“Quite a few activists and researchers were fired for asking Google to be a bit more ethical”
In the examples the author cited, those activists and researchers were fired for reasons going beyond “asking Google to be a bit more ethical. And frankly it doesn’t seem ethical for activists, researchers, and the author to deliberately misconstrue facts whenever they believe they’re on the right side of an ethical (and sometimes just political) issue.
Exactly, perhaps people read too many comic book with these one dimensional villains to think that an employee asked Google to be more ethical and then they fired them right there and you could hear an evil laugh in the background.
I've looked at some of the correspondence between Timnit and others and it's fairly unprofessional on her part if you ask me and it has this self righteousness tone to it too.
Another very funny thing people believe is that Google wanted to do "evil" things but their perfectly crafted plan fell apart at the last second when they realized that they have "do no evil" in their code of conduct. After much scrambling the order came from the top to quickly remove it to unblock the evil plans.
That said, there sure are jerks at Google like any other company but I've realized most these dramatic stories regurgitated by the media is BS
She was publishing studies critical of Google, Google wanted at least the right to have studies of the company reviewed and to include more updated information etc. which is extremely gracious on the part of Google.
She gave the company an ultimatum i.e. 'Change the Terms or I Quit' so they said 'We're not changing the terms, we accept your resignation'.
It's galling how many people (and their supporters) think that they have a right to be employed by a company and then publicly criticize them with misleading information, unmitigated or reviewed by the company.
That’s true for academics getting grants (being able to publicly critizise the one that gave them). Why would it not be true for academics being directly bankrolled by a company?
No - the foundation of your question is a bit misaligned. The major difference is that as much as corporate research likes to pretend they’re academics, they’re not. In corporate research you’re not after some unalienable absolute truth that benefits society - you’re doing research that can benefit the company that is paying your salary. Sometimes those two align and that’s great. But anyone that thinks that an experienced researcher at a corporation will have the seem freedom as a tenured professor is missing something.
In exchange for this tradeoff you get a lot better salary and don’t have to feel like all of your time is spent chasing grants.
> you’re doing research that can benefit the company that is paying your salary
So you are doing research within a slightly more defined scope. If you still call it research then the results are completely separate from your affiliation.
I'm not privy to the full details so idk but I provided my perspective based on what I'd witnessed which makes me believe she was fired/forced resigned because of how she approached issues at work (giving ultimatums / looking for identities of her reviewers, harassing LeCun on Twitter, etc)
> Another very funny thing people believe is that Google wanted to do "evil" things but their perfectly crafted plan fell apart at the last second when they realized that they have "do no evil" in their code of conduct.
Of course the CoC is not binding, but having that in there was a statement, which meant, roughly interpreted, "the impact of our work should not be net negative". The removal of the line itself did not change anything, of course, but it was a clear statement that this priority had been dropped. I think it's quite obvious why this is seen as not great.
Second, if you were wearing "I'm a good guy" badge on your shirt and later on decided to go rogue and work for the enemy, would you then remove that badge right away? Of course not unless you're truly incompetent at being bad.
I'd understand it if they got rid of the wording because of how much headache it creates for them.
Btw, that's basically why government and business officials often resort to saying the least informative statements. Because every word and sound bite will be used against them.
Um - she wasn’t fired because of her paper - she required they out an anonymous reviewer or she wouldn’t be able to work there.
It’s pretty likely the reviewer would have been harassed - a good thing perhaps if you have one view of thing - but a bad thing from another perspective.
Google refused and accepted that she writing be able to continue working there
I think it's worth reflecting on why we're so quick to believe the corporate version of events. Corporations have an astronomical amount of power and influence. Aren't you at least a little curious whether it's true?
And if there was nothing wrong with her paper, that raises the question of why an anonymous reviewer was grilling her about it.
Her paper got got flagged during internal review. There wasn’t time to revise it because she submitted it late. She then wrote an angry rant to her coworkers that claimed she was oppressed and told them to stop working. And then she demanded the name of the anonymous reviewer and threatened to quit if the company did not meet her demand.
Your defense is that most papers were submitted late without a problem. Perhaps most papers don’t need revisions, but this one did, and there’s no excuse for how she handled it.
I’ve had papers rejected from my company’s internal review process. It sucks, but it’s not the end of the world. You make some revisions and submit to the next conference. You don’t go ballistic and tell your coworkers their work is meaningless and threaten to quit unless you get your way.
(The subthread below contains the refutations, so I won't copy them here. I left a reply at https://news.ycombinator.com/item?id=28966286, but you might want to climb it upwards to see why this is mistaken. Suffice to say, her paper passed the internal review, and was subjected to a second review that wasn't policy at the time.)
If it's in violation of your contract they can reject your resignation, but most people are At Will employees which means either side can terminate employment.
This is generally false. You can’t be forced to work and a specific performance judgement is unlikely. If you resign and had a contract you can’t work elsewhere, may need to pay damages. But slavery, indentures etc mostly illegal these days
> I think it's worth reflecting on why we're so quick to believe the corporate version of events
I reflected on it and found that it mainly happens when the person affect have their entire work life on basis of their politics as opposed other victims of corporate malfeasance who get plenty of support/ empathy from general community.
Timnit is unethical as hell. She's also the worst kind of being unethical: being unethical in the name of doing good. She second guessed the motives of her coworkers. She attributed every difference to her paper or claim to racism or white supremacy or oppression or any combination out of the three. For all I know, only a true racist would view everything through a racial lens.
I had some sympathies with her when the initial story broke out. But the continous flaming on Twitter, harping and retweeting her supportive followers and the performative activism for a select set of racial topic makes me feel she cares only about her own set of interests.
> This is basically mainstream teaching now (see critical race theory). It's like witch burning…
>> She attributed every difference to her paper or claim to racism or white supremacy or oppression or any combination out of the three.
Critical race theory is not yelling to claim every problem is due to racism or oppression. While I don’t agree with all tenants of the recent evolution of critical race theory, I do with the historical purpose of it.
> A collection of critical stances against the existing legal order from a race-based point of view — Roy L. Brooks
Fast forward to today and you have a more broad and insistent approach developing with Critical Race Theory. Primarily due to all of the push back on having legitimate societal conversations and decisions about systemic discrimination since the civil right movement in the US (even going back to the 1968 Kerner Commission).
> The secular high priests of this ideology do and are.
While the focus now on structural determinism, the critiques of liberalism or the use of narratives and lived experiences of CRT might upset you. CRT is not at all simply what you described.
> Critical race theory is not yelling to claim every problem is due to racism or oppression.
It largely is these days.
> While I don’t agree with all tenants of the recent evolution of critical race theory, I do with the historical purpose of it.
I don't. It's divisive hateful racist pseudoscientific nonsense that relies on bullying and threats to shut down its critics.
> Fast forward to today and you have a more broad and insistent approach developing with Critical Race Theory. Primarily due to all of the push back on having legitimate societal conversations and decisions about systemic discrimination since the civil right movement in the US (even going back to the 1968 Kerner Commission).
It's always funny when they talk about systemic discrimination while also being the ones trying to institute segregation, keep too many "undesirable" races out of colleges, and generally agitate for institutional racism.
> While the focus now on structural determinism, the critiques of liberalism or the use of narratives and lived experiences of CRT might upset you.
My conscience is quite clear, I have never lusted for policies of racism. I have never complained about too many Asians in colleges, for example. I am not a racist. I stand with greats like MLK Jr on the right side of history, and opposite them. I'm not offended by their attempts to invert definitions to absolve themselves and bully me, I'm only offended by their racism.
> CRT is not at all simply what you described.
I disagree, it is very much like what I described.
From what I read at the time, she wasn’t fired for the content of the paper (though she initially made it sound like this before more details became public). She gave the company an ultimatum to reveal the name of an anonymous reviewer or she would quit, and they took her threat seriously.
She submitted her paper late for internal review when it was too late for revisions. It might be the case that most papers are submitted late for internal review, but most papers don’t expressly criticize the employer’s work. That’s an important distinction, and it was naive to not anticipate that the paper might get extra scrutiny.
Regardless, being upset because you didn’t get your way doesn’t give you carte blanche to do unprofessional things that will get you fired. You can’t say “Give me X or I quit” and not assume there’s a chance your bluff will be called (especially when “X” is a demand for the company to violate its own policies).
You're factually wrong. I was literally able to view her paper, and it's approved status, in the internal tool where papers are approved for publication.
This directly contradicts what Jeff Dean wrote: “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” (emphasis mine).
Which makes it more likely to be the truth, not less likely. It would be very foolish to write that if it were demonstrably false, as was claimed above. And I have to assume their lawyers aren’t very foolish.
But this is all another deflection. Whether the paper was rejected by a regular process or an irregular process, it doesn’t condone her actions that followed.
> Whether the paper was rejected by a regular process or an irregular process, it doesn’t condone her actions that followed
This is where we'll have to agree to disagree.
As you can see in this thread, she was treated unfairly. Someone confirmed to you that this is the case, and you're still arguing that her subsequent behavior was unjustifiable.
I think her subsequent behavior was a result of the unfair treatment. I'm not sure how I would've reacted in her place, and I have a lot of empathy for a scientist who is being scrutinized outside of normal policies.
Her paper was anodyne. No one would've blinked if it was published. And I think you keep saying that it's okay for a research department to stop publication of scientific work regardless of whether the paper was mistaken or not.
You've said that it needs revisions. I work in ML. It seemed not to need any revisions, as far as I could tell. Google's reaction seemed far out of proportion for what the paper claimed.
You can say "Well, that's life. Google didn't do anything wrong." The former is true, but I'm skeptical of the latter.
"As you can see in this thread, she was treated unfairly. S"
This is not true.
The fact that she may or may not have had extra scrutiny than others, within regular policy or not is completely besides the point.
If a company, for whatever reason, wants to have 'another review' of the paper, it's completely within their purview.
"Her paper was anodyne. No one would've blinked if it was published."
Our view of whether it was notable or not is irrelevant. If the company wanted to put the brakes on it, especially because it related to Google, then that's their choice.
It's hugely arrogant for her to make the claims she did - let alone publicly (!) - by all right she should have been fired for that, she was luckily the conditions enabled her to even try to establish the premise of her own resignation.
A company being legally allowed to do something doesn't implicitly make that thing fair or ethical.
You can simultaneously believe that Google has the right to quash any paper, and believe that doing so arbitrarily is unfair. So yes, she was treated unfairly. And that's absolutely relevant to people's decision making. Like, what do you think it's irrelevant in relation to?
Its not clear what you mean by publicly, nothing was done outside of Google until after she was fired.
After the, (my words), harrasment of Lacun on Twitter, the sort of belittling of the internal award she was given (you'd have to really think they were evil to give her an award just to gaslight her), plus the short notice ultimatum to a very senior person (one level down from Jeff Dean right?)....no offense but from the outside she seems exhausting to manage.
The abstract came out, and honestly the paper was so so. Obviously it went against googles business model, but it'd be a wayyyy bigger scandal if they burried it.
Maybe Jeff Dean is a liar, but his comment was the paper missed incorporating significant recent research. As head of the group, and just out of respect towards the institution and to science, pondering it for more than 48 hours before bouncing on vacation or heaven forbid presenting at the next conference would have been reasonable.
Other people circumventing policy doesn't mean that it is fine for her to do it. It is really hard to judge as you do without more data on the circumstances of the other people circumventing the policy and not getting punished, right now we have to take their word that what she did wasn't worse than what most others did. We can't know that. Similarly one engineer force pushing code to fix an issue might be fine, but another engineer doing the same thing might not be fine and the second gets investigated, is that unfair? We can't tell, there isn't enough information here.
Because plenty of other researchers working at Google regularly publish papers critical of Google and are just fine? Even Timnit had already done that many times, she just got into trouble this time but all the other papers she published was fine. If Google really cared about these researchers publishing critical pieces then Google would have shut down that entire group rather than just firing her.
Right but the claim is that Google instituted a new policy, and applied it inconsistently without telling people, until it blew up in their faces when she was fired. Only after that did they formalize the policy, and saying that she was fired for circumventing and then questioning a policy that ostensibly didn't exist when she was fired doesn't have the same ring to it.
But you claim she was fired for being critical of Google. Maybe what you are saying here is right and she was just the first person fired since nobody followed the new guidelines and at some point they will face consequences, it could just as well have been anyone else. But I don't see how that would make Google look bad, it is normal to fire one of the offenders as an example when a group refuses to follow new guidelines, either you do that or you fire everyone.
And hers case was especially easy, since she said she will leave the company if the policy isn't changed. Google didn't want to change the policy. What Google might have been wrong about here is if they reported this as her voluntarily leaving, even if you could argue it is Google firing her. But nobody really cares about that that detail, it isn't important and the story wouldn't blow up if that was the only thing they claimed Google did wrong.
> And hers case was especially easy, since she said she will leave the company if the policy isn't changed. Google didn't want to change the policy.
The policy seemed to be made up specifically for her paper. Her paper already passed the standard internal review, and she alerted them months before the conference that her paper was going to be submitted.
I have a feeling her continuous rants on Twitter is because her overall demands and feeling of entitlement were stood down. She demanded things which would have been just another kind of shit show and targeted harassment.
A social-justice themed charade is being made out of just being salty at getting fired.
I have mentioned this long time ago and would do so again: No matter how righteous be the cause, threatening your employers and instigating a mutinous conduct with fellow employees doesn't go down too well.
Both sides could have done better. Both sides includes Timnit's subsequent conduct as well.
I interpreted "with fellow employees" as in cooperation with fellow employees, so that wouldn't be a one-person strike. Although I guess it could be interpreted as against fellow employees.
In this case, I think the initial ultimatum delivered to management was in the vein of “I will quit if you don’t do X” (not “we will quit”), but she was also trying to collaborate/mobilize with others in the company to pressure management. So a bit of both?
I have read the paper. There were several things wrong with it. The most glaring example was comparing the power consumption of training a machine learning NLP model with a gas guzzling jet.
Not once in the paper does she mention that Google has been carbon neutral since 2008, because that would trash half her paper. The paper seemed of desperation to make non existent arguments.
I wouldn't be surprised if the people who worked on BERT generated more CO2 commuting to work than they did training the model. It seems like a drop in the bucket.
"We first consider environmental risks. Echoing a line of recent
work outlining the environmental and financial costs of deep learn-
ing systems [129], we encourage the research community to priori-
tize these impacts. One way this can be done is by reporting costs
and evaluating works based on the amount of resources they con-
sume [57]. As we outline in §3, increasing the environmental and
financial costs of these models doubly punishes marginalized com-
munities that are least likely to benefit from the progress achieved
by large LMs and most likely to be harmed by negative environ-
mental consequences of its resource consumption. At the scale we
are discussing (outlined in §2), the first consideration should be the
environmental cost."
This whole paragraph is such a reach, and so absurd that it seems like a parody. Are we supposed to believe that marginalized communities are being harmed by Google training NLP models using green energy? Or will they benefit from better translation from English language content into marginalized languages?
Timnit has zero interest in what actually benefits marginalized communities. She uses them as a prop to build a career in high paying AI field where, apparently, the sub field of ai fairness has a low bar for publication.
Yes, I never said that the power consumption calculations were wrong. The paper wasn't about power consumption, but CO2 emissions leading to global warming. She compared the power consumption of model training to jet fuel exhaust and used jet fuel exhaust to tie the power consumption to CO2 emissions. Of course, she didn't mention that Google is carbon neutral and its data centers are primarily powered by green energy. I have not seen a more dishonest paper in CS.
Electricity from the grid can come from green sources, and in the particular case if Google, it actually does. Comparing electricity consumption for a one time training with jet fuel usage is absurd.
Moreover, the paper doesn't mention any possible benefit of building such nlp models. Increased accessibility of web content across languages and for disabled people.
Google, simply through its search, maps and navigation saves a lot of wasted carbon. Complaining about the power consumption of ML models is absolutely ridiculous.
Her paper reeked of desperation to find something, anything within the AI space that could be posted as an ethical dilemma.
Reading that paper convinced me that Timnit could bring absolutely no value to AI as a field, in terms of ethics or otherwise.
"Carbon neutral" via buying offsets and renewables made electricity is better than nothing but also a half truth. Their energy demand is displacing other consumers of that electricity, and for example, who will then consume eg fossil made electricity instead.
That's not how economics of scale work. This is zero sum thinking.
Google buying green electricity makes green electricity cheaper, not more expensive. More producers are incentivized to produce green electricity as Google's demand increases continuously. Other companies like Microsoft and Facebook go green as well, and the economies of scale kick in. Prices start collapsing. This is actually happening right now.
If I go solar today with my home, it is not a negative action that displaces someone else. It actually expands the solar industry and enables cheaper solar in the future.
Without early customers, green energy would never have taken off.
> Google buying green electricity makes green electricity cheaper, not more expensive.
The marginal cost of puchasing more energy might go down (or it might go up if you are pushing up the demand). But not the total expenditure on energy.
> Prices start collapsing. This is actually happening right now.
Thermal coal prices are at all time record highs right now. It's 4x the price it was a few years ago when we were all hearing about how solar has become cheaper than coal.
It's not that solar generation suddenly became far more expensive for some supernatural reason. It's that demand for electricity outweighs the installed capacity of renewables.
Coal prices are high because there is a lot of electricity demand not being met by existing renewables. Because everybody bought the green electricity. Which is not only making the green electricity more expensive, it's causing more coal to be burnt to keep up.
Like I said before, that's not how economies of scale work in the long run. Following your recommendations will reduce green energy prices in the short run and promote CO2 emissions in the long run.
The negative externalities of CO2 are long term and not short term. So your complaints are counter productive.
The fact that there is a learning effect doesn't mean it's good to consume more electricity. If you want to argue positive sum, you would need to show a the consumed green electricity displaces more than the equivalent amount of fossil electricity plus co2 fp of measures used to handle variability of renewables vs nearly flat demand from data centers. Which seems hard.
I mean we might as well shut down electricity to all of our homes if we want to make this argument. And we could argue with each other over snail mail.
But I suspect that will actually be worse from CO2 emissions perspective.
That's strawman, some electricity use is needed because we depend on it for survival and the rest is various levels of nice to have.
But its use has environmental costs and we are in big trouble with climate change. As long as there is fossil made electricity flowing in the grid, marginal changes in demand can be viewed as contributing to its production, because electricity is fungible (modulo transmission considerations).
That's a fat big POV that very few will take seriously. Being carbon neutral absolutely incentivizes green energy. Making up a wall of insurmountable demands that very few people in this world, except hunter gatherers, can possibly meet does not help but hurts efforts towards CO2 reduction.
Carbon neutrality works by subsidizing the costs of reducing carbon emissions at source. Trying to shit on that makes you regressive, not progressive.
Let's not get too emotional about things and keep to the HN guidelines.
I do understand and somewhat agree with the "perfect is the enemy of good" thinking, but it's still good to be aware and keep outselves honest about the difference between them.
The marginal CO2 perspective is in fact an establilished thing in the industry, see eg [1] [2].
Regarding "insurmountable demands", not sure what you're referring to here, the discussion was about whether there are significant indirect co2 & environmental costs to "carbon neutral" electricity as defined by Google, which I argued there are.
You can use the same argument to argue that EVs are a scam, because some electricity is generated by coal.
The point is that electricity can have a green source, liquid carbon fuel can't. You absolutely cannot compare the two. FWIW, Google's data centers are primarily powered by green energy.
> Google's data centers are primarily powered by green energy.
But not since they declared themselves carbon neutral (which they originally accomplished with certificates). Note the original claim was 'carbon neutral since 2008'. Google also uses a large amount of resources besides electrical energy, so claiming to be neutral is a scam unless they actually sequester co2 permanently.
> You can use the same argument to argue that EVs are a scam, because some electricity is generated by coal.
It's not the same argument. But I do argue that EVs are not carbon neutral.
> Google also uses a large amount of resources besides electrical energy, so claiming to be neutral is a scam unless they actually sequester co2 permanently.
I am going to have to ask for a citation for this. They claim to be carbon neutral, not that they don't use any resources at all.
>> > Google's data centers are primarily powered by green energy.
> But not since they declared themselves carbon neutral
Nobody has made this claim
> Note the original claim was 'carbon neutral since 2008'.
If the now died down Timnit reporting did not convince you that Timnit didn't apply the right tactic, if we agree to set aside the legitimacy of her cause in the first place (which itself was and still is under intense debate), then there does not seem any good way to unify our views.
Dr. Gebru was fired for having approached a lawyer about workplace issues. (She told Jeff Dean and others that she was doing so in an email made public). That triggers Legal to retaliate and look for a reason to fire the employee, immediately. Jeff Dean let the conflict drag out 6+ months because he thought he had control of the situation (and he was wrong). Without Jeff Dean’s persistence, Dr. Gebru would probably have been retaliated upon as swiftly as those in the Alphabet Workers Union. That’s how Legal and HR work—- they retaliate if litigation looks imminent. (Levandowski lasted so long because he was dealing with not Jeff Dean but Larry Page).
Google handled Dr Gebru’s exit differently than that of Anthony Levandowski, Andy Rubin, David Drummond, etc etc because Jeff Dean made the call in timing and he was really that clueless about public optics, especially with Tensorflow’s huge following. Moreover, Dr Gebru didn’t have the bankroll to fund a legal team like any of the others listed.
A lot of the posts here suggest Google is struggling with scale. Dr Gebru’s case definitely would have been handled differently were Google a smaller company. Also if there wasn’t COVID—- keep in mind she was fired a couple days before Thanksgiving during COVID. The whole country was a mess then.
> Dr. Gebru was fired for having approached a lawyer about workplace issues.
That is just very dishonest. She gave an ultimatum. Execs accepted the ultimatum. You are gaslighting other people on this topic. At the very least, you should've acknowledged that this is up to an interpretation.
Regardless whether she was fired or quit, she was toxic as hell. She dragged Jeff Dean into an unrelated matter in public and implied Jeff Dean supported personal attack.
If a direct report did this to me, I would've wanted them to leave. In this case, Gebru wanted to leave (according to her ultimatum). Jeff Dean wanted her to leave (I don't doubt this. nobody wanted to work with a colleage who attacked them on twitter). I see a happy ending for both.
Gebru and her supporters have been dishonest about the whole thing.
The email about the lawyers is months before the “ultimatum.”
The reason Dr Gebru’s case is so divisive in tech (especially in this thread) is because so many are agnostic or ignorant of the role of ethics in tech; and almost every employee is ignorant of how Legal/HR execute retaliation. Anthony Levandowski is a literal criminal and yet is still enjoying tens of millions of dollars of Google’s money. Same with Andy Rubin. Even Michael Church is rich. If one percent of the energy spent fighting Dr Gebru was put into fighting the real scumbags in tech… you’d have something more like Google before it got greedy with scale.
Is there any reason to think Michael Church earned more than Dr. Gebru? Michael Church only worked at Google for 6 months, so I would guess none of his stock vested.
Can you explain why you think dragging a low level IC who worked at the company less than year over 10 years ago and afaik never really did anyone any harm into this is appropriate?
She said we was consulting a lawyer about discrimination in the workplace. That’s all you need to say to trigger Legal to retaliate. If you’re Anthony Levandowski bartering with Larry Page, though, Legal takes a back seat. It doesn’t even matter when Levandowski instigated a collision on the 101.
In the examples the author cited, those activists and researchers were fired for reasons going beyond “asking Google to be a bit more ethical. And frankly it doesn’t seem ethical for activists, researchers, and the author to deliberately misconstrue facts whenever they believe they’re on the right side of an ethical (and sometimes just political) issue.