Tensions in Google's ethical AI group increase as it sends demands to CEO • Hacker News Board

Re: Tensions in Google's ethical AI group increase as it sends demands to CEO

#221

Post by thu2111 » Thu, Dec 17, 2020, 4:30 PM UTC

panpanna wrote: post #168

hn_throwaway_99 wrote: post #139

Earlier quoted context omitted.

Again, I'm not really trying to defend Google one way or the other. At the end of the day it seemed like Gebru published a paper that was quite critical of how Google intends to make a sizable portion of their revenue, and anyone should understand that when you do that in a large company, it's going to get extra scrutiny. And it could very well be that the whole purpose of Google's "ethical AI" team in the first plac…

So basically Gebru was fired for turning Googles fake ethics group into a real ethics group?

Her paper didn't make any reasonable points about real ethics, it was mostly wafer thin pseudo-intellectual moralising of the form "but what about the trees!" without any attempt at rigorous cost/benefit analysis. At its core ethics is about balances and tradeoffs: something creates some problems for one person, but creates some benefits for another person, is it moral? Is it right?

As a consequence her paper made a lot of points that came across as absurd like railing against models due to the electricity cost of training them, even though those costs are small relative to completely ordinary things everyone takes for granted, like driving to work and back.

Re: Tensions in Google's ethical AI group increase as it sends demands to CEO

#222

Post by woah » Thu, Dec 17, 2020, 4:31 PM UTC

uh_uh wrote: post #128

woah wrote: post #125

Earlier quoted context omitted.

Google tried to prevent her from publishing a paper, which is outside of their normal practice, then dissembled about what happened (pretending internal reviewers are anonymous, for example)

Didn't they push back on the publication because the paper was lacking in certain regards? IIRC it failed to take into account relevant recent papers, for example. Which probably could have been fixed if they didn't submit as late as they did.

According to current and former Google employees on various machine learning forums, Google does not do an academic review on their employee’s papers. That is the job of the journal reviewers. Google usually just checks whether any proprietary IP has accidentally made it in. For them to claim anything different is comically dishonest.

Re: Tensions in Google's ethical AI group increase as it sends demands to CEO

#223

Post by googlethrow3134 » Thu, Dec 17, 2020, 4:31 PM UTC

motohagiography wrote: post #110

This issue reminds me a lot of what we do in security and privacy, which are ostensibly engineering disciplines, but are really technology governance functions. Ethical AI is also a governance function, and like hackers in security, there is an invisible yet stark limit to the value activists can provide in actual governance roles. Litigating demands is just not how governance gets done. The irony of Google hiring so…

[flagged]

Re: Tensions in Google's ethical AI group increase as it sends demands to CEO

#224

Post by hansvm » Thu, Dec 17, 2020, 4:31 PM UTC

iab wrote: post #131

tw04 wrote: post #95

Earlier quoted context omitted.

Because the entire reason for hiring someone like this is to find issues with what you're doing and stop them. The entire position is a pre-emptive move to stop government from stepping in to regulate the company. If you fire the person the second they talk about something they feel you're doing wrong, it kind of defeats the entire purpose (IMO). You literally hired this person to the e-brake on your AI program and t…

Exactly - your last paragraph is the crux for me. They could have been fired for no reason, and no notice under the most common at-will agreements. The system does not favor employees at all, and I wish we were all more angry at that.

Is the problem that you can be let go at any time or that if that happens at an inopportune moment you can lose everything? With better social safety nets we (in the USA) wouldn't have to tie our well-being to a Lovecraftian body of law entangling employment with every other aspect of our lives, and it wouldn't be a big deal in the first place if employment ended.

Re: Tensions in Google's ethical AI group increase as it sends demands to CEO

#225

Post by lawnchair_larry » Thu, Dec 17, 2020, 4:32 PM UTC

jasonvorhe wrote: post #187

hn_throwaway_99 wrote: post #117

Wow, Google's ethical AI group sounds like the most toxic group of individuals I've ever heard of. I'm going to sidestep the debate about whether Gebru was "fired" or she "resigned", because that debate primarily centers on how one defines the words. But clearly she did do this: She sent a letter that made a bunch of (IMO) outrageous demands, outrageous because it essentially attacked other people at the company, and…

That's quite the statement considering that there's lots of PoC, women, queer people, sexual abuse victims, etc who have been suffering through toxic work environments for ages. What makes this group "the most toxic group" you've ever heard of?

Considering that she has started falsely calling her former coworkers including Jeff Dean a bunch of racists who fired her for being black, I’ll agree that she sounds extremely “toxic” and essentially impossible to have in your team. I don’t really understand where you get the idea that those minority groups have it any worse these days.

Re: Tensions in Google's ethical AI group increase as it sends demands to CEO

#226

Post by pron » Thu, Dec 17, 2020, 4:32 PM UTC

mpweiher wrote: post #214

pron wrote: post #186

Earlier quoted context omitted.

I don't think anyone would claim corporations are democracies, but the view that the corporation is its CEO and/or its shareholders isn't so obvious, either. There are clearly cultural differences here (e.g. see the situation in Germany [1]), but I don't think that the American view of what a corporation is and who it serves is necessarily the only reasonable one. I don't know any more details on this matter than tho…

As a German who has tested the labor laws himself: behaviour such as that displayed would absolutely be grounds for getting fired without the notice period ("fristlose Kündigung") here. And it's actually pretty hard to legally fire people. The reason is "Verlust des Vertrauensverhältnisses", i.e. loss of trust. You start making demands coupled with threats, and public ones at that, you are no longer an employee.

But if employees are well-represented on the board then such actions might not be necessary in the first place. In this case, they might have determined that they have no other recourse to change the company's behaviour.

Both firing them or dealing with them are expressions of management's political power, and making their demands public is an expression of the political power of those employees, and it's up to both sides to figure out what result is best for them, a result that will likely be determined by the relative political power of the sides. I see no moral prerogative that says that management's exertion of political power is good, but workers' is bad.

Re: Tensions in Google's ethical AI group increase as it sends demands to CEO

#227

Post by cbozeman » Thu, Dec 17, 2020, 4:32 PM UTC

GVIrish wrote: post #94

ergocoder wrote: post #58

Earlier quoted context omitted.

I have no idea why this matters. Timnit gave an ultimatum (i.e. do this or I'll leave). Google agreed to her condition of parting way. This is like I go to my boss and ask for a $1 billion dollars. Otherwise, I'll leave. Then, my boss decides that I should leave since the boss can't give me $1b. Then, I keep telling my friends that "well, I'm actually fired.". huh?

She gave a set of conditions and stated that if they couldn't be met, that she would work with Google to schedule a resignation date. Google then cut off her corporate access and terminated her employment immediately, ergo, firing. If someone says, do this or I'll jump, and the other person pushes them, that's not the same thing as the person taking the leap on their own. The result might be the same, but the way the…

She wasn't immediately fired, her resignation was immediately accepted. Difference.

Re: Tensions in Google's ethical AI group increase as it sends demands to CEO

#228

Post by commandlinefan » Thu, Dec 17, 2020, 4:33 PM UTC

BitwiseFool wrote: post #133

hn_throwaway_99 wrote: post #117

Wow, Google's ethical AI group sounds like the most toxic group of individuals I've ever heard of. I'm going to sidestep the debate about whether Gebru was "fired" or she "resigned", because that debate primarily centers on how one defines the words. But clearly she did do this: She sent a letter that made a bunch of (IMO) outrageous demands, outrageous because it essentially attacked other people at the company, and…

I've noticed a trend where toxic people hide behind a veneer of a righteous cause, usually something related to social justice. It's effective and usually short circuits the public's willingness to see the nuance of the issue.

> toxic people hide behind a veneer of a righteous cause

Motte and Bailey: https://en.wikipedia.org/wiki/Motte-and-bailey_fallacy

Re: Tensions in Google's ethical AI group increase as it sends demands to CEO

#229

Post by mpweiher » Thu, Dec 17, 2020, 4:34 PM UTC

pron wrote: post #226

mpweiher wrote: post #214

Earlier quoted context omitted.

As a German who has tested the labor laws himself: behaviour such as that displayed would absolutely be grounds for getting fired without the notice period ("fristlose Kündigung") here. And it's actually pretty hard to legally fire people. The reason is "Verlust des Vertrauensverhältnisses", i.e. loss of trust. You start making demands coupled with threats, and public ones at that, you are no longer an employee.

But if employees are well-represented on the board then such actions might not be necessary in the first place. In this case, they might have determined that they have no other recourse to change the company's behaviour. Both firing them or dealing with them are expressions of management's political power, and making their demands public is an expression of the political power of those employees, and it's up to both…

No, they are expression of the fact that employees are employed and paid by the company to work for the company.

You can agitate against companies all you want, it's a free country. But you can't expect to be paid by them while you do it.

Not sure what "morality" has to do with it.

EDIT (Since parent added the bit about boards): No, having employee representatives on the board changes nothing about this. The board is legally (and ethically) obligated to work for the benefit of the company.

Re: Tensions in Google's ethical AI group increase as it sends demands to CEO

#230

Post by bonoboTP » Thu, Dec 17, 2020, 4:35 PM UTC

red75prime wrote: post #211

bonoboTP wrote: post #87

Earlier quoted context omitted.

She's Eritrean-born and identifies as Black. Races aren't an exact science but social constructs. There's a gradual continuum in people's appearance along geographical areas, linking Sub-Saharan west Africa where most slaves were taken from in the Atlantic slave trade, to East Africa, to North Africa and the Middle East, to the Mediterranean coast, to central Europe to Scandinavia to Siberia to Central Asia to East A…

It doesn't mean that a person cannot genetically belong to a well defined group with little admixture.

The question is how geographically localized that "well defined group" is. Are Eritreans in the same well defined group as Ghanans or as Lebanese? Are Cypriots the same as Lebanese or Italians? Are Italians in the same well defined group as Greeks or as Germans? Are Germans like Slovaks or like the Sami?

You can surely say that the overlap in a Ghanan and a Sami person's ancestry is quite small regarding recent millennia. But I don't see well defined groups at all. The reason it seems so in the US is because of the huge Western European (German/British) immigration early on and the large population of descendants of West African slaves. These two are sharply distinguishable (discounting for intermixing effects) but there are no "well defined groups" such as "black" and "white" when looking at it globally, not just in America. They are linked continuously via North/East Africans/Arabs/Turks and Mediterraneans.