When will this madness end? There will always be people offended by something. I'm pretty sure there are individuals claiming that NOT suggesting their pronoun deprives them of their identity, dignity or whatever.
> There will always be people offended by something.
That does not mean every instance of offense should be ignored or dismissed, does it? How about addressing specific grievances instead of this extremely vague proclamation?
How do you determine if the grievance is valid or not? Sometimes "addressing specific grievances" should be to recommend the greivers apply for therapy.
There are people up in arms about Maryland's flag because the checkered pattern is a confederacy symbol which somehow triggers them. The best course of action for those people is not to change the Maryland flag but rather to get them therapy so they can learn how to live without getting offended.
> How do you determine if the grievance is valid or not?
I can't give you an algorithm in advance that will let you decide every political and social issue ever. That's obviously not possible.
There are heuristics, like. You can put yourself in the other person's shoes, try to empathize with them to figure out where they're coming from, and work out what their values and their reasoning is. You're a human being, a social creature whose most distinctive characteristic is extremely complex and nuanced communication. Your brain was made for this.
> There are people up in arms about Maryland's flag because the checkered pattern is a confederacy symbol which somehow triggers them. The best course of action for those people is not to change the Maryland flag but rather to get them therapy so they can learn how to live without getting offended.
I'm gonna let you in on a secret. If you treat everyone with contempt and dismissal, then every conclusion you come to about them will be contemptful dismissal.
>Sometimes "addressing specific grievances" should be to recommend the greivers apply for therapy.
People often assume I'm a man on reddit. A lot of the time I'll politely let them know that not everyone on reddit is a man. Are you suggesting that I ought to go to therapy instead?
If you became offended that someone assumed you were a man, then yes, I would recommend therapy. Politely correcting them? Of course not, that's a completely reasonable thing to do.
Consistently taking offense where none was intended warrants therapy in my book.
If you became offended that someone assumed you were a man, then yes, I would recommend therapy. Politely correcting them? Of course not, that's a completely reasonable thing to do.
Consistently taking offense where none was intended warrants therapy in my book.
The madness is gendered words to be buried so deeply in our psyche in the first place. There's limited need for those assumptions but due to history we let them rise to the top.
Its perfectly straight forward to write evocative and passionate prose without concreting to a gender. It makes sense that when we build general purpose tools we encourage them to choose the neutral words more readily because it increases their accuracy without any loss of meaning.
Are you really so terrified of a world where "meeting him" is "meeting them"?
>There will always be people offended by something.
Trying to remove bias from a system is about trying to remove bias from a system. People tend to agree that removing bias from a system rather than intentionally creating a system that discriminates.
@Quarrelsome: "Them" wouldn't work here, because it's one specific person which has a concrete gender. 99% of people would assume that the investor is some company with multiple people if you say "them".
"You" and "They" is proper usage, it has been well before this whole dialog came into play. It's a catch-all. There's nothing about a made-up pronoun that captures an ambiguity about an individual, let alone their individuality. They have to express that themselves. Pronouns are not what jeopardizes or validates individuality.
I agree on "they" but not "theM". Also I would only use "they" when talking about some generic person that really doesn't have a specific gender (or any other specific characteristics).
If I'm writing an email about my investor I know who that is and what gender they are. Getting "them" as completion option is almost never what I would wan't in that case.
(Above I even used "they" subconsciously, because It was just an example and I don't actually have an investor to talk about here.)
The connecting thread through all of them but the last is that there is some cost or difficulty to doing something about them, whereas the last is just a ticket in the bugtracker. Wouldn't it be nice for Google if all the activism in the world was directed towards things that could be solved at no cost by a junior engineer?
> I'm pretty sure there are individuals claiming that NOT suggesting their pronoun deprives them of their identity, dignity or whatever.
This would be similar to how feminist movements in languages that systematically distinguish men from women push to get the language changed so that women are referred to as men ("there shouldn't be any difference between a waiter and a waitress"), while feminist movements in languages that don't draw the distinction push to get the language changed so that women are referred to explicitly as women ("we should acknowledge that women can be waiters too").
If feminism demands one of these things, it surely can't also demand the other.
"they" are different groups that might use the same label. They can absolutely use the same justification, to argue for opposite changes.
Their justification could be as simple as, "We want to be respected," and different groups of feminists disagree with each other about how best to achieve that goal.
Sure, it's even worse if you personally land in the crosshairs, but from your example it sounds like the feminists are working in different places with different languages. And perhaps the problem of gaining respect requires different solutions in different areas.
The nature and beauty of language is that it changes. There is nothing static about it. The problem with chronocentric thinking is that you expect things to remain the same forever. Language changes, that's just how it is.
I don't believe this is a move based off of the sensitivity at Google's workplace, especially given what we know now of Google's policies around workplace harassment. I would hesitate to extend it to Google's culture as a whole, or any political movement that is being partaken in here.
If you think people who are working to improve the social climate in which we find ourselves are in it for the money, I can only conclude you don't know any of them.
Assume there are 80% included in a game and 20% frozen out. Here it makes sense to change the rules of the game to include the other 20%.
Yes it might slightly inconvenience the 80% but the switch from not being included at all to playing the game for the 20% is huge. So, overall it might be a net positive.
Problem is when you get to a 98/2 split. Is it worth it to inconvenience such a large majority for the benefit of a very small minority. And how much are you going to change the rules?
Is it worth the reduced enjoyment/efficiency of the game?
At some point the trade off will become too much, and even moral/emotional arguments will not be enough to silence the dissent.
Asking people to refer to you how you want to be addressed
IS madness. You must then ask every person how to refer
to him/her/them/it/cis/trans/pan before you assume standard
English is good enough.
Your madness is my ideal. I really enjoy living in a time that recognizes more than a binary sense of gender identity. Our language should change to address this. Just like it has changed to recognize that we use computers now.
Being offended when an algorithm picks one of two words that represent 99.9+% of the population is madness.
Also the words in question are not used to address a person. When I am speaking to a person, regardless of their gender identity, I use the second person pronouns you, yours, or yourself. These are already gender neutral. The requested gender neutral pronouns (xir/xe/ne/???) are used in the third person and lower the value of communication as there is no asexual physical presentation, people present as men or women based on their fashion and physical characteristics. Going to a restaurant and saying to the attendant that you are here to me "that xir over there" does not help them figure out which table to take you to.
Not to mention that people who feel gender neutral are still male or female (or have experienced a physical developmental disorder, humans are dimorphic after all.)
Where would it end, if we just kept inventing pronouns? Everyone getting their own pronoun? Like, perhaps a unique name? In the end, there'd just not be any pronouns. Problem is, pronouns are useful. They compress text.
One time I wrote a paper in high school where I didn't use any pronouns and always referred to Abraham Lincoln as "Abraham Lincoln", just so I could pad the paper as much as possible with the full length of Abraham Lincoln's name.
"Abraham Lincoln, in Abraham Lincoln's famous Gettysburg Address, suggested that Abraham Lincoln could..."
> Google’s technology will not suggest gender-based pronouns because the risk is too high that its “Smart Compose” technology might predict someone’s sex or gender identity incorrectly and offend users
Nobody cares when touch keyboards predict the wrong pronouns, so why would the same service in an email client be any different?
Bias is one of the new hot topics in ML. People act surprised that a face recognition software trained on a database of white people is unable to properly classify pictures of people from all around the world. The algorithm is only learning the data it was fed.
Touch keyboards are (mostly) reliant on the user's history, so they offer a ready made "it's your own damn habits" defense that's less accessible for Google.
A larger issue is that touch keyboards predict using single words and pretty quickly spiral into Markov chain incoherence if you let them draft whole messages. Smart Compose tries to offer syntactic and even semantic understanding. If I type "The investor is here, do you want to meet", my phone keyboard is working off 'meet', and will probably suggest 'up'. Smart Compose is working off 'investor' also, and apparently arrived at 'him'. (It's not clear to me from the article if this was actually tracked to specific biases, or if the predictions just adopted the default masculine.)
But most of all, I'm guessing Google's core motivation is explained by the offhand comment that Smart Reply uses the same logic. A gendered pronoun suggestion probably isn't going to spark a scandal, but a feature that encourages you to send Google-drafted text unedited might be subject to a lot more scrutiny.
No one actually cares, but a small minority will throw a temper tantrum for political leverage.
Google doesn't actually care either. I hate call it virtue signaling but that's totally what it is even though that term is completely worn out at this point.
Edit: Why is my personal opinion flagged? Am I really such a terrible person for thinking that way? We are so tolerant, but yet people can't have different opinions?
The substantial point was that I do not think that gender is just an important thing to be so worried about. Googles fear of gender bias is just not reasonable, at least not in that case.
Why can I not express that feeling in a short comment? Would it have been better or any clearer if I wrote a "substantial" paragraph about it?
I got the feeling that "formality" is more of an excuse to remove non-echo voices from here. Almost like in a political subreddit...
> Robots don’t have genders—they’re metal and plastic and silicon, and filled with ones and zeroes ... The problem is that even if a robot isn’t gendered, and even if it doesn’t look human or even animal, you’ll tend to want to gender it.
Alternate wording: "Google's Gmail autocomplete no longer assumes gender."
From the article:
> Gmail product manager Paul Lambert said a company research scientist discovered the problem in January when he typed “I am meeting an investor next week,” and Smart Compose suggested a possible follow-up question: “Do you want to meet him?” instead of “her.”
So really, this is a fairly reasonable decision whereby it doesn't infer gender from signals such as profession.
The article doesn't address this, but I wonder what it would do given an explicit gender, e.g. "She's the investor I am meeting next week."
True, doing antecedent linking would be neat. I think that would be a good solution in order to get highest 100% accuracy. Although, if it linked the antecedent incorrectly, it could still put them in controversy.
Yeah, this is a hard problem and they probably still have mistakes in front of them. Still, though, I think "we misidentified the antecedent in this sentence" is easier to forgive than "we incorrectly assumed a gender based on profession". The former is just a grammar parsing bug, while the latter carries a lot of cultural baggage.
'Vincent is in the lobby, do you want to meet them now?' Them who? Vincent and his unannounced friends? Maybe Vincent is actually 3 kids stacked on top of each other in a trench coat?
They is not singular. Replacing gendered pronouns with 'they' shits the flow of the sentence and creates ambiguity. There's a whole class of expressions that just stop working without a singular pronoun, hence the proposed 'ze' or whatever it is these days.
And what do you propose for all the languages where nouns and verbs and adjectives are gendered? In Russian, using gender-neutral phrasing when speaking about someone is an insult that implies that they're a thing rather than a person.
"They" is commonly used as a singular third-person gender-neutral pronoun in English — so much so that it's often unconscious. I glanced through your comments; you used it yourself just yesterday [1]:
> Anecdote time: a friend from one of the nation's major universities was involved in a road authority funded study to evaluate the safety impact of newly installed speed cameras at several intersections. [...] They found there was none.
Edit: after I replied, you edited your comment to add the caveat "except sometimes in 3rd person". "They" is always a third-person pronoun, and there are no criteria that would make it acceptable only in a subset of situations.
Edit 2: now the caveat is gone. Can you just reply to me rather than changing what I'm replying to retroactively?
Your comment-stalk sample is bad though because that could easily refer to the administrators of the study or everyone who ran the study. The sentence you are trying to link to it is about the friend being involved in the study, not being the one coming up with findings.
> Vincent is in the lobby, do you want to meet them now?' Them who? Vincent and his unannounced friends? Maybe Vincent is actually 3 kids stacked on top of each other in a trench coat?
My them example was in reference to the parent's phrase, not a general prescriptive. And in your example, while them would be a bit odd, from context you can gather them refers to Vincent alone. If there were further people one would likely be saying Vincent and his children are in the lobby, do you want to meet them now? In either case, I was primarily referring to they.
> And what do you propose for all the languages where nouns and verbs and adjectives are gendered?
To my ear, as a 34 year old native English speaker from Scotland, the phrase "Do you want to meet them?", meaning non-gendered singular pronoun, sounds absolutely totally fine and not strange in even the slightest way, and always has done since I was a child.
"Do you want to meet them?" is perfectly acceptable. People seem to have no issues using it to refer to me, somebody who doesn't feel particularly at ease with referring to themself as either "him" or "her".
I'm not going to repeat myself - my opinions on that topic are recorded in the comments of the article about Richard Stallman's opinion of "singular they".
If this is a system that groks the meaning of the input text and constructs original text to match, then sure, it could just use non-gendered language.
But that seems unlikely - we're probably talking about an engine that just recommends whatever occurs frequently in its corpus (hence the article talking about the team black-listing swear words and so forth). In which case adding gendered pronouns to the list seems reasonable enough, and non-gendered forms will presumably show up regardless (to whatever extent they appear in the corpus).
> it doesn't infer gender from signals such as profession
I'm struggling to tell from the article if Smart Compose ever actually did that. It's certainly implied, but the explanation given about gender balance in different fields is offered by the reporter. The parts actually attributed to Googlers only say they're being cautious about gendering in their outputs and couldn't find a fully-acceptable solution. Which leaves me curious they found specific semantic biases or instead noticed general issues like "our corpus taught the NLG to use a masculine default and simple fixes like randomizing pronouns created new issues".
(Either way, it seems like a reasonable decision. I'm just wishing for a more technical breakdown out of engineering curiosity.)
The other interesting point I saw is that Smart Reply apparently runs on the same system as Smart Complete. Since that offers fully-formed ideas, I imagine gender issues there could be vastly more embarrassing than single-word autocompletions.
When's the Hebrew language version coming out? There are many languages for which this "solution" is completely impracticable. On the other hand, there should be no problem with Turkish or Persian.
English is probably one of the harder languages to get gender right in, since only pronouns are gendered.
In more gendered languages, the problem is easier, since the AI could just look at nearby words to make a guess. In those languages, when it gets it wrong, people would likely assume that it just picked the wrong referent, instead of assuming gender.
This is fine. Guessing gender based on things like name is terribly error prone. Imagine if you have a name like Taylor. But I’m sure everyone who wants to be annoyed by this will find a way to be annoyed.
Now that you point it out, I'd imagine that aside from any bias pronoun choice would also be one of the least accurate suggestions offered. When somebody writes "my cousin is in town, do you want to meet", it's basically impossible to do better than a coinflip chance - and that's if offering a pronoun next was even correct. So it's no real surprise that they didn't spend a bunch of effort on debiasing a feature that would still be wrong if they succeed.
I'd like to point out how long-hanging fruit this is for Google to address gender imbalance and gender inequality, especially given Google's much larger issues with their policies around inappropriate sexual behavior. You'd think more significantly, Google could work on establishing better policies and a more equal workplace culture.
> Men have long dominated fields such as finance and science, for example, so the technology would conclude from the data that an investor or engineer is “he” or “him.”
Err, so the AI correctly predicts that statistically you probably intend "him", but we limit the utility of the tool because that would be discriminatory? I know it will get it wrong sometimes, and you can say it re-inforces stereotypes, but if it will get it right most of the time based on strict statistical inference, seems like it could at least be configured. It seems to be a case where the AI is a bit too accurate... I don't disagree with their decision, I think given the circumstances it's actually a brilliantly safe move and keeps out of the fray as much as possible.
> “The only reliable technique we have is to be conservative,”
I know this is just whimsical, and a horrible logical equivocation on my part, but that's kind of funny that a large tech company decided being conservative might actually be useful in some cases to help protect against liberal outrage...
And I find it ironic that in this discussion of choosing words carefully, we still use the basically meaningless (because of all of the different ways to use them) words "conservative" and "liberal". The words have nonpolitical meanings, and political meanings. And even the political meanings are inconsistent.
Yes, the goal of the modern Civil Rights Movement has basically been to say "in these areas, even though Bayesian inference produces statistically valid results, we as a society have decided that it's immoral to use it."
What axes are OK to use Bayesian inference on and what are not is a philosophical and historical question with lots of practical implications across politics, economics, actuarial science etc. It's really worth thinking about. But here's a start: in general, people are more forgiving of inferring data from mutable traits than immutable ones.
For people who are eager to think this is a don't-assume-gender charade: try misgendering a cis person next time you're out, and see how much they don't care. _You_ are who this is for
I've never been offended by getting misgendered, but hearing people get misgendered does, at times, bother me in the sense that it can make the sentence difficult to understand.
Gender is a big thing to get wrong in person. I am a man, If someone referred to me as "she" or "her" I would correct them. I wouldn't expect any less from anybody.
When it's a machine doing it though, some unknown piece of metal spitting out a form message, it seems it would be easier to forgive.
The gender neutral is just a dangerous ideology. The example in the article is not even good:
Article example: Type the neutral Turkish phrase “one is a soldier” into Google Translate and it spits out “he’s a soldier” in English.
Ok, since armies worldwide are mostly composed by men due to their physical strength and biologically design to fight. Is it unfair in any way to assume a "he" rather than a "she"? Not at all.
It's just naturally that genders have different preferences and some words are associated with one or another. This politically correctness bullshit is just a way to control what people can say and think.
This gender thing is just retarded. Women today enjoy greater benefits and freedom than ever, makes no sense to police the languages. In German, Portuguese, French, etc. words have gender like male, female and neutral. In English there is no gender in the word but you can attach it by pronoun. Honestly, no one feels oppressed by this. If in a group there is 1 man and 2 women, (in Portuguese) we call the group "eles" which is "male they" and it's a cultural thing and no one cares about this.
Gender in word has an immense importance to poetry where you can invoke cultural properties of each gender in romantic languages. This gender ideology wants to people to thing that men and women are the same thing but (f* obviously) are not.
The underlying issue is a political one that it's making it's way to several fields. This is not about red pill or hate, it's about common sense.
The translation "He's a soldier" will probably be accurate 99% of the time because... you know... armies are mostly made by men. It's likely that in the training dataset, "She's a soldier" doesn't appear as often.
If the autocomplete/translation suggestion is not accurate, you can drag the cursor, click and edit it. Problem solved.
The autocomplete is not meant to be biased or 100% accurate all the time. Being an AI, will need a huge dataset of text to provide the autocomplete. Gender is just one variable that is there, what matters for the product is that it will get the intention right most of the times.
Ironically your comment indicates that you are in a bubble. You reflexively dismissed them by saying "red pill", which is precisely what your bubble tells you how to react to anything that says "Jordan Peterson".
That's not your own thought and it's not an honest or meaningful rebuttal to the OP. Take your own advice by watching some full length Peterson lectures with an open mind. You may find that you've been mislead.
I've seen his videos. I've even read one of his books and saw him speak in London on his book tour. Every interesting point he raises is overshadowed by his pseudoscience chaos is feminine and order is masculine nonsense.
He is sexist. His work espouses sexism.
> That's not your own thought
But please, tell me more about what goes on inside my own head.
"It's a very rare woman who at the age of 30 doesn't consider having a child her primary desire. And the ones that don't consider that, generally in my observation there's something that isn't quite right in the way that they're constituted or looking at the world."
- Jordan Peterson[1]
"It's a very rare woman who at the age of 30 doesn't consider having a child her primary desire."
Evidence shows this to be the case. It is rare.
"And the ones that don't consider that, generally in my observation there's something that isn't quite right in the way that they're constituted or looking at the world."
I see how that sounds sexist, but he says the same thing about men that don't want kids. It's also important to note that he qualifies the statement as "generally", not as a rule. Acknowledging there are exceptions. It's not the case for everyone.
He says that men are misguided for prioritizing work or money over kids. He actually makes the point that women are better at recognizing what leads to happiness, fulfillment and meaning. Which is family. That's his understanding based on decades of observation and research. But again generally speaking and there are exceptions.
It's a tool that's meant to predict things, and they weren't able to successfully predict something that is used extremely frequently in lots of conversations. So they decided, after 'several' other attempts, that they were best off not trying to make suggestions.
If you can't do something right after repeated attempts, don't do it.
I don't know whether other people do this, but when using voice recognition software, such as Google's, I "learn" what it is likely to understand, and then modify my statements to be more easily understood by the software. This is very useful when sending text messages, for example. I am able to quickly send a message in just a moment.
However, because of Google's limitations, the language I use is modified. Their "AI" is shaping my use of language, and thus, how I communicate with others. Multiply this by millions (billions?) of people, and this could have a real impact on culture.
What is Google's responsibility, in this regard? Certainly they shouldn't ignore the ways their technology could affect society. They should be making conscious, deliberate decisions. This is dangerous territory; and not an easy one to navigate. I am glad they seem to recognize that.
These times are indeed hyper-sensitive, and ridiculous exaggerations do occur. This is not one of them.
Gender dysphoria is a real, crippling, dangerous mental illness. Maybe in the future there'll be some pill or simple brain surgery that fixes it. Today, there's none of that.
Gender reassignment surgery generally helps. Calling people their preferred gender helps, as a complement or substitute to surgery. It's a hacky, inelegant solution, perhaps disgusting to some, but it fucking works! Goind the extra mile to avoid misgendering people is reasonable.
"Imagine if we could give depressed people a much higher quality of life merely by giving them cheap natural hormones. I don’t think there’s a psychiatrist in the world who wouldn’t celebrate that as one of the biggest mental health advances in a generation. Imagine if we could ameliorate schizophrenia with one safe simple surgery, just snip snip you’re not schizophrenic anymore. Pretty sure that would win all of the Nobel prizes. Imagine that we could make a serious dent in bipolar disorder just by calling people different pronouns. I’m pretty sure the entire mental health field would join together in bludgeoning anybody who refused to do that. We would bludgeon them over the head with big books about the side effects of lithium."
If automatic suggestion tools had more than a markov chain like understanding of language they likely would have been able to identify this as a use case and had a culturally appropriate UI. As is, completely blocking references to gender is the only way to fix the issue.
Heaven forbid that the millions of mindless sheeple using “Smart” Compose actually have to read and edit their email before hitting send.
If Google really cared about combatting bias, they would just use “she/her.” At least it would be grammatically correct.
It’s not about taking a political/social stance to them, it’s about making life convenient for their users by removing any form of cognitive processing necessary
Is it just me or does it seem like the only people getting upset in these situations are the people offended by others trying not to offend others? It just doesn't seem like a big deal either way. It's weird that people automatically view actions like this as some sort of moral command, and thus feel so strongly about it. At the end of the day, it's just a design choice to reduce inaccuracies.
It's not something I would've thought to do myself, but I can see the reasoning for it. Bias in AI is a real issue, and it's wise to consider it earlier than later. Sometimes it's something more culturally visible, like in this situation, but oftentimes it can be much more subtle and insidious. This step isn't going to fix much but it's part of a bigger effort to make considering bias one of the priorities. Ignoring bias in our AI will make our AI more human in all the bad ways.
On the other hand, Smart Compose already seems like a bad idea to me. It's good for their AI but bad for humans. This pronoun action feels like a micro optimization for this technology's social impact, while the entire feature itself is a small net harm for society, in my opinion. It's a subtle dampening of our personal voice and nuances.
Yeah, people are getting worked up about a good faith trade off in software features made by a company trying to create products useable by, preferably, every single person on the planet. And since gender is something were are - right now, whether you like it or not - revising our understanding of, it makes sense to wait until you have a better solution to the problem.
I mean, autocorrect can only take you so far anyway, your gonna have to change it a bit anyway and I assume most people with two brain cells knows the gender and preferred pronouns of the people mentioned in the convo and knows how to use them properly even if they don't have a clue about the pronoun debate. This is an edge case, but still something they saw value in avoiding a mistake in.
I read the article and I find this fascinating. I am happy they are giving it some thought. Pronoun political correctness (a.k.a. thoughtfulness or respectfulness) aside, if our society's training data leads AIs to make assumptions, we should be aware of it and be careful to detect it.
An incorrect pronoun isn't going to kill someone -- but we are gradually handing over more and more decisions to technologies that rely on AI/ML. Perhaps investigations into incorrect pronoun assumptions can lead to improvements in assumption errors in other areas (e.g. you think your self driving vehicle doesn't need radars and that only using cameras is sufficient? Just because it looks like a big fluffy cloud doesn't necessarily mean you can safely fly/drive through it).
Back to just pronouns though: if someone says "My teacher assigned me to read Act I of Macbeth tonight", most people would avoid a reply that uses he/she until an indication was given or they'd just ask "Who's that, Ms. McFadden? Yeah, I know, she assigns way too much homework!". If a human can be smart enough to get it right, then I'm glad folks are working on AIs getting it right too (or, for now, not making assumptions until they can get it right).
The fact that AIs are advanced enough for us to be thinking about these kinds of details is wonderful! :)
126 comments
[ 2276 ms ] story [ 3017 ms ] threadThis is getting ridiculous.
That does not mean every instance of offense should be ignored or dismissed, does it? How about addressing specific grievances instead of this extremely vague proclamation?
There are people up in arms about Maryland's flag because the checkered pattern is a confederacy symbol which somehow triggers them. The best course of action for those people is not to change the Maryland flag but rather to get them therapy so they can learn how to live without getting offended.
I can't give you an algorithm in advance that will let you decide every political and social issue ever. That's obviously not possible.
There are heuristics, like. You can put yourself in the other person's shoes, try to empathize with them to figure out where they're coming from, and work out what their values and their reasoning is. You're a human being, a social creature whose most distinctive characteristic is extremely complex and nuanced communication. Your brain was made for this.
> There are people up in arms about Maryland's flag because the checkered pattern is a confederacy symbol which somehow triggers them. The best course of action for those people is not to change the Maryland flag but rather to get them therapy so they can learn how to live without getting offended.
I'm gonna let you in on a secret. If you treat everyone with contempt and dismissal, then every conclusion you come to about them will be contemptful dismissal.
People often assume I'm a man on reddit. A lot of the time I'll politely let them know that not everyone on reddit is a man. Are you suggesting that I ought to go to therapy instead?
Consistently taking offense where none was intended warrants therapy in my book.
Its perfectly straight forward to write evocative and passionate prose without concreting to a gender. It makes sense that when we build general purpose tools we encourage them to choose the neutral words more readily because it increases their accuracy without any loss of meaning.
Are you really so terrified of a world where "meeting him" is "meeting them"?
Trying to remove bias from a system is about trying to remove bias from a system. People tend to agree that removing bias from a system rather than intentionally creating a system that discriminates.
If I'm writing an email about my investor I know who that is and what gender they are. Getting "them" as completion option is almost never what I would wan't in that case.
(Above I even used "they" subconsciously, because It was just an example and I don't actually have an investor to talk about here.)
ML driven automation widening wealth disparity? Eh.
ML ushering in the age of the persistent surveillance state? Eh.
ML being used to manipulate elections and hijack democracies? Eh.
ML might auto-suggest wrong gender in a text editor? STOP. EVERYTHING. THE RISK IS TOO HIGH IT MIGHT OFFEND SOME USERS.
This would be similar to how feminist movements in languages that systematically distinguish men from women push to get the language changed so that women are referred to as men ("there shouldn't be any difference between a waiter and a waitress"), while feminist movements in languages that don't draw the distinction push to get the language changed so that women are referred to explicitly as women ("we should acknowledge that women can be waiters too").
If feminism demands one of these things, it surely can't also demand the other.
Different groups of people call themselves "feminists," and don't agree with each other about changes they'd like to see in the world.
It's kind of odd to me that people (like you apparently) don't understand.
I think they all agree that they'd like other people to change as an acknowledgement of their power.
I also think that they can't both use the same justification to argue for opposite changes. What did I miss?
"they" are different groups that might use the same label. They can absolutely use the same justification, to argue for opposite changes.
Their justification could be as simple as, "We want to be respected," and different groups of feminists disagree with each other about how best to achieve that goal.
Sure, it's even worse if you personally land in the crosshairs, but from your example it sounds like the feminists are working in different places with different languages. And perhaps the problem of gaining respect requires different solutions in different areas.
1. People expect public organizations to not be bigoted.
2. People are generally fairly bigoted.
Therefore, if you're an organization training NNs on data produced by 2, you have to do some editing, or you're going to get in trouble over 1.
Including, apparently, the fact that some people are making an effort to care about other people.
The majority of people who "care" would have no cushy job otherwise.
This applies to programmers implementing useless features, social scientists, politicians, administrators and many more.
Assume there are 80% included in a game and 20% frozen out. Here it makes sense to change the rules of the game to include the other 20%.
Yes it might slightly inconvenience the 80% but the switch from not being included at all to playing the game for the 20% is huge. So, overall it might be a net positive.
Problem is when you get to a 98/2 split. Is it worth it to inconvenience such a large majority for the benefit of a very small minority. And how much are you going to change the rules?
Is it worth the reduced enjoyment/efficiency of the game?
At some point the trade off will become too much, and even moral/emotional arguments will not be enough to silence the dissent.
cf. https://medium.com/incerto/the-most-intolerant-wins-the-dict...
How much revenue would Google be losing over such a decision?
Even if something only affects 2% of the population, the cost of getting it wrong may not be worth the benefits.
Newspeak. When I first read 1984 I thought it so farfetched as to be laughable. I'm not laughing any more.
The analog between using computers and non-binary self-identification is a strawman/woman/person/thing. Which straw<something> will be acceptable?
Also the words in question are not used to address a person. When I am speaking to a person, regardless of their gender identity, I use the second person pronouns you, yours, or yourself. These are already gender neutral. The requested gender neutral pronouns (xir/xe/ne/???) are used in the third person and lower the value of communication as there is no asexual physical presentation, people present as men or women based on their fashion and physical characteristics. Going to a restaurant and saying to the attendant that you are here to me "that xir over there" does not help them figure out which table to take you to.
Not to mention that people who feel gender neutral are still male or female (or have experienced a physical developmental disorder, humans are dimorphic after all.)
Guessing something with a 50% probability of being wrong is not a useful feature.
> I'm pretty sure there are individuals claiming that NOT suggesting their pronoun deprives them of their identity, dignity or whatever.
Yeah but the end result of everyone wanting to feel special is that in the end, no one will.
One time I wrote a paper in high school where I didn't use any pronouns and always referred to Abraham Lincoln as "Abraham Lincoln", just so I could pad the paper as much as possible with the full length of Abraham Lincoln's name.
"Abraham Lincoln, in Abraham Lincoln's famous Gettysburg Address, suggested that Abraham Lincoln could..."
Nobody cares when touch keyboards predict the wrong pronouns, so why would the same service in an email client be any different?
A larger issue is that touch keyboards predict using single words and pretty quickly spiral into Markov chain incoherence if you let them draft whole messages. Smart Compose tries to offer syntactic and even semantic understanding. If I type "The investor is here, do you want to meet", my phone keyboard is working off 'meet', and will probably suggest 'up'. Smart Compose is working off 'investor' also, and apparently arrived at 'him'. (It's not clear to me from the article if this was actually tracked to specific biases, or if the predictions just adopted the default masculine.)
But most of all, I'm guessing Google's core motivation is explained by the offhand comment that Smart Reply uses the same logic. A gendered pronoun suggestion probably isn't going to spark a scandal, but a feature that encourages you to send Google-drafted text unedited might be subject to a lot more scrutiny.
Google doesn't actually care either. I hate call it virtue signaling but that's totally what it is even though that term is completely worn out at this point.
No it's not.
Edit: Why is my personal opinion flagged? Am I really such a terrible person for thinking that way? We are so tolerant, but yet people can't have different opinions?
On HN the idea is: if you have a substantive point to make, make it thoughtfully; if you don't, please don't comment until you do.
https://news.ycombinator.com/newsguidelines.html
p.s. Flamewar comments, like you posted at https://news.ycombinator.com/item?id=18541549 and elsewhere, will get your account banned on HN, so please don't do that.
Why can I not express that feeling in a short comment? Would it have been better or any clearer if I wrote a "substantial" paragraph about it?
I got the feeling that "formality" is more of an excuse to remove non-echo voices from here. Almost like in a political subreddit...
https://www.wired.com/story/robot-gender-stereotypes/
> Robots don’t have genders—they’re metal and plastic and silicon, and filled with ones and zeroes ... The problem is that even if a robot isn’t gendered, and even if it doesn’t look human or even animal, you’ll tend to want to gender it.
From the article:
> Gmail product manager Paul Lambert said a company research scientist discovered the problem in January when he typed “I am meeting an investor next week,” and Smart Compose suggested a possible follow-up question: “Do you want to meet him?” instead of “her.”
So really, this is a fairly reasonable decision whereby it doesn't infer gender from signals such as profession.
The article doesn't address this, but I wonder what it would do given an explicit gender, e.g. "She's the investor I am meeting next week."
Is the idea just to eliminate all useful suggestions that contain genders?
Say do you want to meet them? would work fine in this case.
They is not singular. Replacing gendered pronouns with 'they' shits the flow of the sentence and creates ambiguity. There's a whole class of expressions that just stop working without a singular pronoun, hence the proposed 'ze' or whatever it is these days.
And what do you propose for all the languages where nouns and verbs and adjectives are gendered? In Russian, using gender-neutral phrasing when speaking about someone is an insult that implies that they're a thing rather than a person.
> Anecdote time: a friend from one of the nation's major universities was involved in a road authority funded study to evaluate the safety impact of newly installed speed cameras at several intersections. [...] They found there was none.
[1] https://news.ycombinator.com/item?id=18532386
Edit: after I replied, you edited your comment to add the caveat "except sometimes in 3rd person". "They" is always a third-person pronoun, and there are no criteria that would make it acceptable only in a subset of situations.
Edit 2: now the caveat is gone. Can you just reply to me rather than changing what I'm replying to retroactively?
My them example was in reference to the parent's phrase, not a general prescriptive. And in your example, while them would be a bit odd, from context you can gather them refers to Vincent alone. If there were further people one would likely be saying Vincent and his children are in the lobby, do you want to meet them now? In either case, I was primarily referring to they.
> And what do you propose for all the languages where nouns and verbs and adjectives are gendered?
I don't. I was explaining they usage that a lot of english speakers seem to not be aware of. I agree with this linguist: https://slate.com/human-interest/2018/08/singular-they-prono...
"Do you want to meet the <noun>?"
"Do you want to meet <name>?"
"Do you want to meet ___" <prompt for autocomplete from set of pronouns>
etc.
I'm not going to repeat myself - my opinions on that topic are recorded in the comments of the article about Richard Stallman's opinion of "singular they".
People use English this way constantly and it's well understood.
But that seems unlikely - we're probably talking about an engine that just recommends whatever occurs frequently in its corpus (hence the article talking about the team black-listing swear words and so forth). In which case adding gendered pronouns to the list seems reasonable enough, and non-gendered forms will presumably show up regardless (to whatever extent they appear in the corpus).
I'm struggling to tell from the article if Smart Compose ever actually did that. It's certainly implied, but the explanation given about gender balance in different fields is offered by the reporter. The parts actually attributed to Googlers only say they're being cautious about gendering in their outputs and couldn't find a fully-acceptable solution. Which leaves me curious they found specific semantic biases or instead noticed general issues like "our corpus taught the NLG to use a masculine default and simple fixes like randomizing pronouns created new issues".
(Either way, it seems like a reasonable decision. I'm just wishing for a more technical breakdown out of engineering curiosity.)
The other interesting point I saw is that Smart Reply apparently runs on the same system as Smart Complete. Since that offers fully-formed ideas, I imagine gender issues there could be vastly more embarrassing than single-word autocompletions.
In more gendered languages, the problem is easier, since the AI could just look at nearby words to make a guess. In those languages, when it gets it wrong, people would likely assume that it just picked the wrong referent, instead of assuming gender.
Err, so the AI correctly predicts that statistically you probably intend "him", but we limit the utility of the tool because that would be discriminatory? I know it will get it wrong sometimes, and you can say it re-inforces stereotypes, but if it will get it right most of the time based on strict statistical inference, seems like it could at least be configured. It seems to be a case where the AI is a bit too accurate... I don't disagree with their decision, I think given the circumstances it's actually a brilliantly safe move and keeps out of the fray as much as possible.
> “The only reliable technique we have is to be conservative,”
I know this is just whimsical, and a horrible logical equivocation on my part, but that's kind of funny that a large tech company decided being conservative might actually be useful in some cases to help protect against liberal outrage...
What axes are OK to use Bayesian inference on and what are not is a philosophical and historical question with lots of practical implications across politics, economics, actuarial science etc. It's really worth thinking about. But here's a start: in general, people are more forgiving of inferring data from mutable traits than immutable ones.
I know this is annecdotal, but you did ask.
Edit:
In case, one is curious or unbelieving. Consider that my name is Nikita.
When it's a machine doing it though, some unknown piece of metal spitting out a form message, it seems it would be easier to forgive.
Here you go Google, I fixed it for you
Jordan Peterson had to fight back this: https://www.youtube.com/watch?v=s_UbmaZQx74
The gender neutral is just a dangerous ideology. The example in the article is not even good:
Article example: Type the neutral Turkish phrase “one is a soldier” into Google Translate and it spits out “he’s a soldier” in English.
Ok, since armies worldwide are mostly composed by men due to their physical strength and biologically design to fight. Is it unfair in any way to assume a "he" rather than a "she"? Not at all.
It's just naturally that genders have different preferences and some words are associated with one or another. This politically correctness bullshit is just a way to control what people can say and think.
I get that AIs need to be moderate to avoid things like this: https://www.technologyreview.com/s/610634/microsofts-neo-naz...
This gender thing is just retarded. Women today enjoy greater benefits and freedom than ever, makes no sense to police the languages. In German, Portuguese, French, etc. words have gender like male, female and neutral. In English there is no gender in the word but you can attach it by pronoun. Honestly, no one feels oppressed by this. If in a group there is 1 man and 2 women, (in Portuguese) we call the group "eles" which is "male they" and it's a cultural thing and no one cares about this.
Gender in word has an immense importance to poetry where you can invoke cultural properties of each gender in romantic languages. This gender ideology wants to people to thing that men and women are the same thing but (f* obviously) are not.
Yet you're linking to Jordan Peterson videos. Maybe it's time for you to explore the world outside of your red pill bubble for a bit.
The underlying issue is a political one that it's making it's way to several fields. This is not about red pill or hate, it's about common sense.
The translation "He's a soldier" will probably be accurate 99% of the time because... you know... armies are mostly made by men. It's likely that in the training dataset, "She's a soldier" doesn't appear as often.
If the autocomplete/translation suggestion is not accurate, you can drag the cursor, click and edit it. Problem solved.
The autocomplete is not meant to be biased or 100% accurate all the time. Being an AI, will need a huge dataset of text to provide the autocomplete. Gender is just one variable that is there, what matters for the product is that it will get the intention right most of the times.
That's not your own thought and it's not an honest or meaningful rebuttal to the OP. Take your own advice by watching some full length Peterson lectures with an open mind. You may find that you've been mislead.
He is sexist. His work espouses sexism.
> That's not your own thought
But please, tell me more about what goes on inside my own head.
His wife and daughter would also disagree with the sexist claim. As well as his many female patients who he has helped with their careers.
> But please, tell me more about what goes on inside my own head.
I didn't mind read. You literally typed out a common dismissal. The sexist comment is also that.
[1] https://www.youtube.com/watch?v=NV2yvI4Id9Q&t=7m30s
"It's a very rare woman who at the age of 30 doesn't consider having a child her primary desire."
Evidence shows this to be the case. It is rare.
"And the ones that don't consider that, generally in my observation there's something that isn't quite right in the way that they're constituted or looking at the world."
I see how that sounds sexist, but he says the same thing about men that don't want kids. It's also important to note that he qualifies the statement as "generally", not as a rule. Acknowledging there are exceptions. It's not the case for everyone.
He says that men are misguided for prioritizing work or money over kids. He actually makes the point that women are better at recognizing what leads to happiness, fulfillment and meaning. Which is family. That's his understanding based on decades of observation and research. But again generally speaking and there are exceptions.
right after that he says there are some women who not maternal
It's also unacceptable to use HN primarily for ideological warfare.
https://news.ycombinator.com/newsguidelines.html
It's a tool that's meant to predict things, and they weren't able to successfully predict something that is used extremely frequently in lots of conversations. So they decided, after 'several' other attempts, that they were best off not trying to make suggestions.
If you can't do something right after repeated attempts, don't do it.
However, because of Google's limitations, the language I use is modified. Their "AI" is shaping my use of language, and thus, how I communicate with others. Multiply this by millions (billions?) of people, and this could have a real impact on culture.
What is Google's responsibility, in this regard? Certainly they shouldn't ignore the ways their technology could affect society. They should be making conscious, deliberate decisions. This is dangerous territory; and not an easy one to navigate. I am glad they seem to recognize that.
These times are indeed hyper-sensitive, and ridiculous exaggerations do occur. This is not one of them.
Gender dysphoria is a real, crippling, dangerous mental illness. Maybe in the future there'll be some pill or simple brain surgery that fixes it. Today, there's none of that.
Gender reassignment surgery generally helps. Calling people their preferred gender helps, as a complement or substitute to surgery. It's a hacky, inelegant solution, perhaps disgusting to some, but it fucking works! Goind the extra mile to avoid misgendering people is reasonable.
"Imagine if we could give depressed people a much higher quality of life merely by giving them cheap natural hormones. I don’t think there’s a psychiatrist in the world who wouldn’t celebrate that as one of the biggest mental health advances in a generation. Imagine if we could ameliorate schizophrenia with one safe simple surgery, just snip snip you’re not schizophrenic anymore. Pretty sure that would win all of the Nobel prizes. Imagine that we could make a serious dent in bipolar disorder just by calling people different pronouns. I’m pretty sure the entire mental health field would join together in bludgeoning anybody who refused to do that. We would bludgeon them over the head with big books about the side effects of lithium."
From http://slatestarcodex.com/2014/11/21/the-categories-were-mad...
If Google really cared about combatting bias, they would just use “she/her.” At least it would be grammatically correct.
It’s not about taking a political/social stance to them, it’s about making life convenient for their users by removing any form of cognitive processing necessary
It's not something I would've thought to do myself, but I can see the reasoning for it. Bias in AI is a real issue, and it's wise to consider it earlier than later. Sometimes it's something more culturally visible, like in this situation, but oftentimes it can be much more subtle and insidious. This step isn't going to fix much but it's part of a bigger effort to make considering bias one of the priorities. Ignoring bias in our AI will make our AI more human in all the bad ways.
On the other hand, Smart Compose already seems like a bad idea to me. It's good for their AI but bad for humans. This pronoun action feels like a micro optimization for this technology's social impact, while the entire feature itself is a small net harm for society, in my opinion. It's a subtle dampening of our personal voice and nuances.
I mean, autocorrect can only take you so far anyway, your gonna have to change it a bit anyway and I assume most people with two brain cells knows the gender and preferred pronouns of the people mentioned in the convo and knows how to use them properly even if they don't have a clue about the pronoun debate. This is an edge case, but still something they saw value in avoiding a mistake in.
An incorrect pronoun isn't going to kill someone -- but we are gradually handing over more and more decisions to technologies that rely on AI/ML. Perhaps investigations into incorrect pronoun assumptions can lead to improvements in assumption errors in other areas (e.g. you think your self driving vehicle doesn't need radars and that only using cameras is sufficient? Just because it looks like a big fluffy cloud doesn't necessarily mean you can safely fly/drive through it).
Back to just pronouns though: if someone says "My teacher assigned me to read Act I of Macbeth tonight", most people would avoid a reply that uses he/she until an indication was given or they'd just ask "Who's that, Ms. McFadden? Yeah, I know, she assigns way too much homework!". If a human can be smart enough to get it right, then I'm glad folks are working on AIs getting it right too (or, for now, not making assumptions until they can get it right).
The fact that AIs are advanced enough for us to be thinking about these kinds of details is wonderful! :)