Google’s AI-powered ‘inclusive warnings’ feature is very broken • Hacker News Board

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

#841

Post by noduerme » Sat, Apr 23, 2022, 5:23 AM UTC

Spivak wrote: post #830

epivosism wrote: post #812

Earlier quoted context omitted.

"Avoid using gray-box, graybox, or gray box to describe testing." "grayed-out, greyed-out, gray out, grey out Don't use. Instead, use unavailable." WTF are these people on about. Oh it's because they hate certain color words such as "black" and "white", words used by humans to indicate certain reflectivity properties of light; since it would be weird to only prohibit words which contain white or black, they just exte…

It’s not a race thing, it’s a disability thing — “greyed out” is only something that makes sense to someone who can see, it’s confusing the actual concept with the visual indicator of the concept. It’s the same with “the box is checked” to mean “selected” or “redlining a contract” to mean “negotiating.” The point of inclusive language isn’t some weird political points contest. It’s writing in a manner that’s clear to…

>> The point of inclusive language isn’t some weird political points contest. It’s writing in a manner that’s clear to people of lots of different backgrounds and doesn’t rely on just having to know English idioms or have common shared experiences.

I don't think efficiency and universal comprehension is the point at all. Deaf people understand the phrase "loud and clear" just as well as hearing people. Blind people understand "redlining" as well as sighted people do.

Policing every bit of language to cater to every potential group robs the language of its color, which is something I suppose you couldn't say anymore. Robs it of its beauty. Robs it of its poetry. Destroys its capacity for nuance. In the name of what?

Can you possibly be serious? Is your idea of perfect communication something that omits every daily experience of every person who can see, hear, read, and walk? Should these concepts never be mentioned?

How is it "violence" to someone in a wheelchair to talk about walking to the store? They know they're in a wheelchair. They know other people can walk. You think preaching "avoiding the need" for idioms that refer to walking is somehow fixing inequality? It's daft and ludicrous virtue signaling, and worse, it's insulting to everyone you think you're protecting.

[Edit] The worst is, it's a power play. Google should not nudge or dictate the language you use in any document, let alone a private text document! Words are our tools, human tools, to use as we wish to express what we want, the way we want, and you and Google have no right to police them.

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

#842

Post by rayiner » Sat, Apr 23, 2022, 5:24 AM UTC

seanmcdirmid wrote: post #805

rayiner wrote: post #746

Earlier quoted context omitted.

You can’t overlook the role that white progressives play in selectively amplifying minority voices. I can’t get on CNN to give the “Bangladeshi guy take” on rising crime by saying stuff most Bangladeshi people agree with. I have to provide a Bangladeshi perspective that is appealing to the white progressives that make the decisions at CNN. There’s a deep synergy here that exploits the fact that most Americans have li…

> And I’ll tell you that most desis are closer to Tulsi Gabbard than Harris. Including the Russia stuff—given India’s long-standing affinity with Russia. No surprise there: Modi is a conservative populist who could make Trump blush, and is very popular (and polarizing) in India. I’m not sure how the translates to Indian Americans, Indian politics isn’t very portable.

I’m talking about Indians in America. They represent an elite slice of Indians. :9 most aren’t fans of Modi, and most aren’t populists.

But at least with respect to pre-Fox News Gabbard, the areas where she departed from white progressives were recognizably desi to me. You don’t celebrate your abortion or make your sexuality part of your identity in Indian American communities. Likewise, there tends to be a strong impulse towards realism in foreign policy that accepts brutal dictators like Assad as necessary in some cases to avoid a breakdown in order. and there is a strong Russian affinity, since India was aligned with the Soviet Union during the Cold War. Meanwhile, Harris quite closely tracked white progressive orthodoxy.

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

#843

Post by Rastonbury » Sat, Apr 23, 2022, 5:24 AM UTC

nxm wrote: post #616

la6472 wrote: post #548

Earlier quoted context omitted.

I support it and I don’t work for google. I like this feature that encourages more empathy in this strife and hate filled world.

So Google execs now decides how we speak… brave new world. You like it until it corrects you for not being woke enough

Where is the line drawn? MS Word suggests style corrections too, despite the meaning being totally understandable.

As an enterprise feature, enterprises would interested in this because they wouldn't want their employees communications to get them cancelled or put on woke twitter. Despite their writing style guide[1] being a bit too woke for my liking i'm sure some people find it a good reference, especially if they don't have capacity to create their own corporate guide.

[1]https://developers.google.com/style/word-list

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

#844

Post by noduerme » Sat, Apr 23, 2022, 5:32 AM UTC

oefrha wrote: post #828

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

Geez, that word list. Here’s an idea. American Big Tech should form a Uniban Consortium to streamline the process of adding and banning new harmful words of the month. Then your OS update could come with support for, say, Unicode 20.0 and Uniban 6.0: five new emojis and fifty new words banned! Btw, the word main should probably be put on a future batch; too similar to “man” to be comfortable.

Better yet, Google's AI will just write your Docs for you from now on. Letting people type their own text files is unsafe, and may cause harm.

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

#845

Post by rougka » Sat, Apr 23, 2022, 5:33 AM UTC

onion2k wrote: post #635

rougka wrote: post #503

Earlier quoted context omitted.

I want to offer greenlist and redlist And to those who say this is not inclusive to the colorblind, I say traffic lights

Greenlist and redlist would fail in a global company. Plenty of cultures use red as a color that promotes positivity rather than a negative color. Japan has blue rather than green for their "Go" color. And so on.

I still think traffic lights would work as that’s almost a universal global standard (isn’t Japan the only exception?) and it works well in English

But I think we can agree that the goal when choosing these names isn’t really being concise and easy to use

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

#846

Post by noduerme » Sat, Apr 23, 2022, 5:36 AM UTC

Spivak wrote: post #838

epivosism wrote: post #834

Earlier quoted context omitted.

Yeah, I appreciate the sentiment and I don't think most of it is out of ill-will. I just think it's ineffective. I'd rather just accept the need to educate every generation of its own responsibility for treating people fairly - rather than pretend that fixing words would make a difference. There are also cases like "white-glove" where the phrase doesn't just represent the stereotypical "white = good & black = bad" -…

I completely agree, although in my own writing I would probably avoid white-glove not because of any race association but again because it’s an idiom that you have to have existing context to know. If your upbringing didn’t make you associate white gloves with fastidious butlers then you will probably be confused. You might instead associate cleanliness with the yellow gloves worn by many cleaning staff.

Or people could learn something by, for example, encountering the phrase and looking up the historical context.

The extension of your logic is that all writing should be geared to people with no knowledge, no historical reference, no firm understanding of the language, no visual or auditory cues (how do you talk about music without harming someone? Seriously, I ask this as a former music columnist). That is not language. [Edit: It's at best a graphic, except even that would be prohibited]. That is the conveyance of minimal information, like, 'don't stand here (in the place formerly known as being in front of a door).' It's completely absurd. It's inhuman. It's bent and wicked.

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

#847

Post by sio8ohPi » Sat, Apr 23, 2022, 5:36 AM UTC

elpakal wrote: post #65

I was extended an on-site interview at Google and after reading this and the comments I think I'll pass.

This particular feature is indeed half-baked and dumb, but FWIW, much of what I've seen claimed in the comments is outright false.

If you know anyone within Google, I suggest privately asking their experiences, rather than sampling only those people who are mad about Google on HN.

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

#848

Post by epivosism » Sat, Apr 23, 2022, 5:36 AM UTC

oefrha wrote: post #828

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

Geez, that word list. Here’s an idea. American Big Tech should form a Uniban Consortium to streamline the process of adding and banning new harmful words of the month. Then your OS update could come with support for, say, Unicode 20.0 and Uniban 6.0: five new emojis and fifty new words banned! Btw, the word main should probably be put on a future batch; too similar to “man” to be comfortable.

Yeah, the emoji manipulation of this is much more subversive and also continuously going on. Forbidding emojis which get converted to suggestiveness. We're left with eggplant and water droplets, which they know they can't remove, but they also are on guard against adding anything new that would allow more expressiveness in that area.

If they could they'd ban the letter X to prevent people typing "sex"

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

#849

Post by tintedfireglass » Sat, Apr 23, 2022, 5:38 AM UTC

rayiner wrote: post #779

screye wrote: post #761

Earlier quoted context omitted.

Desi here, like Tulsi more than Kamala. I'd say Kamala is more disliked than the average white person too, because she panders so incompetently. I have not heard great things about Pramila Jayapal either.

What tickled me is that the same white people who played up Harris’s south Asian background—the only ones to do so—in the 2020 primary attacked Gabbard specifically for views she shared with lots of other Indians: being a bit socially conservative, opposing US intervention in Syria, and having a soft attitude toward Russia. It was almost as if they knew nothing about desis.

Exactly.

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

#850

Post by season2episode3 » Sat, Apr 23, 2022, 5:41 AM UTC

donkeybeer wrote: post #788

rmbyrro wrote: post #745

Earlier quoted context omitted.

These people remind me of crusaders a thousand years ago, fighting the "infidels" all over. In fact, the whole body of thought governing them reminds me too much of the Latin Church. I have no words to express how I detest the kind of thinking behind them.

How many people have these people killed and how many wars have they started in the name of inclusivity?

It's hard to tell tone from text, but I'm going to assume you're not being facetious here... I'm pretty sure the answer is 0.