Almost every time I get a call from TELUS about a new service or promotion, it’s someone from the Philippines or India. A lot of them speak English fluently, but the accent and phrasing can be pretty different from what I’m used to, and I don’t always catch everything they’re saying. Sometimes I feel like I’m guessing a big chunk of the conversation, which makes me not want to engage, especially on sales calls.
It matters more when I’m the one calling them for billing or technical support. In those cases, clarity really counts, and it can get frustrating when I have to keep asking for repeats or try to piece things together.
Honestly, I’d love something like this for my own speech too. I’m Japanese and have a fairly strong accent, and it would be nice if people could understand me more easily without having to guess.
I've tried to keep the habit of talking about things in the third-person when I'm on the phone with someone: instead of saying "you messed up the billing" I say "BigCo messed up the billing".
It's a small mental reminder that it's not the fault of the person I just happen to be talking to.
- Is it dehumanizing to Filipinos that Filipinos probably now do their job more efficiently without having to learn an accent that they are not exposed to?
- Is it dehumanizing to Filipinos that they no longer enjoy having their accent heard as a externality of a counterfactual arrangement?
- Is it dehumanizing to the customers that the company does not expect their customers to be cosmopolitan enough to understand a foreign accent with ease?
- Is it dehumanizing to the customers that the customers are now more sensorily shielded from a current-day reality regarding globalized providers of service?
- Is it dehumanizing, not due to this decision itself; but the globalized arrangement, to Canadians that they cannot expect to hold such a job and get by in Canada? Or perhaps to Filipinos, that such a job might be low-paying in their own country (or in respect to non-domestic goods that need to be purchased from outside their polity)?
- Is it dehumanizing, regarding not this decision, but the offshoring decision, that such decisions can be made without consent by employees and contractors?
On the other, too many customers are complete racist dicks to people who they perceive as not "belonging to their country". I... don't think this is the solution to that problem (people will just start applying their racist views elsewhere), but it could be argued by some that it might help.
I'm still against this, don't get me wrong - we absolutely should not be doing this to anybody. I can understand the appeal, though.
Are you going to also fight the good fight for Chinese and Japanese depictions of and reactions to black people, for example? Because those caricatures are certainly worse.
But I think so long as people are given the choice it's not dehumanising at all. Just like how I choose to speak a little slower if speaking to someone who doesn't speak English very well when it becomes clear they're struggling to follow what I'm saying.
So in a way it's actually more human than completely ignoring the reality of a situation like that. Same as that first human binding the leg of another.
If I have a hard time with accents and someone has a thick accent, the technology is not too different from the sci-fi babblefish concept, automatic translation for the recipient. It is always presented as an enabling technology.
I have no expectation that sci-fi analysis of a potential technology is correct or complete. But I do think we can think about why this feels so different.
In this case I think neither recipient nor speaker has opted in, and I think deceptively at that. It would feel different if the recipient is turning on an assistive technology because they are having a hard time understanding, or if the speaker is turning on an assistive technology because they are having a hard time doing their job.
Technology is supposed to make life easier and better
When calling support in my own country it is much faster and easier, because they intuitively understand the type of issue I’m having and can better relate. I question if changing the voice would make it more frustrating, as I’d have similar issues without the obvious explanation as to why it’s happening.
This is a different kind of way of using AI to eliminate local jobs and allow them to more easily outsource it to countries with low labour costs and poor labour conditions.
While I would appreciate being able to understand them better, I would not at all support this. You could maybe make an argument that using this with local staff could have some merit. As at least then they are not exploiting cheap foreign labour. There are still people living within the country of the caller who may still have strong accents like in the example you gave about yourself.
Obviously not enough of them. Most are used to under-bidding and being stretched to take the lowest possible price.
Its not for the person on the other end.
I used to do phone tech support, and:
1. Lots of my female coworkers would end their shifts in tears because men would yell at them for no reason. A male voice would absolutely make the job more bearable for them.
2. Singaporeans hate Australian accents more than anyone over here hates indian accents. I had a nearly 100% strike rate with singaporeans demanding local tech support, calling me names and hanging up.
From GP
> Almost every time I get a call from TELUS about a new service or promotion
I’d hate to see accents removed in movies and e.g. YouTube review videos. But sales and customer service have lost their humanity long ago. At least the call center workers will receive less bigoted hate and hard-of-hearing customers will be less confused.
My reason for mentioning this, is that there are going to be weird bugs in any such system. Systems hallucinate. Misunderstand words. I can see accent removal meaning that different words are the result, and context can mean those different words could be a disaster. This immediately opens up liability, because it doesn't matter if it was a computer, a human, or who, a company is on the hook.
It also doesn't matter if another company is providing this service, your contact is with Telus. Telus may sue their company, but you're going to go after Telus. A company could agree to all sorts of things without meaning to, make fraudulent statements, and yes they are liable and always have been. That also includes hate-crime related legislation, harmful insults, snide comments, and here's the fun part...
The person on the other end doesn't even know what they're saying to the person. Not accurately. This is supposed to be seamless, so they'll think that what they're saying is coming through correctly. And continue talking.
Yes, humans can do all of these things. But often there's a manager walking around the room, listening, and would hear someone raising their voice, yelling at the end-user, swearing, making inappropriate statements. This would stand out.
Yet here we have a system altering what's being heard, and no one is directly in the loop on that. No manager. No person on the floor.
Frankly, I hope this explodes in their face. Hard. I want to see them sued so hard, that no other company tries to ever interfere with human conversation again. Go full AI? OK. Full human? OK. But this nonsense???
Absolutely not.
However, thanks to this AI 'assistance' its becoming what was actually intended to be said by the people and what was made up the LLM, with some people creating wordy pages long LLM babble.
This also prevents non-native speakers from actively getting better, which is a core issue with AI general.
Also I think people who are not native speakers are often overly concerned with how much other people are bothered by broken English and accents (as long as accents are clear enough that the point can be understood)
OP likely just has more self-awareness than most in being able to be honest about it.
You must be very lucky to always get "a lot" of fluent English speakers.
Just this week I was speaking to Microsoft (well, their Indian outsourcer, of course).
As is the case 99% of the time, the guy was not at all fluent.
I'm not being rude here. I live in a large city in a Western country, I have friends and colleagues who are Indian and I encounter Indians in day-to-day life. These people all speak English in a truly fluent manner. Yes they still have the strong accent, but guess what the accent has never caused me a problem.
Telus thinking they can magically fix the lack of fluency through AI because the "problem" is the accent ? Now that IS being rude and disrespectful.
Thank you for your attention to this matter.
A) this will be used to hire non Canadian with minimal language skills and will be bad for the local labour market without objection from customers
B) accents are troublesome but the biggest issues were people that don’t have the same cultural standards for things like, not lying, not dumping calls that were hard, or doing a good job with complicated systems and accurately logging cases truthfully.
So many problems are created by poor workers (opps we deleted the customers account, oops I transferred them too you).
These were problems that were so bad they had to have specific cultural training for specific nations to get people to the Canadian standard, and many failed. But hey, cheaper labour!
Now I clean houses, and there is so much competition from people from abroad who are flooding the market and undercutting prices and I don’t get government subsidies to live in a hotel…
Canadians get a lot of scam calls from Indian call centres. Whether it's furnace cleaners or somebody calling about a fraudulent amazon package you supposedly ordered, it's usually somebody with an Indian accent. It's reached the point where many people simply hang up if they hear an Indian accent on the line. If you're trying to do telemarketing, possibly using the very same call centres that run these scams, that's a huge barrier.
Telus, for its part, is absolutely shameless in its use of aggressive telemarketing. I'm not surprised that they're one of the first companies to employ this sort of innovation. Unfortunately, this tech will likely spread to the scammers almost immediately, assuming it didn't originate with them.
As an aside, here's one of my favourite games to play with telescammers: Pick one word to say over and over again, but attempt to give it a variety of natural inflections, ambiguities, etc. so that it sounds like you're not just saying one word. Then see how long you can keep the scammer on the line. Start your stopwatch the moment you start talking to a human. I once managed over three minutes with the word, "Fuzzy-cuffs". Every minute of their time you waste could be a minute somebody's Grandma isn't being scammed.
e.g. This was in the news yesterday, but there's basically always a scam of the week/month going around: https://www.cbc.ca/news/canada/toronto/scammers-rogers-custo...
The solution is to hire people who can be understood, and empower them with the authority to be effective. Elaborate and misleading schemes to mask the problem are disrespectful and insulting to your customers. If the job involves speaking with customers, candidates for it should have the communication skills required. I've dealt on a regular basis with foreigners / ESL'ers who are perfectly capable of speaking English, French, etc. even with an accent.
If the job involved regular heavy lifting, would you hire someone incapable of doing so?
Can’t help poor education.
It's not inherently racist, just very lazy, but it can quickly lead to racism.
Related last year:
AI Accent Conversion for call centers (48 points, 70 comments) https://news.ycombinator.com/item?id=43514141
Call centres using AI to 'whiten' Indian accents (8+6 points, 0+6 comments) https://news.ycombinator.com/item?id=43246376 https://news.ycombinator.com/item?id=43292311
As soon as I hear the "Mr Firstname and how are you today?" I hang up.
Call spammers have not worked out that a formal polite greeting is a big giveaway.
That said, Sarvam, Gnani, and a number of other Indian AI companies are working on dialect aware TTS for localization usecases.
Is this actually a thing? (Translating to American, it's the culinary equivalent of crepe-pizza-burger-clam chowder.)
a sweet korma, or a vindaloo are my most favorite.
Scam calls sounding "more legitimate" because it passes the (unfortunately racist) filters most people have.
I realized quickly how it was changing my thinking process to devote so much to each word.
Anyone have the original source?
Most modern call centers / support is completely pointless and almost nothing would be lost by not having them. This is assuming that you'd provide just a half decent self service and have actual information available, written in a clear, easy to read, language.
You say they're not good enough, they smell, they don't fit in, but you take their culture, their clothing, their food and rebrand it as scandinavian, high fashion, chic fitness, pumpkin spice. They do the things you value but for their skin color.
You pay them colored people wages, with colored people working conditions with no social mobility outside of where they live, but you literally rob them of voice.
Your lack of ability to see "why this is dehumanizing is" why you're replacing yourselves with A.I. "AI is better" f*k. AI is controlled by a few platform owners. Once everyone is replaced with AI they're jacking up the cost and no one of any color is eating. Just a few rich.
So yeah, i can understand why you think it isn't dehumanizing. You don't see when you do it to others, or when we do it to ourselves.
It's hard to decipher anyone when you can hear 30 other people in the background and the audio is choppy.
I wonder if it doesn’t make more sense to do it on the client side. I would love to have an app installed on my machine that does this for me, because then I also have the option to turn it on and off.
* https://www.bbc.com/worklife/article/20160317-inside-the-sec...
I had to give them various pieces of information that had to be accurately transcribed, e.g. a Hide My Email - generated email address.
Normally for this like this, I tell rep that I'm going to read the full email, then spell it out using the NATO phonetic alphabet, then read it out again, and this usually works great.
This particular rep was entirely unfamiliar with the NATO phonetic alphabet and couldn't reliably make out make bog-standard North American accent, so I spent probably five minutes on the phone to just read off my email address with various iterations of "T AS IN TANGO"... "did you say M as in mango?". By the end I still was not confident that they'd accurately taken down my email.
I don't think AI accent-altering would have fixed this exchange.
They sounded a tinge strange, like they’ve almost crossed the uncanny valley, only to succumb at the final 3% stretch.
I was suspicious, but their ability to understand my complex request and the relatively low latency make an LLM -> TTS or e2e voice model unlikely.
This post finally solved the mystery.
Still, they could just give the employees training to learn additional accents.
The English accents around the world were left behind with the subsets of English people were taught to be able to aspire to entry level administrative jobs.
Someone recommended this to read, not sure if anyone else has read it: https://archive.org/details/educationascultu00carn
It feels like it bears some underpinning and contextual relevance.
Oldschool callcenters often had a human! Now I "interact" with AI ...
Canada, USA, doesn't matter - if our taxes subsidize a market or entrench a player within an important market (telecom, physical infrastructure, etc), they should be mandated to keep the money in local economies.
I'm American, and find it deeply offensive if a company wants to offshore despite getting tax breaks, government protections against new market entrants, etc.
I'm not paying tax money so a utility can raise prices, pay its executives more, spend more on lobbying, and outsource labor to 3rd world or developing countries. I don't give a fuck how well those folks in those countries speak English.
2) if that's not an option, have a pick-up-the-phone agent pick it up
