thing, but related to sexual harassment [0]. I've never heard anyone fired as a result of airing their grievances. And to the contrary it sounds like, at least to the public media, they really tried to embrace and discuss what was at the heart of these encounters.
There's a problem at Google, and I don't think they solved it by firing this researcher.
[0] https://hub.packtpub.com/microsofts-metoo-reckoning-female-e...
Managers act as agents of the company and the amount of legal assfucking that statement opens the company up to is enormous.
Well thank god I don't work for you. Most tech companies don't fire everyone engaging in employee activism or they would have much larger labor issues to deal with.
While it sounds difficult for her, I can't imagine how Google's behavior here amounts to not acknowledging someone's humanity.
Reading the HN comments section sometimes feels akin to doing archival research on letters to the editor during the peak of the civil rights movement.
It's astonishing how consistent the tropes are.
I think it is very illuminating to read letters from that period, because they aren't just vitriolic racists like schools often teach. These are people trying to come off as "very reasonable", offering a civil, moderate path forward for the South, etc., etc. It's fascinating to see that rhetoric being deployed to defend what is an obvious evil and compare to who is using that rhetoric now.
You are the one bringing up comparisons to Nazis, so it appears that you have now fallen prey to Godwin's law.
Don't really know how else to respond to vague accusations of engaging in a "divisive and dogmatic strain of diversity advocacy."
So let me make some arguments for you to consider.
-To declare that certain people (on the basis of their minority status) have inherently privileged knowledge of boundary-less, undefined, fuzzy social dynamics is a divisive assertion. It seeks to divide authority along identity lines.
-To advocate that certain people (on the basis of their minority status) should actively be given privileged decision-making status over collective processes or structures is a divisive and dogmatic advocacy. It seeks to divide power along identity lines.
-To denounce as ethically compromised the people who disagree with you on complex open topics, such as the above two notions, is outright dogmatic. To assume one's perspective is so undeniably true that discussion can only be aggression is a dogmatic position.
I contend that all three of these dynamics are rampant in the preferred methods and among the prominent leadership of the currently en vogue Racial Justice and DEI movements (lots of crossover).
To tie it back to the start: I've observed Timnit engaging in these things, fanning these flames and boosting others who do so too.
But note that my comment implies there are other ways to advocate for diversity and justice. The broad BLM/Kendi/DiAngelo/Gebru style is not the only form and theory of diversity advocacy.
A simple alternative, for example, would be to call attention to the same problems without 1) preemptively attacking other people based on their own immutable identity characteristics, and 2) trying to draw new lines and hierarchies of privileged voice and power.
Lots of people (I suspect the majority of the sympathetic) practice this alternative. However, the louder, Timnit-aligned strain seems to enjoying social privilege in this area at this moment. How ironic.
In order for the comparison to be informative, it must also be accompanied by some argument about whether or not the accusations are accurate, which sort of takes us back to the original question.
I was only trying to point out the parallels in bad argumentation between now and then.
Top level comment said, paraphrased, "it's a shame people dismiss her as a Twitter personality when she has a highly respected body of work in the field."
My reply was to press that, in my view, she's not being dismissed as "just a Twitter personality", but rather her Twitter behavior is worthy of examination in itself.
Finally, I've never read civil rights era letters to the editor, so I'm not referencing or borrowing any specific rhetoric from there. I'm using bog standard English language words to describe the arguments and advocacies that I am seeing in the present.
Actually, you're right, but you're making the opposite point than you think you are. I wasn't alive in 50's or the 60's (and I was too young to even remember the 70's), so my whole life I just took for granted that what the history books had to say about racial inequalities at those times was true. It's hard to miss, though, that people are saying the exact same things today when I can look around and observe that they aren't true and haven't been at any point in my lifetime. It doesn't make me question my perspective on the world as it is today, it makes me wonder how true the history books reflected reality at that time.
> it makes me wonder how true the history books reflected reality at that time.
What parts of the history books do you think might not be true?
Stanford Business School professor Jeffrey Pfeffer wrote a book about this[1] in particular.
[1]: https://www.goodreads.com/book/show/24331490-leadership-bs
A book about how toxic this is and how it needs to change, right? This small detail seems sort of missing from your comment.
"Corporate culture" isn't fixed, the whole point is that it needs changing.
Hacker news also requires a certain amount of politeness in the way you speak and it's one of the many reasons the discourse is better than twitter.
Doesn't this cut both ways? Google employed her to critique the company's ethics; she is critiquing the company's ethics - as such, she is literally doing the job she was employed to do.
If Google doesn't want that role to exist or wants a puppet to do it, why did Google employ her in the first place?
Maybe Yann LeCun was wrong, the way she conveyed the message disregarding any knowledge he might have on this was definitely not civil.
She could have chosen a different way to address the complaints here, not by getting overtly confrontational. Maybe the review group got it wrong, maybe they were right. But that is not how you handle disputes.
I agree that people should try to be kind and charitable to each other, and this conversation didn't show that.
That said, I think we also have to be cognizant of the way "civility" is not objective/neutral, often relies on codifying the norms of a dominant majority, and also preserves the status quo.
It might not have been very "civil" for instance when Google employees walked out of the office, but I also don't think it was "civil" when Google paid $90 million to someone fired for sexual harassment.
Sometimes big problems can't be met just with civil discourse, it's why many of us still voted for Biden (for instance) even after he told Trump to "shut up."
If they had fired him under more favorable circumstances his payout would have been significantly higher.
I don't understand your point of view here. If I sell something to someone else, I'm not expecting to get more payout for it fifteen years down the line unless some kind of residual was explicitly negotiated in the original sale agreement. Acquisitions would never happen at all if the acquiring company was on the hook for unbounded amounts of money decades later just because they made it successful. This point of view seems very un-businesslike. Andy Rubin was not owed that $90M, it shouldn't have been paid to him, and Google has admitted to its mistake and will not be paying out in similar cases going forward.
2. It doesn't matter what their deal was. It matters what public perception is. I'm talking entirely about public perception.
Regardless of what their deal was, Google has to send a signal that _impacts future deals_. If the general developer public thinks that Android was acquired for too little, then we will all be reluctant to sell our projects to Google. That's the whole point.
It has nothing to do with him thinking Google treated him fairly. It has to do with whether we think we could keep a company going long enough to make more money than his payout on _our_ work.
Google needs to make more Android-like deals in the future.
> If the general developer public thinks that Android was acquired for too little, then we will all be reluctant to sell our projects to Google. That's the whole point.
I'm still not understanding the point here. If you agree to sell your company for price $X, it doesn't matter who you're selling to; you've sold for that amount. If you think you should get more than $X, then don't sell to Google, or anyone else for that matter, for only $X. If you want less up front but more money over time as part of an incentive structure, then negotiate for that as part of the acquisition (which Andy Rubin did do and was paid out for). All of this is independent from a severance payout fully 15 years after the acquisition.
I think you'd be hardpressed to find a founder that wouldn't sell to Google for what they consider the right price if he didn't get $90M. Do you think any football players don't want to play for the Patriots, because Aaron Hernandez wasn't supported by them? People get that all bets are off with criminally poor behavior.
https://www.nytimes.com/2018/10/25/technology/google-sexual-... https://www.theverge.com/2018/10/25/18023364/google-andy-rub...
Also, take a step back and consider how there was a worldwide walkout of 10k+ employees. That wouldn't have happened if Google's hands were tied here and they had been contractually forced to pay that money out to Andy Rubin. The outrage was because they knew about his misconduct and yet still chose to let him "retire" and give him a big amount of extra money, rather than firing him and giving him the $0 he was legally entitled to.
In settling on terms favorable to two of the men, Google protected its own interests. The company avoided messy and costly legal fights, and kept them from working for rivals as part of the separation agreements.
Real life isn't a melodrama. I suspect there's plenty of nuance here.
> If someone walks away with 8-digit sums, there were contracts involved
The NYT article is pretty unequivocal that that is false, even if you want to pull out quotes talking about different issues.
> paying them millions of dollars as they departed, even though it had no legal obligation to do so.
They had no legal obligation to pay the severance.
If you want to discuss a different reason to pay $90 million to someone fired for sexual harassment, fine.
But your original comment was unequivocally wrong, let's just recognize that.
In this case, the norm in question was respectful argumentation grounded in empirical evidence. Why does it matter that that norm belongs to a dominant majority?
It matters
a. if the norm is actually universally applied - I find that this norm of respectful argumentation is not universally applied.
Remember: this is a twitter argument entirely unrelated to what Timnit is under discussion for now. Elon Musk and Linus have had unrelated arguments where they engaged in very offensive argumentation (calling their opponents pedophiles, telling them to kill themselves) without any evidence, yet I rarely see these issues brought up on unrelated issues about them. But for Timnit, it's a scarlet letter she has to bear.
b. I think we substantially underestimate how frustrating it might be to be a scientist of color where people are constantly discounting your expertise (because, yes, Timnit is an expert.) I'm sure seeing that same sentiment for the ith+1 time can evoke strong emotions and we are not necessarily seeing the context.
Regardless, maybe not an incredible look for Timnit but not super huge in the context of a large body of work.
https://news.ycombinator.com/item?id=23739076
https://news.ycombinator.com/item?id=18000698
https://news.ycombinator.com/item?id=18083859
https://news.ycombinator.com/item?id=16508921
https://news.ycombinator.com/item?id=8415603
https://news.ycombinator.com/item?id=2372096
I think the threads in articles specifically about Linus are better comparisons to this, rather than articles about computing that involve him. But, hey, folk like Torvalds and Gebru both really rub me the wrong way, so it's reasonable that I might be selectively remembering threads where HN said I'm right and that he's an asshole (and I did a lazy googling of "linus torvalds asshole site:news.ycombinator.com" which is biased too), so who knows.
That's how woke Twitter works: you define the group of people who are allowed to speak as only you and your allies. You pursue the meta-argument whenever someone not in your group attempts to speak.
Aggressive, sure, but also seems:
a. reasonable for AI ethics experts to "lead" in issues surrounding AI ethics
b. not the cardinal sin of discussion that it appears to be viewed as here.
You'll see people here defending Linus or Elon Musk for saying objectively way worse stuff than this.
Somehow Timnit is defined by this discussion, but Elon isn't defined by calling a rescue diver a pedophile with no evidence or Linus telling someone to kill themselves.
> the expert in your field and you have work for this person, who controls your career and compensation
Must be harder when your CEO boss has a tendency to call domain experts pedophiles, yet I see Musk praised frequently here.
Your description above sounds like gatekeeping to me - not saying that is the researcher’s intent - but it doesn’t seem to make for good science
If you work in an area that is written off, you need a strong advocate, and from everything I've seen, she was that strong advocate both for herself and for those who reported to her.
She says talking down to an actual expert in ML who gave an opinion on ML
I can see how this would get frustrating fast if your opinion is held among many Google engineers.
To use an example that comes up: do we solve the ethical qualms of a system that predicts incarceration rate based on facial structure by getting a better dataset? Is it even ethical to attain such a "better" dataset, what does "better" even mean in the context?
A: Training data is a cause of this bias
B: But not the only cause
A: True
A: Training data is a cause of this bias
B: I am so sick of this. You are not permitted to speak.
A: I am deleting my Twitter account.> You are not permitted to speak.
I really don't think this is a fair reading of her tweet. I don't see any way to read her actual tweet claiming Yann can't speak at all on the subject. I do see a plea to engage more with domain experts (which to reiterate: my understanding is that Yann has a history of ignoring on the subject)
But if you change to a different topic like privacy then these same people really get all of this nuance that they completely seem to miss here...
Also, to quote LeCunn: “I’d like to ask everyone to please stop attacking each other via Twitter or other means. In particular, I’d like everyone to please stop attacking @timnitGebru and everyone who has been critical of my posts… Farewell everyone.”
What is the evidence that this has to do with the content of the paper?
That's called firing.
> when someone gives 2 weeks notice.
But she didn't give 2 weeks notice. This seems to be the key detail.
If you give an employer an ultimatum and they don't want to give in to your demands then is that giving notice or quitting on the spot or something else? Should her managers/google legal have waited for her to tell them when exactly she was going to leave?
Yes, if they wanted her to resign they should have waited for her to resign rather than fire her. If they fired her, seems odd to call it a resignation.
As an employee, I agree, it is twisted to call it a resignation.
As an employer, of course that's exactly what they'd do with any employee who says "I quit, I'll tell you exactly when" and then proceeds to blast out a disruptive email.
You cannot file for unemployment aid if you resign or are terminated for cause. You can only get assistance if your termination is the result of your position being eliminated, otherwise known as a reduction in force. This is valid in California and I believe true in every other state in the US.
> Frequently after notice of a quit, layoff, or discharge has been given, either the claimant or the employer will act to accelerate the separation. Such action can alter the character of the separation.
https://www.edd.ca.gov/uibdg/Voluntary_Quit_VQ_135.htm
I don't really think unemployment is relevant in this case. I am interested in whether she technically quit or resigned though.
Let you continue to have access to company resources? My expectations are they would immediately terminate access to all company resources, pay me 2 weeks, send me a letter wishing me good luck on my future endeavors, and send an email to my co-workers saying I resigned.
Now later I might think, shoot maybe I shouldn't have threatened to quit if they didn't meet my demands because I really liked my job. But I don't think we should pretend this isn't exactly what 95% of all corporations would do in a similar situation.
I think the wishing of luck was circumvented by the media firestorm. Her first tweet was about losing corporate access which is usually the first thing H.R. does once a decision has been made.
The employee remains employed in your scenario. Just without duties or access. Terminating employment immediately and not saying anything about severance implies not paying.
Both the positive aspects of her contributions and the clear negatives of her approach were rightly pointed out by various people (as well as problems with process and politics at Google); perhaps your reading of a long and complicated morass of comments was unintentionally selective.
I personally find HN to be generally liberal-leaning but not excessively so, and have noticed a welcome openness to centrist opinions that might once have been dismissed here out of hand as right-wing (and are still being dismissed elsewhere) that I take as a good sign of an increasingly aware local commentariat with respect to the excesses of the more ideological elements of the left and right.
The big online communities I'm a part of are twitter, reddit, hacker news, and facebook.
It's true that hacker news seems to be different from those when discussing culture war topics. With hacker news being father to the right than twitter and reddit, and more center right than my facebook feed.
One conclusion is hacker news exists in a right wing bubble. The other is that twitter and reddit exist in a left wing bubble. Given that the median voter in the last election voted for Biden, but just barely, and maybe voted for a republican senator. And even the people arguing her being let go was justified are probably pretty center or center left. I would guess your "elsewhere" exists in a ideological bubble that is much stronger than the one hacker news exists in.
I am in a bubble where I mostly talk to college educated people and people who have educated themselves in a similar tradition or are well-read. Perhaps it is snobbish of me, but I am mostly okay with that bubble for discussion forums.
Conditioning on that type of person, who make up most people on HN, Reddit, and Twitter, HN seems like it is in a right-leaning bubble.
> Given that the median voter in the last election voted for Biden, but just barely, and maybe voted for a republican senator
Your assessment of the median voter is being super distorted by the geographic layout of the US, btw. The median voter did not vote for a republican senator.
Even if in the long run it turns out your position becomes the mainstream, that isn't necessarily nearly the case now.
This is why your original "elsewhere" is being questioned as it's likely not a representative baseline in that regard.
Whatever HN is, it certainly isn't right-leaning.
> Conditioning on that type of person, who make up most people on HN, Reddit, and Twitter, HN seems like it is in a right-leaning bubble.
If most of the people you talk to are college educated people who live in a city than you most definitely exist in a bubble. Remember 47% of the country voted for Trump(I doubt Hacker news came anywhere near this number). Of your friends how many voted for Trump? As an example I'm a college graduate in his 30's who lives in the city. 0% of the 20 people I talk to most often voted for Trump, and maybe 4% of the the extended 200 people I know, voted for trump. And of that 4%, 3/4ths voted for him begrudgingly. And I live in a southern state that voted for Trump. So when you look at hacker news and encounter opinions that are different than your own it might seem like Hacker News exists in a bubble, but I think it's far more likely your life exists in a bubble that's stronger than hacker news.
> Your assessment of the median voter is being super distorted by the geographic layout of the US, btw. The median voter did not vote for a republican senator.
The median senate voter would have voted for a Republican but that's mostly because California and New York didn't have senate races this year.
Why am I getting so many comments like this - I am pretty explicitly stating that I am aware that I am in a "bubble", my social circle isn't perfectly representative of America. That said, I think that restricting the forums I frequent to one's with primarily people who are well educated is a defensible practice.
My claim is this: among the sort of forums where people are actually having discussions about Timnit Gebru's firings (ie. highly educated crowds of people who pay attention to the goings ons in AI ethics, Google, etc.), HN skews right-leaning.
All this talk about the "median American" is irrelevant to what I'm talking about because I don't really care to be in a discussion forum with the "median American", particularly post-Thanksgiving.
There are two groups of people who are interested in this firing. People who are very interested in race and race relations. These people skew extremely left wing.
And people who are interested in technology, ai and machine learning who are probably center left.
The machine learning subreddit and hacker news are both part of this second group and only look like it's skews right leaning because the other discussion skew so far left leaning.
No academic of integrity could accept that as a work environment, doubly so if you were hired as Ethical AI Lead. It simply indicates to me that Google as an organisation is no longer capable of having a serious Ethical AI Lab.
Even if we take it as fact that she was somehow going to be releasing something very damaging and that Google was weary of that (which I already think is quite generous to her), I would imagine that her role as a researcher is not to be an in-house whistleblower but rather to be someone who helps Google mitigate the biases etc that they may be perpetuating. If that's the case, her first job is to serve Google and help them address these shortcomings and her second role (if there even is one) would be to publish this for the community at large.
Most companies will not let people publish their findings at all, and I think this is something that sets Google and a few other companies apart from the pack. The fact that there even is a mechanism to share results outside of your employer is really rare.
So, even if Google said "wow, this is spot on and will make us look terrible", they should be 100% able to tell her "don't publish this, instead let's work to fix this and then possibly publish a post mortem".
Google is ultimately a company, not a democracy.
Saying a brilliant asshole is a net negative isn’t trivializing their work. It’s saying the work isn’t the most important factor.
Timinit has been showing her aggressive side without much self awareness for a long time. Instead of relying on her research work and data, she routinely try to shut down the opposite side's "matter of fact" statements.
In contrast, the proclaimed "bad" guy, Jeff Dean, never do that. I know a co-worker who is level 4 engineer, reviewing Jeff's code changes, with Jeff never show a slight hint of his status. At the moment Jeff is level 11 engineer.
If Timinit could be that patient, then given enough time, she will make impact equal or beyond Jeff as in pure engineering. Instead, she tried to let loose her emotions.
Her continue staying at Google and the ethical AI research community would become a long term risk and overall negative factor, given the trajectory. That's my pessimistic prediction.
I see different claims from Google researchers, but it seems that at least for some internal review is used to access quality of the work [1,2]. It might potentially harm the freedom of the researcher, but once again, unless I missed something, there is no evidence that Google asked her to withdraw the paper and never resubmit.
As I already wrote in different comment, I can not think of a single reason to ask for reviewers identities unless she wanted to confront them personally.
[1] https://twitter.com/julianibarz/status/1334623651719643141 [2] https://twitter.com/vivnat/status/1334710506045669376
There is an endless river of souls being callously, unfairly fired every day. The only reason we are even talking about this one is because it is controversial, and has been blasted out on social media. Gebru seems like a smart cookie to me, but being smart, hardworking, just, kind, or any other number of other adjectives simply doesn't matter. Google, just like every other profit driven company on the planet, only cares about the value you provide. They are only different in degree from companies running sweatshops in SE Asia, and if the law allowed it, there would be _zero_ hesitation to bring that reality back here in the US. People in tech would get whiplash from how fast the industry would pull that u-turn.
We want to believe there is a social contract, that Google, or the tech industry, or America, or some group of people, somewhere, is above the competitive fray, and would not stoop so low as to abuse us for profit. Even at the very heights of privilege in the richest country in the world, working at one of the richest, most enlightened companies in the world, you are still getting exploited, and nothing will change that as it stands today. Google will probably keep investing in AI ethics, as long as people keep believing that actually makes them committed to some kind of standard.
In Jeff Dean's email, he made one point that could be pretty salient: she submitted the paper for approval right at deadline. If that's true, that is poorly handled.
The ultimatum is easily a step too far.
And all of this right before a vacation/holiday.
That's not poorly handled on her part. If she'd submitted it late, that would be her problem. If getting a submission right at the last valid time to make a submission (the deadline) is an issue then it's a problem caused by whoever set the deadline, not the submitter.
JD> Unfortunately, this particular paper was only shared with a day’s notice before its deadline — we require two weeks for this sort of review — and then instead of awaiting reviewer feedback, it was approved for submission and submitted.
Her job involves ethics, if you're not stirring the pot a bit you're probably not doing ethics and just acting as PR for your company?
*Also, while my reply is source/fact free your comment is as well, so I think they're both reasonably at the same level.
I know her from her communications with LeCun, and from the emails she herself (or someone allied with her) leaked, that were posted in the original article outlining her "firing".
From those things, I know enough that I would avoid her at all costs because of her toxicity. You can have the loftiest ideals and goals in the world - that doesn't mean you can act with impunity and any action of yours is unquestionably good.
The story could just be "well known Google ethics research quickly let go over paper dispute" but instead the "also she had an altercation in July which really riled people up."
Does the twitter dispute context help as much as her contributions to her field and whatever the specific paper and terms of its publication?
No clue personally, but now I know it's a "bad thing" and she's a "rabble-rouser that likely had it coming".
I think we should definitely acknowledge that politics do play a role in these mega corporations. This is especially true in the ML field, where choosing something as "innocent" as a training data set would most likely have political and societal implications.
But that being said I don't think the combative cancel culture is helping anyone.
Linus still has a job, AFAIK. So does Lennart Poettering. Both have plenty of defenders on HN.
That's not to mention people like Peter Thiel and his fans.
And a bunch of them weren't actually hostile to his team, they just got taken out of context. He'd refer to a userspace program they had to keep from breaking as a "braindead abortion" or something and people would report it as if he said it to/about a kernel developer.
I feel like the industry has at least made big leaps forward. Went I joined (during the great recession) it felt way worse, but I've also found a great company, so maybe I'm insulated in a kind bubble.
Firing her much sooner than the date she requested seems really foolish on Google's side. Arguably, one of the primary reasons a profit-oriented company would hire someone for this kind of job is for the good PR. If they offer to resign why would you not simply accept the terms of resignation, why would you rush it to happen faster.
Even if she is acting in bad faith Google could have saved themselves a big headache by at least giving her a month or something, even two weeks, to exit the company.
Yes, she can complain on Twitter and still do some harm; but without her, management at least has a chance to control the internal conversation.
NOTE: I am not trying to take sides as I don't know enough information. Just saying I believe there are good reasons why you'd let someone go immediately.
Edit: also you put the stuff about 'agreeing a leaving date' instead of 'I will quit' in your ultimatum to make it clear that you're dead serious about leaving and are trying to be reasonable rather than threatening, not because you need next month's paycheque.
https://syncedreview.com/2020/06/30/yann-lecun-quits-twitter...
I bet one could get "hundreds" of Google employees to rally behind a flat earth and many more for the abolition of mayonnaise.
Thats objectively one way or the other - it's not a matter of opinion.
There's still a ton of people, including her, who think that she was fired. The amount of pushback from those folks for saying that she resigned has silenced a lot of people, from what I've seen.
It would be courageous to do the opposite.
Americans really want full-blown Communism, but are too afraid of speaking out for fear of retaliation from jack-booted capitalist thugs.
Prove me wrong.
Or, more likely, they don't are about the whole "woke" movement and don't want anything to do with it.
Perhaps the people who are not speaking out in agreement with you simply don't agree with you.
I do think there's a larger set of questions being asked right now, around how organizations deal with differing viewpoints among their constituents, and how constituents dissent: https://jakeseliger.com/2020/12/03/dissent-insiders-and-outs....
Especially considering that we don't know how many people would have "rallied" in favor of her firing (if there were such a thing), even if we could believe that they could do so without fear of reprisals as those rallying "against" it can.
The flip side is that you have to respect the researcher, who frankly has more credibility than you do as an organization in this space.
Even to corporate researchers, demanding (through an HR procedure? bizarre) the researchers retract a paper without a discussion or an opportunity to revise it (which is still possible since the paper is only in review!) is highly unusual. To do that to a researcher whose work you loudly (and seemingly proudly) advertise, is insulting to them and the broader team they hired. And to do it over not citing literature? Unheard of and clearly indicative of something else going on. This seems like some kind of turf war.
You can read the abstract of the paper in question here - it is incredibly anodyne, though obviously does take a critical view of Google Brain's work (BERT in particular); https://www.reddit.com/r/MachineLearning/comments/k69eq0/n_t...
You submit a final version for review. You cannot change a paper significantly after it was accepted.
I say this from personal experience, I once published last minute for a conference and was violently chewed out by my bosses because of things they disagreed with in the paper.
I don't get why anyone is surprised that Google doesn't want a paper saying ‘we should spend less money on ML’ published under their name, when the whole company is ML driven and pushing for growth there.
What other large company would allow that sort of thing?
‘Train for less time,’
> investing resources into curating and carefully documenting datasets rather than ingesting everything on the web
‘on less data,’
> carrying out pre-development exercises evaluating how the planned approach [...] supports stakeholder values
‘while spending less’
> and encouraging research directions beyond ever larger language models
‘on smaller models.’
Explain
Unless the underrepresented people got their deserved social and economic status, these so called fairness ethical study would just be building a mansion over empty air.
I am also wholly assured that the rich and powerful progressives know very well of this truth. And they are very willing to indulge these self-righteous elite intellectual. As long as they do not roll up their sleeves and get to help the unprivileged to rectify the system.
> Unfortunately, this particular paper was only shared with a day’s notice before its deadline — we require two weeks for this sort of review — and then instead of awaiting reviewer feedback, it was approved for submission and submitted.
There seems to be something between the lines here. How was it approved for submission? Did Gebru or one of the other authors approve it? And if so, would that be unusual?
> A cross functional team then reviewed the paper as part of our regular process and the authors were informed that it didn’t meet our bar for publication and were given feedback about why. It ignored too much relevant research — for example, it talked about the environmental impact of large models, but disregarded subsequent research showing much greater efficiencies. Similarly, it raised concerns about bias in language models, but didn’t take into account recent research to mitigate these issues. We acknowledge that the authors were extremely disappointed with the decision that Megan and I ultimately made, especially as they’d already submitted the paper.
Does this contradict what Gebru has said so far? Didn't she say she wasn't given any feedback?
If you say "I quit unless...", and they don't agree to the conditions—you've quit. If I'm running a business, I certainly have no use for an employee who has quit under fairly unfriendly conditions. I definitely wouldn't want someone who has quit spending another month or so using corporate resources for what is now their own personal agenda.
I don't think Google is in the clear here either, but don't I think calling her a fired AI researcher is realistic. The story here is an AI researcher quit because Google wouldn't let her publish.
Also "AI researcher was let go after submitting resignation letter" is pretty close to accurate.
But if we had to choose between "AI researcher resigned" and "AI research was fired" I think the first would give you a more accurate idea of questions like
"did she submit a resignation letter" (yes), or "who initiated the break up" I feel resignation tells you she initiated it, which I think that is closer to the truth Google initiating it.
For all we know she is still getting paid for X weeks but was asked to not come in. Unless she files for unemployment it doesn't really matter though.
If you weren't serious about resigning, you shouldn't have made the threat.
Google has just about 100,000 employees + 120,000 temps/contractors. If hundreds - lets say 500 employees - sign something (aka click a button), currently we're sitting at .2% of the company that feels a certain way.
What other petitions could you get .2% of the company to sign onto?
"Google staff rally behind overthrowing the US government to create a Marxist utopia"
"Google staff rally behind mandatory veganism in the US"
"Google staff rally behind anti-vaxxers"
etc
https://googlewalkout.medium.com/standing-with-dr-timnit-geb...
Even if they were, their number would comprise a rounding error of Google's total head count.
I'd like to know what was technically controversial.
For an extreme example: I could say "I got tazed and arrested for doing 5 mph over the speed limit". That sounds insane. That cop should be fired and in jail! What I haven't told you is: I was doing 5 mph over the speed limit, and I was just going to get a warning from the cop, but I became combative and tried to physically fight the cop, at which point I got tazed and arrested.
It's also possible she was fired for the paper. But she has published many papers while at Google without issue, so it would have to be something stunningly different than her previous papers.
I'm not fully informed about what she wanted, but I don't think it was particularly unreasonable, the controversy more has to do with headlines and spotlights about this person and her conduct/history in the past.
Criticizing one's employer was possible within very defined circumstances, and always privately, but doing so in a public square was a sure route to termination. This seemed reasonable at the time.
Has the relationship of employee/employer changed so?
Researchers are used to having more freedom. IBM didn't pretend to be egalitarian and fun. And part of the controversy is Google firing her but characterizing it as accepting her resignation.
- you have an employee who (in her letter) hired attorneys to sue you. - you have an employee who was given an award for impact who openly mocks it and thinks it’s a joke, which detracts from culture - you have an employee who went on a Twitter tirade and send several pages of information teaching a leading person in your space how to apologize. Humiliating. - you have a paper of questionable quality that maybe needs more review and maybe doesn’t but at a minimum appears to be BERT related which could impact your business. It’s not unreasonable to want to look at it. - you have an employee whose tone is very aggressive. It seems reasonable you’d be apprehensive of working with her. It doesn’t seem googles point is to make a model that is biased towards any group, etc, so you should be on the same team? - you get a volley email which requires information (give me all the conversations you had) which could be used to publicly shame the head of a business unit (Jeff Dean) - the employee emails, says she’s going on vacation, and then expects you to be waiting for her when you get back
I don’t know. It’s probably mixed. But it seems she’s well respected and given significant leeway. If we strip race and sex from this (if you can do that, perhaps it ruins it, but this is why this is a throwaway): this seems like a very high drama employee. Let’s say there was no hidden bogey man and we take Jeff at his word: more time was needed to give comments on the paper. From there, a) either paper is fine and she submits. Obviously the initial publication space for which she missed the internal review deadline is out, or B) if the paper is not fine and Google makes reasonable points to improve the quality, you incorporate them, then it’s published, or C) if the paper is not fine, but Google is making bogus claims to surpress you because of (in this thought experiment) some racial bias or business model conflict, then you can escalate internally or leave or fight in the media.
The problem is you didn’t let a)/b)/c) play out. By jamming the list serve as a way to influence and over ride the internal system the employee is trying to circumvent the operation practice of the group and it undermines leadership. Note, I’m not claiming she is incorrect in her work being suppressed potentially. I’m just claiming if I was her manager my inner voice mental response would be: “shut the fuck up and stop being such a shit disturber. I was willing to work this out with you, even though you could have followed the normal process everyone else does, but now you created a company internal hurfuffle so I need to loop in MY manager (Jeff) and the last time you interacted with public exposure with someone of that high stature (the Twitter incident) you caused a highly abraisive outcome. Ugh. What do I do? Oh you resigned unless you publicly humiliate my boss? Yeah okay, let’s just accept that, I’m not being paid enough to baby sit you. I’m at Google brain to do ai research, not manage temper tantrums. Ok, so will talk to Jeff, tell the team you resigned to help you save face (which she spun btw, lol) and oh yeah we should probably shut off your devices to avoid more damage. This sucks. Ugh I wish that list serve and email to Jeff didn’t go out. “
That’s a jaded view. You can be potentially (or not) repressed and simultaneously an asshole. Would any of you want to work with someone like her? Who low key threatens your job or attempts to publicly humiliate you if she doesn’t get what she wants?
Note, using a throw away because it’s too scary to post this publicly. Because I can’t judge what the “public consensus” is (and it seems to be changing), I can’t tell if the above is a 1/10 and like lunatic level, or 4/10 and highly offensive, or maybe a 6/10 and just highly insidious (by design). The bigger point is this incident shows how Twitter flames and self censor ship can perhaps make certain topics hard to discuss.
Sorry if this was offensive. But the “HR policy drone” view seems otherwise underrepresented.
Turns out, Google would feed each and every one of their employees, feet first, into a giant wood-chipper if that was the most cost-effective, legal way of disposing of them.
Turns out, the tech industry isn't really all that structurally different than any other industry.
Ignore that pipe full of blood and guts y'all, none of those people were marginalized. Back to work!
On one hand, I'm glad that people are sensitive enough to at least pretend to care about this, but on the other hand, they're completely deluded about the nature of work, and the nature of their relationship to their employer. It must come as a shock when, after spending a decade in school and moving directly from a university dorm/cafeteria to a corporate one, all of a sudden they can just ask you to leave. Maybe Google should start charging tuition.
She was working on a known-to-all controversial research project, which was asked to be retracted via an anonymous feedback doc delivered via HR! That is a very screwed up thing to do. All reactions from her from that to me are totally justified.
Seriously, google staff are walking out / protesting / rallying behind stuff every week. How can you afford to pay 120,000 folks who seem to be focused on being outraged, striking and walking out all the time.
I participated in a walkout ONCE when I was much younger, over an issue that was pretty clean cut (this was before me too, things had to go WAY WAY over the line - we'd call it a crime now - to get a walkout going). It was still a bit cheesy, but ...
Do I have this right - a google employee doing public attacks against other google staff / how to apologize etc? And googlers want this kind of culture, where having a discussion get's you social media blasted? It just seems these cultures are getting very unhealthy. Why do this all working for google?
Google must just have boatloads of cash to be able to afford this all. Do customers for GCP ever worry that some staff will protest what they are doing and take some kind of direct action?
For many, the way these employee activists operate is in fact very unhealthy and distracting. But with leaders often letting these disruptive activities and ignoring the fact that these activists aren’t actually doing work on paid company time, it feels to the majority that they have no psychological safety to speak up. And so they remain silent while a vocal aggressive minority takes over.
"I will/would resign if" reserves her the right to resign; it does not have the same meaning as "I am resigning if"
Compare to e.g.: https://thehill.com/policy/healthcare/511062-trump-vaccine-o...
We can't take this ultimatum as a resignation.
If they succeed you should assume Google is dead, sell your stocks and backup your data.
Jeff not only has a monumental reputation and wealth, he's also under Brin's and Page's protection who own more than 50% of Google.
Hi everyone,
I’m sure many of you have seen that Timnit Gebru is no longer working at Google. This is a difficult moment, especially given the important research topics she was involved in, and how deeply we care about responsible AI research as an org and as a company.
Because there’s been a lot of speculation and misunderstanding on social media, I wanted to share more context about how this came to pass, and assure you we’re here to support you as you continue the research you’re all engaged in.
Timnit co-authored a paper with four fellow Googlers as well as some external collaborators that needed to go through our review process (as is the case with all externally submitted papers). We’ve approved dozens of papers that Timnit and/or the other Googlers have authored and then published, but as you know, papers often require changes during the internal review process (or are even deemed unsuitable for submission). Unfortunately, this particular paper was only shared with a day’s notice before its deadline — we require two weeks for this sort of review — and then instead of awaiting reviewer feedback, it was approved for submission and submitted.
A cross functional team then reviewed the paper as part of our regular process and the authors were informed that it didn’t meet our bar for publication and were given feedback about why. It ignored too much relevant research — for example, it talked about the environmental impact of large models, but disregarded subsequent research showing much greater efficiencies. Similarly, it raised concerns about bias in language models, but didn’t take into account recent research to mitigate these issues. We acknowledge that the authors were extremely disappointed with the decision that Megan and I ultimately made, especially as they’d already submitted the paper.
Timnit responded with an email requiring that a number of conditions be met in order for her to continue working at Google, including revealing the identities of every person who Megan and I had spoken to and consulted as part of the review of the paper and the exact feedback. Timnit wrote that if we didn’t meet these demands, she would leave Google and work on an end date. We accept and respect her decision to resign from Google.
Given Timnit's role as a respected researcher and a manager in our Ethical AI team, I feel badly that Timnit has gotten to a place where she feels this way about the work we’re doing. 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. I understand the frustration about the pace of progress, but we have important work ahead and we need to keep at it.
I know we all genuinely share Timnit’s passion to make AI more equitable and inclusive. No doubt, wherever she goes after Google, she’ll do great work and I look forward to reading her papers and seeing what she accomplishes.
Thank you for reading and for all the important work you continue to do.
-Jeff
However Google's passive-aggressive response wasn't terribly great either ("We respect your decision to leave Google ... and we are accepting your resignation.")
It's a firing packaged up as fake "respect".
I wish both sides learned more about how to de-escalate situations.
i'm not involved with google in any way except being their product and user of their services... but neither side inspires my confidence here.
"We feel that at ultimatum is not an effective way to discuss this issue and so we are hereby terminating your employment effective today."
No need for this passive-aggressive "respect your decision" [but didn't actually respectfully let you confirm your decision] bullshit.
Just checking to make sure my lexicon's up to date.
Let's focus the discussion around if that was a good or bad decision, to not comply with her conditions. She fired herself, there is no discussion around that.
Google needs to decide what it wants to be. Given its choices, I think it's clear what they want. They just need to say it.
It'll help employees and potential employees know if they want to work there.
I don't want to work anywhere that pushes any ideology so far that they break the law. That goes for if it's in pushing their bottom line as well, so regulatory gray-area companies like Uber & AirBnB are also a no-go for me.
Tech firms and startups are just businesses. Breaking laws/regulations is not a requirement to operate a new business.
Also running an ethical business goes far beyond adhering to rules and regulations, as many of them don’t currently penalize unethical practices.
yes, plenty of reasons, you're part of a team, you got to work with the team or you're just using them
What they should strive for is a diversity of backgrounds, viewpoints, experiences and have that process account for the differences in how those people arrived at where they currently are.
As far as I understand, all companies need to have a pretty solid overlap of "common opinions" or else you are going to have political deadlock.
If you have examples to the contrary I'm really curious to learn about how they worked.
Diversity in terms of measurable metrics, such as women/minorities/LGBT++ being adequately represented, is simply a benchmark that can be paraded around for appeasement purposes.
What the hell. How can you say that with a straight face?
Of the two best engineers I know, one is is black and the other is trans. They're both 10x engineers.
I didn't.
> They're both 10x engineers.
https://medium.com/@yuval.idan5/the-10x-engineer-myth-is-wro...
https://sparktoro.com/blog/belief-in-the-mythical-10x-hire-h...
Poe's law likely applies here.
Run as fast as you can. And take your data with you.
At the advent of anything, it won't matter. They'll just be some place.
People want to work there still? Separate from just wanting to work (who wants to be unemployed)? Why? It was stupid to be that way as far back as 2003, as they had already IPO'd by then. Stupid like an idiot trying to buy once/since a stock started to surge upward. Stupid. Anything after is in a class by itself.
I wonder how they are viewed among young people now. Both college students and those with 2 or so years of experience. Hmmm..
Maybe it's still the same.