Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Google’s AI-powered ‘inclusive warnings’ feature is very broken

vice.com

Post reply on HN
Print view View original

921–930 of 1001 posts

Re: Google’s AI-powered ‘inclusive warnings’ feature is very broken

#921

Post by knorker »

ask_b123 wrote: post #525
knorker wrote: post #156

Earlier quoted context omitted.

I think they're trying to change language to disassociate "black" with negative meaning. Which is going to be very hard, because the reason for the association is from black being the unknown. It's the night that hides the predators. It's the shadow where the enemy hides. It's where you don't want to put your foot in case there are spiders or a sharp rock. Humans are afraid of the dark. I've not even heard of being a…

On the mention of being afraid of the light I thought that something like that was probably mentioned in the Bible, and sure enough: https://biblehub.com/john/3-19.htm > The Light has come into the world, but men loved the darkness rather than the Light because their deeds were evil. Everyone who does evil hates the Light, and does not come into the Light for fear that his deeds will be exposed… Another thing; even t…

> Another thing; even though this effort might be doomed to fail, do you think there are good reasons to attempt to change language in such a way?

I'm not categorically against it. I'm not in the right country to properly judge "master/slave". Those words have no association with people to me. But I don't care about "master/slave".

Though I can't find it now, I remember encountering cases where actually "blocklist" was a less descriptive term, even misleading.

I think it's great even though "crewed" is a terrible word, I can't think of a better one, and it's good to change away from "manned".

I'm perfectly willing to surrender the word "niggardly".

So it depends.

Banning words like "ninja", "dojo", "white glove treatment", "blind"... like... really?

What I think is doomed to fail is banning any language that involves color.

I'd say some words are in a "grey area", but I've seen people wanting to ban that term.

Mandating language like this is double plus ungood.

Re: Google’s AI-powered ‘inclusive warnings’ feature is very broken

#922

Post by DavidSharff »

meetups323 wrote: post #11

I wonder if this is an example of not knowing your audience (or rather assuming one's self represents a wider audience than it does). The people vouching for this at Google are likely product marketing managers, public relations folks, social media managers, etc. All they do is write corporate garbage all day, and much like we have "nit"s in PR's for formatting, variable names, etc; they likely have similar reviews t…

Love it.

I do think there is more activism at play than you might. At a minimum it provides the fear mechanism that prevents employees from challenging the work, particularly if they aren't even on the project (why stick your neck out?).

However, the code format for prose analogy illuminates the writing quality angle brilliantly.

I'm sure I'm not alone in having distinct communication modes in work documents, each with their own reasons to be excluded from the ML dataset:

1. Keep it simple -- when achieved it is highly effective for communicating a work memo, but would be terribly dry if used in auto suggestions for an English major struggling through crafting moving prose.

2. Lazily verbose -- going long is more expedient than crafting a compact message. It's mostly unrefined garbage. (exhibit a this comment)

3. Everything is awesome (positivity inflation) -- a deluge of great, love, awesome, wonderful, perfect, etc. all applied to far too many things, far too often.

Imagine if an ML code formatter included psudeo code inputs, or was given a data set of every local file change (pre-commit) instead of what is sitting in a main branch? My Docs are filled with things that I'd never commit much less pass a code review.

If Google Docs was only used in corporate settings, and we wanted to double down on our drab corporate communication styles, I guess the feature would make some sense.

Re: Google’s AI-powered ‘inclusive warnings’ feature is very broken

#923

Post by knorker »

rnk wrote: post #732
knorker wrote: post #98

Earlier quoted context omitted.

Not only that, but there are more than three times as many slaves TODAY as ever were in the transatlantic slave trade that is the only one the US knows exists.

That feels like a very misleading point. The world population is almost 8 billion, it was almost 1 billion in 1800 (per https://en.wikipedia.org/wiki/World_population ), so we decreased slavery as a precentage of the world by a lot, yet the total number of slaves has increased. It's tragic that there is modern slavery, but it was horrible the us did it, it's horrible that people want to minimize and argue we weren't…

> That feels like a very misleading point. The world population is almost 8 billion, it was almost 1 billion in 1800 [...] we decreased slavery as a precentage of the world by a lot

I appreciate what you're saying, but my point is actually that the US view is that "slavery has been abolished for 150 years" makes for a very US-centric view of history. Which for a 250 year old country is laughably closed minded both backwards and forwards in time.

The context of what I said is the parent comment (https://news.ycombinator.com/item?id=31124042), specifically related to the US-centricism of the retold statement of "slavery was started by the United States".

The story of the world is not that slavery was created for the process of building the US as a powerful nation, and then abolished and now only dealing with the fallout and reparations. But I've heard this understanding many times in the US, to the point where it seems like the mainstream understanding there.

Re: Google’s AI-powered ‘inclusive warnings’ feature is very broken

#924

Post by raffraffraff »

marton78 wrote: post #594
stefan_ wrote: post #560

Earlier quoted context omitted.

Haha jesus no wonder they added the "AI-powered" hints because who the fuck can be expected to keep track of all this nonsense? > hang, hung > Don't use to refer to a computer or system that is not responding [..] see Avoid unnecessarily violent language.

Oh, that's due to violence? I thought because it evokes phallic associations, as when someone is well hung.

Well now it does. Thanks for nothing Marton.

Re: Google’s AI-powered ‘inclusive warnings’ feature is very broken

#925

Post by spangry »

darkerside wrote: post #880
noduerme wrote: post #803

This link in the article is insanely, crazily, madly, insultingly sickening to anyone who's not a slave. As I read it I felt like it was some kind of parody of Orwell. Disgusting to tell someone what verbiage they use sincerely and in technical terms is now verboten because it's harmful. To whom? Crazy people? https://developers.google.com/style/inclusive-documentation

I don't think it's the worst thing to encourage people too be thoughtful about their word choice. Is it really "crazy", or are you selling an emotional reaction by personifying your topic as a mentally deranged person? I actually think steering clear of this language is helpful in defusing the outrage culture that is so harmful to our online discourse.

I don't think these guys are trying to encourage people to be thoughtful, despite what they may claim - it feels more like one tribe using the power of google to try and control another tribe.

Giving these dummies what they want won't defuse anything. Thus far, doing so has not retarded the progress of this sort of language policing, but has led to increasingly insane demands. It's like handing your lunch money over to a bully in the hopes she will stop bullying you.

I think the way to stop outrage culture is to stop incentivising 'outrage' by refusing to indulge these lunatics.

Re: Google’s AI-powered ‘inclusive warnings’ feature is very broken

#926

Post by mrep »

titzer wrote: post #587
mgraczyk wrote: post #9

In the article they use the example of "Motherboard" and "Landlord" and seem to suggest that these are areas where the Google AI is making mistakes or being overly strict. As a Google employee expressing my own opinion and observations of company culture, I can say that these are 100% not mistakes. Many Google employees are just so out of touch with the real world that they believe it is the duty of Google Docs to ch…

Woke Clippy (to borrow from up-thread) is going to have a field day when it learns German. There. Are. So. Many. Violations.

Oh please give us some examples.

Re: Google’s AI-powered ‘inclusive warnings’ feature is very broken

#927

Post by bigmealbigmeal »

deanCommie wrote: post #868
screye wrote: post #760

Earlier quoted context omitted.

There is a truth to what both of you are saying. Because from what I know Kamala is disliked among both Indian citizens and Inner-city black communities.

Quoted post unavailable.

Bernie Sanders is an immediate and obvious counterexample to your assertion that people only do this for Kamala Harris and Hillary Clinton. There's also Jeb Bush, Pete Buttigieg, Marco Rubio, and many others.

The exclusion of these fairly obvious counterexamples suggests that you are engaging in a form of confirmation bias.

There is reason to believe that people are sexist toward female politicians, but the first-name basis of Kamala and Hillary is not a strong piece of evidence, given that we do that for the aforementioned male politicians (and don't do it for a lot of other female politicians, such as Elizabeth Warren, Sarah Palin and AOC)

Re: Google’s AI-powered ‘inclusive warnings’ feature is very broken

#928

Post by exodust »

dvirsky wrote: post #306
itsyaboi wrote: post #41

Earlier quoted context omitted.

Since you work at Google, maybe you can answer this: does internal tooling exist to flag occurrences of "bad" words in say, code reviews or shared documents?

Former Googler (until a couple of months ago). I've never seen anything like it, but maybe I just didn't use "bad" words in code? There were presubmit checks for typos and such. Also, IIRC I asked people in code/design reviews to rename white/black lists to allow/deny lists, but it might have just been in docs. I did get an angry code review response from a fellow engineer once, after writing in a commit description…

> "I asked people in code/design reviews to rename white/black lists to allow/deny lists"

Odd thing to do. "Black" is synonymous with darkness, hidden, off, empty space, and things of that nature. For this reason "blacklist" works. There is no clash with people, or any disrespect.

Politically correct OCD invents problems. Then dictate to others how they should solve those problems. Add penalties for those who don't follow the prescribed solutions to invented problems, and we find ourselves in 2022 looking at Google's efforts to solve the invented problems.

The author of the Vice article states that by removing master/slave and black/whitelist, it "addresses years of habitual bias in tech terminology".

Accusations of bias are just accusations. We can't "find" bias on the grounds of matching words in technical contexts to the same words in other contexts.

In the S&M world, the context is not actual master and slave, it's a new context that encompasses consent and good times. Likewise, the technical context is not one that involves people, but machines communicating under specific rules which "master/slave" adequately describes.

Re: Google’s AI-powered ‘inclusive warnings’ feature is very broken

#929

Post by UncleMeat »

saalweachter wrote: post #768
UncleMeat wrote: post #593

Earlier quoted context omitted.

MLK explicitly supported affirmative action and other programs designed explicitly to benefit black people in order to make up for past discrimination. I suspect he would be tired of being used as a justification for absolute colorblindness.

There's a long-running joke about MLK giving exactly one speech and never saying anything else.

Not even just one speech. Literally just one line!

Re: Google’s AI-powered ‘inclusive warnings’ feature is very broken

#930

Post by ulwutzor »

mattkrause wrote: post #500
user3939382 wrote: post #108

Earlier quoted context omitted.

If we're going to play this game, I think dissuading people from using "mother" in the example of "motherboard" is offensive. Mothers are the source of human life on earth, and in that capacity are revered and honored. When we say motherboard, we're making an analogy that suggests the board's significance and universal connection to everything. To discourage the term as an analogy for things that are a universal sour…

What about a motherboard is specifically maternal?

In French we say "carte mere" which is the literal translation of motherboard (carte = board and mere = mother).

The thing is, all nouns in French are gendered, and carte (board) happens to be female. So it can only be a mother, and not a father board. And the reason is that it carries smaller boards, like a mom would carry her babies (in most mammal species, as far as I know, the mom is carrying babies even after birth).

A disque (disk) is male. But a floppy disk translate to "disquette" (small disk) which is female : le disque, la disquette. Don't ask me why, but that's who the language works.

This would drive Silicon Valley wokes completely nuts. The table is female, the cloud is male (note that, in Italian, clouds are female). The TV is female, the glass is male, the door is female, the handle is female, the window is female, the roof is male, the wall is male.

Post reply on HN