Live data from Hacker News

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

vice.com

411–420 of 1001 posts

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

#411
post #407

Earlier quoted context omitted.

Being told I use too many adjectives, my word choice is too heavy for my audience or that I used too many clauses is useful. Being passive-aggressively told "blacklist" is a no-no word is not. If you were really clever, you'd let the user set a "sensitivity" level for the writing from "normal human speak" to "corporate PR approved".

It depends on who you write for. As a non-native, non-American speaker, I'd be happy to know if I'm accidentally offending the people I'm writing to.

That's the problem, though: the tool doesn't actually tell you that. The vast majority of Americans are not offended by words like "motherboard".

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

#412
This topic is always going to be highly subjective, but I find the author's examples of how this feature is malfunctioning to be extremely uncompelling.

"Motherboard" was included in Apple's style guide in 2020 as "don't use" (https://help.apple.com/applestyleguide/#/apsg72b28652). Like it or hate it, this is the direction the industry, of which Google is a part, is moving. I'm sure a publication called Motherboard might have an opinion on this, but neither they nor Google are the final arbiters on the language. The I Have a Dream speech example is pretty anecdotal; King was an excellent writer, and I don't think Google is claiming this tool would make a person write like King. And the substitution suggestion in Kennedy's speech is just the way the language use has trended since his time... One can note that the Star Trek franchise changed the saying from "Where no man has gone before" to "Where no-one has gone before" in the time between then and now, as well. The Bible is probably the worst example to pull up for this topic; paging through the over 20 translations on biblegateway.com shows the passage in question is also sometimes translated as "great works", "mighty works", or just "miracles." Which should you use? It depends completely on what you're doing.

Of everything noted, the only possibly concerning one is really the lack of suggestions on a David Duke interview. If I had to guess, I'd chalk that up to lack of training data, and it may be something Google wants to consider addressing.

But at the end of the day, the overall thesis of the report is flawed. "But words do mean things," says the author. Yes, they do. Which is why it could be nice to have an auto-editor lifting up examples of words that might mean something other than the author intended (and then the author can choose to change their phrasing or stick with the original). Nothing about this feature claims to make users magically better at writing; it's an assistive tool to open the possibility, not unlike a spell-checker.

We've had Microsoft Word grammar checker for literal decades. This technology is neither particularly new nor, IMHO, particularly interesting (certainly not interesting enough to kick up this news cycle). It's no more 1984 than some random stranger online offering an opinion on your verbage is 1984.

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

#413

Controversial suggestion but can the California folks just do their own bay area version of products? The rest of us around the world never asked for any of this weirdness Will it suggest "peace be upon him" in Islamic countries next? If not, why not? At least that's an actual religion unlike whatever this weirdness is. How is it inclusive to export this thinking to places that don't want it? Words are really just wo…

> Words are really just words. But they're not though. Words are products of, and inputs into, people's culture and modes of thought. Which is exactly why giving this much global power to one unelected politically extreme group of people is so dangerous.

The best example to drive the point home is the Nazi's "useless mouths". It's a terrifyingly effective term if your goal is to shape public opinion.

Another is the freedom fighter vs terrorist word choice.

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

#414
post #66

Earlier quoted context omitted.

It's in the banned words list, I can share a link internally if you want.

why does a company have a say which words are allowed to be used? i mean there are some bad words you shouldn’t use, but i just can’t grasp why is there something like “motherboard” on that list?

The banned words list they are referring to is about what words should be used in external product documentation and marketing :) Not about what you are allowed to say inside google, or anything like that.

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

#415

Earlier quoted context omitted.

it contains "lord" which is a male term I didn't think of that. I always thought of it as "lording over" something. As in she's the "lord of the land." That said, a landlord and a property owner are not the same thing. Many (most?) apartment buildings are owned by one company and managed by another.

> As in she's the "lord of the land." But she wouldn't be; she'd be the "lady of the land".

Queen Elizabeth II is the current Lord of Mann. And Dame Fiona Woolf was the Lord Mayor of London as was Mary Donaldson before her.

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

#416
post #352

Earlier quoted context omitted.

For anyone reading this, I volunteer. My moderation level would be based on words that were considered bad in the year 2005. I won’t have to do much work and you still get to say that you have someone that’s doing the job. Win-win.

I wonder what your age is in comparison to your chosen ideal of 2005. As I am getting older it is easier to see the patterns in all this. Most people just want the entire world frozen from the time they were young. That includes everything from the cast of SNL to acceptable language. I have seen enough decades of people complaining about policing language to know that we survived multiple waves of this before 2005 an…

I dunno, man. You're right that the term "mainboard" isn't going to kill us, and I wouldn't advise anyone to make this their main crusade in life. But that's a pretty high standard of dismissiveness and I've never seen anyone apply it to language changes that genuinely bug them. If we discovered that Google employees call codebases which have a lot of bugs "gay", and people got angry about it, would you tell them that it's not a big deal because Google has just developed the language a bit?

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

#417

Earlier quoted context omitted.

Encountering this aspect of the culture war always makes me wonder why are tech companies so focused on these gender minorities and not, say, on the disabled groups. There are more blind, deaf, mute (etc, etc) people in the US than transgender people who will bear a child. It would be equally incoherent to attempt to replace all usages of “see”, “hear” with “perceive”.

Trans people are frequently in tech. Go to any college and you'll see the CompSci department is the one with the most amount of LGBT individuals. The same can't be said for people who are blind or deaf.

But wouldn't an underrepresentation of disabled people be cause for concern in the industry? There are more legally blind people in the USA than there are Native Americans yet at the political stage there is very little interest in coming up for those people.

Several impressive videos on social media have shown that programming without using vision or even hands is perfectly possible, there are very few good reasons why such underrepresentation shouldn't be corrected for. In fact, I believe for many disabilities a job in fields like data science or programming would be much easier to adjust for disabilities than many other sectors where interactivity is key.

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

#418
post #376

Earlier quoted context omitted.

Sorry but did you just say one of the reasons you left google was the memegen moderation team? I can't imagine that being a legitimate worry that would impact my employment decisions. It seems there's a group of people who are too far in the other direction too. When I heard that it's preferred to say "allowlist" instead of "whitelist", do you know what my reaction was? "Sure, whatever." And I moved on with my life.…

You're right. I use the new terms and move on. It's really not that bad. However it tends to become a password game. There's a new password every few months. If you know the new password, you get to feel above those who don't. It's as we invent new crimes to charge people with. If no one calls it out (because sure, whatever), it keeps ratcheting up. Then banal conversations turn into minefields. What you say gets inv…

It's also terrible for real long term projects to be continually renaming things and modifying naming conventions. Engineering isn't fashion.

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

#419

> Cutting phrases like “whitelist/blacklist” and “master/slave” out of our vocabulary not only addresses years of habitual bias in tech terminology, but forces us as writers and researchers to be more creative with the way we describe things. > calling landlords “property owners” is almost worse than calling them “landchads,” and half as accurate. It’s catering to people like Howard Schultz who would prefer you not c…

Why not use proprietor?

If you say "I paid my proprietor this month", nobody will understand what you mean. It's simply a different word with a different definition.

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

#420
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…

When the same kind of people banned "master" branch for being offensive I was surprised. Not so much now if I think in their logic
Post reply on HN