Google Docs' AI-powered inclusive writing auto-correct now under fire
- Reference: 1650871633
- News link: https://www.theregister.co.uk/2022/04/25/the_latest_automated_ai_writing/
- Source link:
Language models are used in Google Docs for features like [1]Smart Compose ; it suggests words to autocomplete sentences as a user types. The Chocolate Factory now wants to go further than that, and is rolling out "assistive writing," another AI-powered system designed to help people write punchier documents more quickly.
Assistive writing is being [2]introduced to enterprise users, and the feature is turned on by default. Not everyone is a fan of being guided by the algorithm, and some people find its "inclusive language" ability irritating, Vice [3]reported .
[4]
Words like "policemen" could trigger the model into suggesting it be changed to something more neutral like "police officers." That's understandable, but it can get a bit ridiculous. For example, it proposed replacing the word "landlord" with "property owner" or "proprietor." It also doesn't like [5]curse words as one writer found.
[6]
[7]
"Assisted writing uses language understanding models, which rely on millions of common phrases and sentences to automatically learn how people communicate. This also means they can reflect some human cognitive biases," a spokesperson for Google told Vice. "Our technology is always improving, and we don't yet (and may never) have a complete solution to identifying and mitigating all unwanted word associations and biases."
Fairness in AI is complicated
As experts strive to create the holy grail of a perfect, unbiased intelligent system, fairness in machine learning models is proving to be a tricky thing to measure and improve.
Why? Well, for starters, there are apparently 21 definitions of fairness in academia. Fairness means different things to different groups of people. What might be considered fair in computer science may not align with what's considered fair in, say, the social sciences or law.
All this has led to a nightmare for the field of AI, John Basl, a philosopher working at Northeastern University in the US, [8]told Vox, adding: "We're currently in a crisis period, where we lack the ethical capacity to solve this problem." Trying to fix fairness is difficult, not only because people can't agree on what the term even means, but because the solutions for one application may not be suitable for another.
[9]
It's not always as simple as making sure developers are training on a more diverse, representative data set. Sometimes the impacts of an algorithm are different for different social groups. Although there is regulation in some use cases, like financial algorithms, there is no easy fix to make these models fair.
IBM: Ethics is a major roadblock to enterprises adopting AI technology
IBM CEO Arvind Krishna has risen through the ranks, working his way up over 30 years to lead IBM. He's witnessed booms and busts in the technology industry, and said that although AI is the future, he's careful about deploying its vast capabilities in the real world. Ah, yeah, that'll be why Watson wasn't [10]fully realized .
"We are only probably 10 per cent of the journey in [artificial intelligence]," he [11]said in an interview with the Wall Street Journal. "With the amount of data today, we know there is no way we as human beings can process it all. Techniques like analytics and traditional databases can only go so far."
"The only technique we know that can harvest insight from the data, is artificial intelligence. The consumer has kind of embraced it first. The bigger impact will come as enterprises embrace it." But Krisha admitted businesses are facing hurdles related to machine-learning models often being biased or the technology being used unfairly.
"We've got some issues. We've got to solve ethics. We've got to make sure that all of the mistakes of the past don't repeat themselves. We have got to understand the life science of AI. Otherwise we are going to create a monster. I am really optimistic that if we pay attention, we can solve all of those issues," he said. ®
Get our [12]Tech Resources
[1] https://support.google.com/docs/answer/9643962?hl=en
[2] https://workspaceupdates.googleblog.com/2022/03/more-assistive-writing-suggestions-in.html
[3] https://www.vice.com/en/article/v7dk8m/googles-ai-powered-inclusive-warnings-feature-is-very-broken
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2YmZxTn6hpSBfxRd375ZPfgAAAI8&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[5] https://twitter.com/k80way/status/1516427480353189902
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44YmZxTn6hpSBfxRd375ZPfgAAAI8&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33YmZxTn6hpSBfxRd375ZPfgAAAI8&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[8] https://www.vox.com/future-perfect/22916602/ai-bias-fairness-tradeoffs-artificial-intelligence
[9] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44YmZxTn6hpSBfxRd375ZPfgAAAI8&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[10] https://www.theregister.com/2022/01/31/machine_learning_the_hard_way/
[11] https://www.wsj.com/articles/ibm-ceo-says-ais-transformation-of-business-will-take-another-decade-11650360601?mod=djemCIO
[12] https://whitepapers.theregister.com/
unfix the unfixable
Trying to fix fairness is difficult,...
It is an impossibility to "fix" fairness. The whole concept of "fair" is a socio-cultural construct and a moving target.
Using machines to do human's work just codifies any current system with all its flaws, biases and problems. It soon becomes clear that nuances of the human interpretation are lacking and the target already has moved. Now we are working with a model that has yesterday's values and is frowned upon by today's users.
You'd need a thinking machine for a better result. But guess what, we already have a thinking machine! It is called a human (*) and it is a biological/chemical machine. Maybe it can improve the AI by replacing the AI?
(*) the amount of thinking embedded in a human may vary from individual to individual. No express or implied guarantees are given for any particular human's thinking ability nor its expression of said thinking.
Orwellian nightmare
How long will it be before a Police Person is sent to your habitation because you used to many non inclusive words in the last 3 months ?
If anyone actually believes that Inclusive words actually change the world, think again because the victim culture will always find news methods in order to reman victims..
Re: Orwellian nightmare
Language does matter: would you, for example, consider saying "sexual deviant" rather than "homosexual"? Phrasing can shape beliefs and understandings (see Chomsky), and I doubt that many people would argue that referring to people with a particular range of skin colours using the n-word is completely fine and will have no effect on societal attitudes.
Unfortunately, some people certainly take it too far and seem to seek to be offended. While I agree that there's a lot of nonsense around, I think it's important not to throw the baby out with the bathwater. (Can I say that?)
Re: Orwellian nightmare
It's your last part that sums it up nicely. "Can I say that?". I know it's a joke, but with the current state of affairs you always have in the back of your mind the thought that you might be saying the wrong thing. You might call it "WrongThink". ;-) This can stifle free flowing thoughts and discussions, and leads to many people who just don't say anything because they are afraid it might be deemed controversial (even at some point in the future).
I don't think the Police will be banging on your door for saying "mother" just yet, but you could well lose your job for saying something which doesn't go along with the current narrative, or perhaps you just don't get a job because someone in HR has quickly scrolled through your public musings and found something that's considered "problematic". (or perhaps an AI has scanned your whole output and flagged anything that might embarrass your potential employer).
This IS an Orwellian nightmare, and it's very likely to be already happening.
Re: Orwellian nightmare
And that is why you stay off social media and connect as little as possible of your online persona to your real world identity. Because no matter what your opinion is, someone will find it offensive. And what is offensive in the future might be completely acceptable at the time.
Re: Orwellian nightmare
This IS an Orwellian nightmare, and it's very likely to be already happening.
It's already happening in Scotland, where the police regularly come down on people for using what they consider to be offensive language.
Re: Orwellian nightmare
So who gets to decide what's acceptable or not, and how often do the rules get updated, daily, weekly, monthly etc ..
How can you consider that ALL homosexuals accept the usage of the word homosexual, what do you do about those that don't accept that word... Maybe some don't even want any word used to describe them as that in itself could be considered exclusive..
There is no end to that method of thinking....It a vicious circle with no winners.... A contemporary exemple
would be how to define people with non white skin.. Is it ok to say Black , Brown, Africa, Midle Eastern, People of Colour or BAME, which is acceptable and which is not ?
I have many coloured and or gay friends/colleagues that admit that they have become disturbed by all these word games. Especially when the games are being played by people that are not concerned by it at all....
Re: Orwellian nightmare
" Unfortunately, some people certainly take it too far and seem to seek to be offended
Even more unfortunately, and more commonly than that, some organisations and individuals appoint themselves to be offended on behalf of others whom they often haven't even asked whether they're actually offended.
Not a good suggestion
'For example, it proposed replacing the word "landlord" with "property owner" or "proprietor."'
All landlords are property owners, but not all property owners are landlords. "Proprietor" is a different thing altogether - the owner of an unincorporated business.
May I suggest "rent seeker" as a suitable gender-neutral term?
Re: Not a good suggestion
It certainly avoids historically misogynist attitudes to acceptable parentage.
More seriously, I think the concern here is that, as spoken communications are increasingly replaced by written communications, a very small number of vast private international corporations are in a position to mediate our conversations and dictate the language in which they are held.
I suspect IBM's CEO has missed the point: the problem is not that we can't process the vast amount of data we now have, the problem is that we have the data in the first place, collected on the off-chance it might turn out to be useful.
The solution to this is not technology, it's rigorous regulation.
Re: Not a good suggestion
In fact "landlords" are not necessarily property owners. We still refer to pub tenants as landlords.
Also, where property is sublet the occupant's landlord won't be the owner.
say the unsayable
It looks like Google has one global 'right' answer here atm.
I am sure it will occur to them that this is not a spellchecker.
So they will attempt to segment the models...
...This can lead into a reverse facebook echochamber hell - your language will be corrected into more and more extreme shapes.
Just as people easily fall down rabbit holes of watching things with increasingly erm 'odd' views
I am sure people easily fall down rabbit holes of saying things with increasingly erm 'odd' speech
Re: say the unsayable
It looks like Google has one global 'right' answer here atm.
Google must already know what you are going to write! It has been analyzing you for many years and knows you. Google should take the next logical step and cut out the user completely while writing. Google knows how to write neutral. Google knows what you want to write. Google knows all. Let google do the writing for you and you only need to press "send". Problem solved.
Linguistics don't have a solution...
"We're currently in a crisis period, where we lack the ethical capacity to solve this problem."
No, you have asserted the existence of a problem that doesn't exist, this is why you are unable to "solve" it.
know the unknowable
Alternatively, if enough data can be obtained on the sender and receiver, it is possible to generate something that is appropriate.
Just think what you could do with something that generated language fitted to your ears modeled as being from a trusted source
Godwins law in 7 comments.
Inclusive writing auto-correct
One feature I definitely won't be using. I've published a couple of novels already and working on my third. The last thing I need is for something to interrupt the flow when I'm mid-sentence. It sounds worse and more annoying than Clippy.
"Our technology is always improving, and we don't yet (and may never) have a complete solution to identifying and mitigating all unwanted word associations and biases."
Not adding unwanted words and phrases would be a good start.
"We are only probably 10 per cent of the journey in [artificial intelligence]"
But we'll get there in 10 years.
I see you have been writing from the future again. It must be nice to write these comments with hindsight.
Of course hindsight says 10 years. It's always been 10 years.
"AI-powered system designed to help people write punchier documents more quickly."
A system which provides the writer with anodyne words & phrases might succeed with "more quickly" but will certainly fail with "punchier".
> but will certainly fail with "punchier"
That's a good point, because particular words/phrases could be being used deliberately for their specific effect.
e.g.: "All the policemen - because of course there were no police women present - were terribly denigrating at the scene."
So, would we expect this google feature to change (or constantly annoy me butting in with it's suggestions) the sentence to:
"All the police officers - because of course there were no police officers present - were terribly denigrating." ?
As someone writing something, especially something at least semi-formal, you should be able to recognise if language being used isn't appropriate, and make the active decision to go looking for better language, by for example, highlighting it and choosing a 'thesaurus'-type function to give a choice of other phrases. If you don't realise that you should be using alternative language to go looking for those alternatives, either it is a situation that it doesn't matter - dashing off a quick casual text to people you are familiar with - or a situation where the reciever of that communication may want to know that you can't recognise that. For example, a job opportunity, where the potential employer would want the 'unvarnished' communication to stand to help them assess the sort of person the author is.
"He's witnessed booms and busts in the technology industry..."
If they're mic booms then I guess he must've spent some time on set during the filming of those busts.