Intelligence wouldn’t have fired her this way, artificial or not.
2 Likes
It seems to me that the duty of peer review in this case should have belonged to the conference. Just as we don’t want drug companies to be able to suppress papers that say that their drugs don’t work for a particular application, we don’t want tech companies to be able to suppress information about the ethics of their technology. Google in particular is in a unique position to do so, whether the papers originate from Google or not.
2 Likes
Indeed. These things are clearly ways to get fired, and I would think that such a smart person would have known that, even subconsciously.
As some point, Google might have been a place where one could question management on internal emails lists. But, that culture seems to have gone away. The paper thing is hard to parse. She seems to be saying that Google sat on the review until the last minute (or later) before rejecting the paper. And, that the internal review was insubstantial. Google says differently. I agree that outing the identities of the internal reviewers is an unusual demand; many academic papers are reviewed anonymously. But, the content of the review should certainly be detailed and shared. I would hope there would be a rebuttal period too, although if the paper was already accepted then there probably wasn’t room to incorporate substantial revisions.
I agree that it seems like she might have been looking for a way out of there.
edit: typo
2 Likes
I’m not sure biotech is the right comparison since there’s a vastly different problem and regulatory apparatus.
But either way, that’s not unique to Google.
When you work for a company and publish you’re speaking on behalf of that company, which means the company gets to have a say over what you say.
I’d be shocked if any big company didn’t have some form of review/approval for internal papers.
Imagine the Google researcher in question wrote a really offensive paper instead. Do you think Google wants that going out under their name?
Which is 2 weeks before the conference/journal deadline.
If you were having trouble with the deadline I’m guessing you could send a draft. Either way, this was Google’s response except they also didn’t approve the paper.
The ‘platformer’ link has Google’s response, but basically the internal reviewers said that the paper was ignoring current research that didn’t support its results.
For instance, the paper said “X is bad because of Y” but it ignored advancements made to X specifically to address Y.
Again, we don’t know if those internal reviewers were making valid criticisms or not but that was their reason.
3 Likes
This wasn’t peer review, it was management/legal/PR declining to sign off on a paper being submitted for peer review because they didn’t like what it said.
3 Likes
Although peer review is perhaps a misnomer if the people reviewing your paper are more senior managers within your employer’s organisation. They’re not exactly your peers are they?
We don’t have much detail. Google apparently says:
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.
Sure and there are good reasons to doubt that it’s a good practice. While peer review itself is highly helpful, there are still huge issues:
- Reviewing a paper is not paid work. It takes several hours to do well and you may be asked to review several papers simultaneously, but it’s counted as volunteer work. It’s time you can’t spend developing your own grant proposals, conducting experiments, writing your own publications, or mentoring new students.
- Reviewers can demand expensive new studies. While sometimes these are reasonable, they can balloon into ridiculous demands very quickly. The current record is for revision experiments costing an additional $400k in compute.
- Anonymity invites assholery, same way pseudonyms do online. A harsh, nasty review is usually unhelpful, but they’re all too common. Reviews that provide constructive feedback and suggestions are much more helpful, but also take more time to produce.
- Reviewers can scoop your work. If they’re your competitors in a field, seeing your raw work can speed their work substantially. This is an asymmetrical risk: you don’t know who they are but they know who you are. In theory, they face the same risks, but if they’re a glam lab with more resources, they definitely don’t face the same costs.
- Reviews are not published. This can gloss over issues in the final papers, lead to repeated work, suppression of negative results, etc.
So maybe peer review shouldn’t be unpaid or anonymous.
5 Likes
I guess that was meant for others… not Google.
Just be corporate!
That’s Dr Gebru to you.
I get this at work regularly. Requests for publication, procurements, contracts, approvals, etc., submitted the day before (or same day) they need to be published, work start, etc. In most cases, it is submitted as a smug fait accompli.
And in most cases, there is a significant issue with the document in question: contract terms are very unfavorable; costs unrealistic, unallowable, or unintelligible; significant conflicts of interest exist; funders, partners, and the company not appropriately cited and credited; information incorrect; negative information “minimized;” etc. Which is why, unsurprisingly, they are submitted at the last second with the hopes that the “urgent” nature of the activity will result in a rubber stamp and no serious review.
It never works. Submitting at the last minute is amateur hour and a flashing red flag that something is suspiciously wrong. Employees get disciplined up to termination, and all documents receive appropriate review, comments, and instruction on how to bring documents into compliance if needed.
I don’t know the full details of this incident, and have no particular faith in Google’s version of events, but if Dr. Gebru did submit a paper for review the day before it was to be published, or submitted a paper for publication without the required review, then she earned whatever disciplinary action Google’s policies reserve for such a policy breach.
(Can you tell I feel strongly about this?)
5 Likes
Lots of things happened this year somehow; nobody had 2 weeks of slack for anything and I heard a lot of people being fine with that (given a hit on irrelevancy or tone deafnessotherwise.)
Where is my PoC for vulns in speculative white supremacy execution, tho.
5 Likes
I agree with your points. The academic community I’m part of is currently going through a reconsideration of our reviewing methods for many of the reasons you’ve laid out. We’d need a whole other thread to get into it.
I was trying to say that Google didn’t seem abnormal in this respect. Although, even at a big company like Google it’s hard to imagine the reviewers are really anonymous. And, it seems like reviewers are senior managers with their own agendas and power beyond mere paper approval. That sounds problematic. It seems like there’s more to this story. If I knew somebody at my org had the expertise to say “Why didn’t you cite X, Y and Z?” then I’d show them my paper on my own. And, if such comments were accurate, I’d be glad to receive them.
This seems like the last act of a play. Even though Google has framed this in a way that is “by the book,” I do not want to normalize it or say that this doesn’t represent a massive failure on the part of management. They hired a talented young researcher and she was doing excellent work. At one point she was working there with the idea that she could get important things done, but eventually became disillusioned. The email she sent is scathing; no reasonable manager should have let the situation get to that point. The style of her dismissal is clearly a punishment and a warning to others, which is further indication of inhumane, authoritarian management.
1 Like
Man. Grow up.
Yes. It MUST be our “inferiority”… /s
15 Likes
7 Likes
9 Likes
It’s a shame that Googles employees are too dim to see that they place the company at risk of litigation when they run their business in a discriminatory fashion.
Or that alienating over half of your potential customer base is a stupid business decision.
That’s just gross incompetence which would get an employee fired in a company run under a meritocracy.
8 Likes
Google was never “woke” nor leftist. While, at one time, it fostered a playful atmosphere, it was very much a lolbrotarian company that has since shifted hard right for one motive: cold, hard cash.
Google’s first 100 employees were all people that Brin and Page knew personally, or who a Google staffer knew and recommended, with two notable exceptions: Marissa Meyer, and Eric Schmidt. One would think that a “woke leftist” company would very much have tried to hire a diverse team well-fitted to finding global solutions, but Google was not that company.
By 2007 lolbrotarian Google was dead, replaced with a hard-right Google; think the purchase of Doubleclick, the YouTube acquisition, the acquisition of Andy Rubin’s Android and repeated crimes.
Even today Google has issues retaining women, and it isn’t because women are more demanding or conscentious – it’s because Google’s leadership are dingdongs bent on covering up harassment and chasing pennies rather than cultivating a team.
Ubisoft, a similar company in size, hires women at a bigger rate than Google… But also tolerates serial abusers and sex pests in management, and has a worse retention rate for women who work than Google.
The problem is emphatically not women.
The problem is dumbass brogrammers who run cover for, promote, and profit from serial harassers, abusers, and bootlickers.
14 Likes
uh yeah dude, whatever 
10 Likes
Is that you, James Damore?
3 Likes
