This would be terrible. I can just about get by with my Google Home now that I know the subset of commands it'll reply to coherently, but I can't imagine the frustration of it being my only option for actually getting something done.
Also, I know this is a tedious point, but if all the call center employees lose their jobs we'll have a mini crisis on our hands.
This is "not applicable", yet the top comment just causally is "putting aside the Automation/Job Replacement fears".
Working in a call center is basically the last easy to get job for people who aren't very intelligent, it's the one job where a physically handicapped but attentive person can get treated like a normal human being. And they all just get tossed in there, "trained" by people who don't know their ass from their elbow either. They don't exactly have it good as is.
But let's "put that aside", because they are not eloquent enough to post here, and the few who think of them in earnest can be easily ignored. And then double think and "struggle" to think how putting someone else's suffering aside could have anything to do with this.
> If machines produce everything we need, the outcome will depend on how things are distributed. Everyone can enjoy a life of luxurious leisure if the machine-produced wealth is shared, or most people can end up miserably poor if the machine-owners successfully lobby against wealth redistribution. So far, the trend seems to be toward the second option, with technology driving ever-increasing inequality.
I can. Just imagine someone talking about getting icecream and giving them the quote in response.
After all we're talking about job loss, which can and probably will hit most of professions at some point in the future. The gp seems to want us to speak out about that? Maybe like this:
> First they came for the Loom Weavers, and I did not speak out —
> Because I was not a Loom Weavers.
> Then they came for the Factory Line Assemblers, and I did not speak out —
> Because I was not a Factory Line Assemblers.
> Then they came for the Truck Drivers, and I did not speak out —
> Because I was not a Truck Driver.
> Then they came for me — and there was no one with Work (Influence) left to speak for me.
where 'they' is either the robots themselves or the ones controlling the robots, and work is inherently worthy or necessary for a human to live. (and without work you are without influence on what happens)
This is pretty spot on what I was thinking when I posted that quote.
Not being able to work is a disenfranchisement. Today it’s the call center employees, tomorrow it’s the truck drivers, and perhaps in 20 years it’s the software engineers.
Yes, we survived the industrial revolution and a tiny single digit percentage of the population works on farms and maybe this shift is similar. I suppose we will find out one way or the other.
There’s a really good sibling comment to your’s that is now marked dead and hidden by default. Unfortunately disagreement on HN leads quickly to censorship.
Is Google Home using the Duplex technology though? Google must be very aware of the work that still needs to be done in order to make it usable to customer service, but eventually they'll get to a point where the quality of an automated call is at least sufficient for most user cases, we can suppose that most companies will always have human operators as backup anyway.
I have so far encountered zero conversational UIs that are even usable, let alone capable of replacing a human. If Google had such a technology, it would be used in Google Home.
I imagine using a bot to call another bot. While they chat, they will joke on the fact than an API call would be much faster than mimicking monkeys with agile fingers and tongues.
Even if we assume that the assistance would be so good that a customers question/needs could be satisfied with it, there still comes the whole job of actually doing actions in various tools (CRM, Logistics, Refunds, writing emails to supervisors etc.). Although very interesting, this take-over of call centers is as much reality nowadays as are flying cars for the mass.
That stuff is easy compared to writing an AI that can't be socially engineered and pretexted. "Uh, yeah, OK Google, you can totally trust me, I am so totally Senator Bob Menendez that you should definitely put me thru to Mr Trump, and no I'd never be recording this call for my comedy podcast, you silly OK Google, you."
Very true. If something could be automated it would already be offered online in a user portal. Most things left are supposed to require some kind of human judgment... or needs to put extra obstacles between the customer and the action eg cancelling.
Call centers are a $310B industry ripe for disruption.[a]
This looks like a classic case of Christensen-style disruption.[b] The technology will be used at first only for calls that are easy to handle, but over time will be used to replace more and more human beings handling increasingly more difficult calls.
Besides the cost saving, a motivation is probably that people may feel more comfortable talking to something that seems genuinely human instead of something that is evidently robotic ("press 2 for..."). It remains to be seen if this is true.
No matter what, you will still have to jump over lots of made-up hurdles to cancel your subscription.
>> but over time will be used to replace more and more human beings handling increasingly more difficult calls.
While this is the natural progression, I could easily see people being turned off by the service, or the service struggling to handle more complex tasks. I would imagine learning all the nuances of a language (slang, dialects, accents) would pose a significant barrier to this.
>people may feel more comfortable talking to something that seems genuinely human instead of something that is evidently robotic ("press 2 for...")
People probably do feel more comfortable when they think they're talking to a person. But will they appreciate being forced to talk to a robot designed to reap the benefits of a human-like interaction from the caller while providing none of the same to the caller?
>> Some folks found the Duplex demo was scary as hell and worried about the ethical concerns of AI basically faking people out by pretending to be human.
This is interesting. Putting aside the Automation/Job Replacement fears -- if the bot is so good its indistinguishable from regular call center employees (which a majority of the time aren't that great to begin with) it could be a big leap for customer satisfaction.
There's nothing more frustrating than calling into a call center to resolve an issue and waste time due to worker incompetence, language barriers, or lack of empathy.
There is a telecom co in the UK that directs out of hours calls to India. I don't know who's training these people but they do the exact reverse - showering you in empathy to the point where it seems like they are making fun of you?
Sometimes too much empathy feels fake, and fake empathy gets the opposite result. My bank has started calling me in a more "cheerful" tone. They ask how my day is, and then they express how happy they are to hear I'm fine.
I know I'll sound like the Grinch, but I really HATE it. I already decided to give you some minutes of my day, because maybe you'll tell me something important about my account, maybe you'll offer me something interesting, even though 80% of the time you'll offer something I don't really want, so we'll waste each other's time. But we won't know until you finish pretending you are interested in my day.
The funny thing about call centers is that customers hate them, even when you are good at providing the service. Well regarded, customer focused companies like American Express offer best in class service -- but something like 60% of customers still don't want to talk to them and prefer email or web.
I think Google has an opportunity to do something interesting here, where you basically offer a voice API for humans to interact with. If they can scale it from conversational interactions down to "shortcut" command interfaces and drop the stupid dance that you do when initially calling, that would be an amazing thing.
The problem with call centers, as an ex-IT guy for several of them, is not the people, but the interface they have to the company behind them.
People expect call centers to be able to do anything. Companies hiring call centers give them 2 or 3 options in a clunky web interface. Or worse, access to a ticketing system. Management does not want the call center to be able to do things, or wants certain metrics (no more than 10 unsubscribings per month - that sort of thing), and this then gets translated in "incentives" for the people answering the calls.
Calling up lots of companies is just a hassle, and you get trained to dislike it due to all the negative experiences.
My best support experiences have been with Apple. I've had to interact with them on the phone a few times and it has always been a highly positive experience. They don't give you any crap or try to sell more stuff.
However it won't exhibit the "please wait, there are 324823 callers in front of you, please listen to this recording of my daughter's cat singing 'let it be' for the next year or two".
Are you sure? This behaviour is actually desirable because the shitty company behind it enjoys the fact that they get less calls as a result.
This problem could’ve been solved for ages without Duplex - simply call, they remember your call and call you back when an operator is ready for you. The reason it’s not used is because shitty companies don’t want this.
The cynic in me hope that Google charges through the roof for this - if barrier to operate a Duplex robocall center is very low, I suspect we'll get bombarded by sales calls from Duplex centers. Not that anyone really answers the phone nowadays though :)
For context, here in sweden it's more or less not allowed to cold call people if they are on a "don't call me"-list, unless it's part of an ongoing or previous relationship, so my ISP can call me for upselling.
Similar here in the UK but it doesn't appear to put off certain companies. Some calls are just recordings pretending to have a conversation (which I'd be surprised if anyone fell for) and others route you to a human. They will never divulge the name or details of the company they represent, I can't imagine their success rate is very good.
I worked as tech support in a call center years ago. In my experience, the biggest barrier to customer satisfaction was not what I did, but the many conflicting and antagonistic rules I had to follow. The only way that can be improved with a bot is if whoever implements the bot has more leverage to point out these problems.
Feel empathy from a machine? The moment anyone realizes or finds out that the "relationship" they just built or experienced over the phone was programmed, any feelings of connection or empathy get tossed out.
The ethical concern isn't about the AI replacing jobs, it's about something calling me and disingenuously posing as a human being. If your machine calls me, it needs to announce that it is AI. I am not interested in replacing humans or human connection.
> it could be a big leap for customer satisfaction.
Exhibit A of everything that is wrong with consumerist, material societies. Everyone is a customer in need of satisfaction in the form of products and services.
Lack of empathy can be dealt with when expected. It also makes a difference when lack of empathy isn't coupled with sarcasm, exasperation, or boredom.
>>Everyone is a customer in need of satisfaction in the form of products and services.
The point of contacting a call center is because you are unsatisfied as a customer. Customer satisfaction is the only factor in a capitalist society. There is no alternative. If customers are not satisfied, you don't get business.
You can argue until you're blue in the face about the evils of consumerism, but it's pretty irrelevant to this discussion.
Unique, inaudible fingerprints for synthetic voices would make it trivial to alert people that they are not talking to a person, and might enable some useful optimizations for ASR.
While it will no doubt be somewhat more competent and have a better grasp of english, do you see a computer as being more empathetic? Here is my favorite lack of empathy quote from a computer:
I'm sorry Dave, I'm afraid I can't let you do that.
Perhaps instead of replacing the call center itself, it could eliminate long IVR sequences, gather better and more pertinent data from the customer and make the humans on either side more productive. As an example, in India you are often asked to press 1 for language X, 2 for language Y. But a smart assistant can just listen to the customer and decide this automatically. This can be a huge win by itself.
> But a smart assistant can just listen to the customer and decide this automatically.
But the customer will not start speaking before the system prompts them to speak or asks them a question. And since it has to do that in one fixed language, the choice of that language will usually determine what language the customer responds in; this is in the Indian context where most people calling into an IVR will be bilingual at least, speaking English and one regional language.
Many call center calls have a negative or zero financial value to the company, which is why the experience can be so horrible now. In the future, rather than wearing down the patience of a human in India to cancel service, you'll have to wear down the patience of "ok google" or "alexa"... good luck with that.
A not entirely sci fi book plot would revolve around indentured servitude (to put it nicely) what does society and the economy look like after a few generations of only being able to cancel a service if you have more patience and more computing cycles than "the cloud"? As a sci fi plot, what does the world look like when its easier and more likely to kick a heroin addiction than to cancel an old gym membership or cut the cable TV service, or request a credit card chargeback?
In that scenario people wouldn't sign up for recurring services, or it would end up being like the new law in Cali where they have to let you unsubscribe on the internet.
Also having a robot repeating: "my customer number is blabla and I would like to cancel my account", "has my account been canceled yet" doesn't exactly take a lot of cycles.
In fact it might be worth it to create cancelation as a service bot, where for a small fee, you can get the bot to call and cancel for you.
Why do IVRs universally suck? They seem incapable of transferring information you already entered and blare obnoxious advertisements every 12 seconds while on hold. It's like the designers of these systems had "make it as painful as possible" in the requirements.
Superficially a high up enough executive would see their business as "accept money, emit 'food'". However that does not work with bartenders or womens hair dressers.
My guess is there's great piles of money to be made in the near future not by automating personality-free transactions, but by convincing people to cough up extra money for personality-added transactions. Note that personality-added transactions are not necessarily in opposition to tech in general, only in opposition to automation. My wife's hair cut lady accepts all sorts of crazy payment processors and social media interaction (I should ask if she takes bitcoin... yet)
Likewise today, getting a fast food burger TODAY is all about turning the personality dial down to zero, if not negative, such that automation is a huge threat, but I suspect in a decade getting anything beyond a glorified vending machine will look like the personality dial is turned up to 10. The future being very unevenly distributed, look forward to the McDonalds dude who flips your burgers in ten years having a similar or larger social media presence in your life than a Food Network celebrity chef has today, or maybe it'll be like "food truck groupies" behave today, but for all prepared food. I'm not sure I want my best social media friend to be the produce manager at my local independent supermarket, but what people get usually has little to do with what they want (and both are usually orthogonal to what they need, LOL).
It's going to be a really weird world when the first thing out of the DMV clerk's mouth is "Have you subscribed to my Instagram yet?"
Since when do you actually have to go to the DMV to get anything done? Aside from the first license (which requires a photo) most things that I can think of can be done online?
And if anyone ever asks me if I have subscribed to their Instagram, I will just say "No."
Add truck drivers (and associated truck stops, restaurants, etc); food packaging and (most of) farming (once we have lab grown meat) retail (already happening); manufacturing and coal mining and there isn't really any jobs left that doesn't require a college degree and few enough of those outside of major cities.
You know that's a good question. I feel like I would want to know though. Another thought I had was some sort of Machine learning Caller ID/CAPTCHA that just emits some garbage that only a bot would be able to dissect. I just don't want to spend my time talking to an API.
There’s going to be miles of text on forums written about how to quickly ask a question or play a tone to test for humanity. It’s going to be a whole new dystopic layer to customer service!
I immediately expected the logical extension of Google Firebase App Indexing, so theoretically having an "OK Google" conversation would have API hooks into the back end of call centers such that I could "OK Google" my way thru reporting a cable TV outage or disconnecting my cable TV or making a doctor appointment or whatever. Instead, its almost the logical reverse of giving the call center a "OK Google" instead of what I expected of giving my "OK Google" a call center, or access to the back end of a call center.
I get angry when I'm forced to speak to anyone/anything to do something that I could have done via a form on a website. Customer service should not be a gatekeeper to changes on my account.
If I'm calling in its because their automated systems can't handle my request. What I don't want is another automated system being unable to handle my request.
What is Duplex, but a voice version of a chatbot. We have had to endure chatbots for ages and we know for a fact that they suck. Why would Duplex be any different?
Also, Google's entire customer support strategy is "let he computers handle it, no humans involved." And it is universally loathed and derided. Why should we assume they are going to offer something to outside corporations that is better than what they use themselves?
It doesn't have to suck, but all evidence points out that it will.
I called AT&T to signal a phone outage for somebody and they had this robot on the line asking questions, me answering them, then you heard the robot's typing sound to give the impression that you were talking to a human.
Thats not to convince you that the thing you're talking to is a human. It's to let you know the call hasn't dropped. People on phones get weird when there is dead air. So instead of it just sitting silently not saying anything while it looks up your account number and cross references it with other various backend systems it play some familiar analog.
People are mentioning that they'd be angry to find out they were talking to an AI. I see people interacting with Google Assistant and Siri in a positive way all the time though, acting amused or amazed when good responses are given. As long as the person knows it's an AI and the responses are good, I think most people will act positively with this.
I, like everyone, hate telephone support. I don't like it mostly because of it's lack of effectiveness, and the very strong impression it leaves that the company doesn't care at all. I'm also one of the people that finds Duplex deeply disturbing (Does anyone not find it disturbing?! That's even more disturbing!)
However, I can see this working quite well for help desk services if the paradigm is flipped. That is if I end up speaking with my electronic assistant who then appears to negotiate on my behalf with the company. As apposed to me calling an unknown entity. It may in reality be the companies service, or even infrastructure that my electronic assistant is running in, but the impression to me would be I'm speaking to my already familiar electronic assistant.
The benefits of this are many, and only possible with something like Duplex. In addition to familiarity, technical support can be delivered at the right customer knowledge level skipping past insulting time wasting support like "is it plugged in?". Also a "continuity of care" model can be adopted by companies to provide a continuation of the relationship in the context of prior interactions. The list goes on.
This might in fact be the one service that is a bad enough experience, that I would gladly welcome our new robotic overlords.
> This might in fact be the one service that is a bad enough experience, that I would gladly welcome our new robotic overlords.
I think this does not solve the underlying problem. Sorry if this is out of scope but lets talk about the recent blender on YouTube thing. Often, it is not that the person answering the phone is an idiot but rather the person answering the phone does not have the authority to do anything. I've received bad bills from T-Mobile where the process goes like:
1. see bad bill in the mail 2. Call T-Mobile and wait 3. Get a person and explain the situation 4. Person looks and agrees that line item shouldn't be there and removes it. Person also says they put a note to prevent this from happening again. 5. Rinse and repeat.
I refuse to believe that multiple people over the course of a year are all incompetent. I must conclude there was either something intentional (bill stuffing?) or the staff was not empowered to fix the situation. Will we empower our new robotic overlords to "fix" the situation? How far will business capitulate on business rules? What happens when something is not right and needs an override? Previously on HN https://news.ycombinator.com/item?id=17350645
> I refuse to believe that multiple people over the course of a year are all incompetent
Those are your words, homie. The parent simply called out call-centers for being ineffective, which I would agree, given my own experience with them. I would contend that the workers are generally skilled and agreeable, rather the hop-skip-dance that needs to take place every time a call-center is used, is painful and ancient. Imagine a situation where your own digital PA already knows your account number, address, phone, etc., allowing you to simply provide a courteous hello before asking your question.
> I'm also one of the people that finds Duplex deeply disturbing (Does anyone not find it disturbing?! That's even more disturbing!)
Why? You presumably exchange text with chatbots all the time, and get mail/slack etc updates from your CI bot as well. Do they bother you?
you get phone calls / talk to people in shops who claim to have your interests in mind when they steer you to one product over another. Does that disturb you?
Oh now I get it, the people who are not disturbed by this contemplate that it'll only be used in innocuous contexts in which we already accept an unknown voice. That's not intended to sound sarcastic, I genuinely mean that now I understand the reasoning.
However, the reason these are not comparable is proof of work/investment. Presumably Duplex can or will soon be able to make millions of calls for any purpose whatsoever. Unlike a human being with a foreign accent perhaps or other alerting auditory factors Duplex massively reduces the cost of experimental manipulation.
One of the most successful hacking techniques is phishing. Its success is directly tied to how compelling it is at counterfeiting a legit communication.
And this does not even touch on larger institutional deception.
It sounds like your concerns are based on what some eventual form of this technology might become. Duplex is incredibly domain-specific, and will only make calls on behalf of Google, not random phishers.
Being worried about Duplex in its current form is like being worried about Siri becoming conscious. Your concerns about malicious AI voices are valid, but that eventual technology will be nothing like Duplex in its current form, and it's still a long way off
The outcomes I speak of, which are nothing compared to all potential exploits, are possible with the existing Duplex technology, in that they do not happen it is only in Google's good gatekeeping. That is too thin a standard for me.
Duplex only works because it was highly optimized for specific domains, like booking haircuts and booking tables. Every conversation in these domains pretty much goes the same way. If you go off the standard reservation script, it falls apart.
I suppose Duplex could maliciously attempt to book a bunch of fake reservations, but the technology behind it really can't do much more than that
Also, I'm not really sure what you mean by "Google's good gatekeeping". 3rd parties have no control over Duplex, other than initiating specific tasks through Google Assistant. If Duplex does anything malicious, it would be because Google went rogue, and there would obviously be massive legal and PR consequences for that.
As someone who runs the budget for my company's call center, I can't wait until this actually happens. Huge savings. Although this probably won't happen any time soon.
So, the company notorious for "Your account has been banned by AI. No revenue for you. No appeal. No explanation." is going to tell the rest of the world how to improve customer service? This is too rich. The hubris is astounding. Does anyone else see this as totally tone-deaf?
So Google is going to data mine your relationship with your customers, find out what's broken, then go build a tool to compete with you and put you out of business.
I can imagine it being very frustrating interacting with an AI for most support issues, but I see a real use case in replacing those gigantic options trees with a semi-interactive AI.
Just a voice saying "Hi, I'm a bot, can you describe your issue?", followed by "Okay, sounds like a whizmo issue, can I transfer you to our whizmo department?" seems potentially both faster and more accurate than a huge options tree. That's a win even if the AI can't directly resolve your issue.
88 comments
[ 6.0 ms ] story [ 187 ms ] threadAlso, I know this is a tedious point, but if all the call center employees lose their jobs we'll have a mini crisis on our hands.
Because I was not a Socialist.
Then they came for the Trade Unionists, and I did not speak out—
Because I was not a Trade Unionist.
Then they came for the Jews, and I did not speak out—
Because I was not a Jew.
Then they came for me—and there was no one left to speak for me.
https://en.m.wikipedia.org/wiki/First_they_came_...
This is "not applicable", yet the top comment just causally is "putting aside the Automation/Job Replacement fears".
Working in a call center is basically the last easy to get job for people who aren't very intelligent, it's the one job where a physically handicapped but attentive person can get treated like a normal human being. And they all just get tossed in there, "trained" by people who don't know their ass from their elbow either. They don't exactly have it good as is.
But let's "put that aside", because they are not eloquent enough to post here, and the few who think of them in earnest can be easily ignored. And then double think and "struggle" to think how putting someone else's suffering aside could have anything to do with this.
> If machines produce everything we need, the outcome will depend on how things are distributed. Everyone can enjoy a life of luxurious leisure if the machine-produced wealth is shared, or most people can end up miserably poor if the machine-owners successfully lobby against wealth redistribution. So far, the trend seems to be toward the second option, with technology driving ever-increasing inequality.
-- Stephen Hawking
There is no "struggle". There is no "try." Think.
After all we're talking about job loss, which can and probably will hit most of professions at some point in the future. The gp seems to want us to speak out about that? Maybe like this:
> First they came for the Loom Weavers, and I did not speak out —
> Because I was not a Loom Weavers.
> Then they came for the Factory Line Assemblers, and I did not speak out —
> Because I was not a Factory Line Assemblers.
> Then they came for the Truck Drivers, and I did not speak out —
> Because I was not a Truck Driver.
> Then they came for me — and there was no one with Work (Influence) left to speak for me.
where 'they' is either the robots themselves or the ones controlling the robots, and work is inherently worthy or necessary for a human to live. (and without work you are without influence on what happens)
Not being able to work is a disenfranchisement. Today it’s the call center employees, tomorrow it’s the truck drivers, and perhaps in 20 years it’s the software engineers.
Yes, we survived the industrial revolution and a tiny single digit percentage of the population works on farms and maybe this shift is similar. I suppose we will find out one way or the other.
There’s a really good sibling comment to your’s that is now marked dead and hidden by default. Unfortunately disagreement on HN leads quickly to censorship.
Call centers are a $310B industry ripe for disruption.[a]
This looks like a classic case of Christensen-style disruption.[b] The technology will be used at first only for calls that are easy to handle, but over time will be used to replace more and more human beings handling increasingly more difficult calls.
Besides the cost saving, a motivation is probably that people may feel more comfortable talking to something that seems genuinely human instead of something that is evidently robotic ("press 2 for..."). It remains to be seen if this is true.
No matter what, you will still have to jump over lots of made-up hurdles to cancel your subscription.
[a] https://www.quora.com/How-many-call-centers-are-there-in-the...
[b] https://en.wikipedia.org/wiki/Disruptive_innovation
While this is the natural progression, I could easily see people being turned off by the service, or the service struggling to handle more complex tasks. I would imagine learning all the nuances of a language (slang, dialects, accents) would pose a significant barrier to this.
People probably do feel more comfortable when they think they're talking to a person. But will they appreciate being forced to talk to a robot designed to reap the benefits of a human-like interaction from the caller while providing none of the same to the caller?
This is interesting. Putting aside the Automation/Job Replacement fears -- if the bot is so good its indistinguishable from regular call center employees (which a majority of the time aren't that great to begin with) it could be a big leap for customer satisfaction.
There's nothing more frustrating than calling into a call center to resolve an issue and waste time due to worker incompetence, language barriers, or lack of empathy.
There is a telecom co in the UK that directs out of hours calls to India. I don't know who's training these people but they do the exact reverse - showering you in empathy to the point where it seems like they are making fun of you?
I know I'll sound like the Grinch, but I really HATE it. I already decided to give you some minutes of my day, because maybe you'll tell me something important about my account, maybe you'll offer me something interesting, even though 80% of the time you'll offer something I don't really want, so we'll waste each other's time. But we won't know until you finish pretending you are interested in my day.
I think Google has an opportunity to do something interesting here, where you basically offer a voice API for humans to interact with. If they can scale it from conversational interactions down to "shortcut" command interfaces and drop the stupid dance that you do when initially calling, that would be an amazing thing.
People expect call centers to be able to do anything. Companies hiring call centers give them 2 or 3 options in a clunky web interface. Or worse, access to a ticketing system. Management does not want the call center to be able to do things, or wants certain metrics (no more than 10 unsubscribings per month - that sort of thing), and this then gets translated in "incentives" for the people answering the calls.
My best support experiences have been with Apple. I've had to interact with them on the phone a few times and it has always been a highly positive experience. They don't give you any crap or try to sell more stuff.
As if Duplex wont exhibit all of these.
This problem could’ve been solved for ages without Duplex - simply call, they remember your call and call you back when an operator is ready for you. The reason it’s not used is because shitty companies don’t want this.
For context, here in sweden it's more or less not allowed to cold call people if they are on a "don't call me"-list, unless it's part of an ongoing or previous relationship, so my ISP can call me for upselling.
Feel empathy from a machine? The moment anyone realizes or finds out that the "relationship" they just built or experienced over the phone was programmed, any feelings of connection or empathy get tossed out.
The ethical concern isn't about the AI replacing jobs, it's about something calling me and disingenuously posing as a human being. If your machine calls me, it needs to announce that it is AI. I am not interested in replacing humans or human connection.
> it could be a big leap for customer satisfaction.
Exhibit A of everything that is wrong with consumerist, material societies. Everyone is a customer in need of satisfaction in the form of products and services.
>>Everyone is a customer in need of satisfaction in the form of products and services.
The point of contacting a call center is because you are unsatisfied as a customer. Customer satisfaction is the only factor in a capitalist society. There is no alternative. If customers are not satisfied, you don't get business.
You can argue until you're blue in the face about the evils of consumerism, but it's pretty irrelevant to this discussion.
I'm sorry Dave, I'm afraid I can't let you do that.
What would the benefit be? Of course the customer will gain a few seconds but I don't see a large improvement.
But the customer will not start speaking before the system prompts them to speak or asks them a question. And since it has to do that in one fixed language, the choice of that language will usually determine what language the customer responds in; this is in the Indian context where most people calling into an IVR will be bilingual at least, speaking English and one regional language.
A not entirely sci fi book plot would revolve around indentured servitude (to put it nicely) what does society and the economy look like after a few generations of only being able to cancel a service if you have more patience and more computing cycles than "the cloud"? As a sci fi plot, what does the world look like when its easier and more likely to kick a heroin addiction than to cancel an old gym membership or cut the cable TV service, or request a credit card chargeback?
Also having a robot repeating: "my customer number is blabla and I would like to cancel my account", "has my account been canceled yet" doesn't exactly take a lot of cycles.
In fact it might be worth it to create cancelation as a service bot, where for a small fee, you can get the bot to call and cancel for you.
My guess is there's great piles of money to be made in the near future not by automating personality-free transactions, but by convincing people to cough up extra money for personality-added transactions. Note that personality-added transactions are not necessarily in opposition to tech in general, only in opposition to automation. My wife's hair cut lady accepts all sorts of crazy payment processors and social media interaction (I should ask if she takes bitcoin... yet)
Likewise today, getting a fast food burger TODAY is all about turning the personality dial down to zero, if not negative, such that automation is a huge threat, but I suspect in a decade getting anything beyond a glorified vending machine will look like the personality dial is turned up to 10. The future being very unevenly distributed, look forward to the McDonalds dude who flips your burgers in ten years having a similar or larger social media presence in your life than a Food Network celebrity chef has today, or maybe it'll be like "food truck groupies" behave today, but for all prepared food. I'm not sure I want my best social media friend to be the produce manager at my local independent supermarket, but what people get usually has little to do with what they want (and both are usually orthogonal to what they need, LOL).
It's going to be a really weird world when the first thing out of the DMV clerk's mouth is "Have you subscribed to my Instagram yet?"
And if anyone ever asks me if I have subscribed to their Instagram, I will just say "No."
Had an Instagram account. It's useless.
Fuuuuck.
I'd be more interested in the latter.
I know I do.
Let's hope Duplex understands the (high volume, distorted) phrase "just let me talk to a human being!!" and transfers the call accordingly.
Also, Google's entire customer support strategy is "let he computers handle it, no humans involved." And it is universally loathed and derided. Why should we assume they are going to offer something to outside corporations that is better than what they use themselves?
It doesn't have to suck, but all evidence points out that it will.
Is anyone working on this? Give it a script and an objective and test it against robot and human callcenters..
It was so slow and weird.
However, I can see this working quite well for help desk services if the paradigm is flipped. That is if I end up speaking with my electronic assistant who then appears to negotiate on my behalf with the company. As apposed to me calling an unknown entity. It may in reality be the companies service, or even infrastructure that my electronic assistant is running in, but the impression to me would be I'm speaking to my already familiar electronic assistant.
The benefits of this are many, and only possible with something like Duplex. In addition to familiarity, technical support can be delivered at the right customer knowledge level skipping past insulting time wasting support like "is it plugged in?". Also a "continuity of care" model can be adopted by companies to provide a continuation of the relationship in the context of prior interactions. The list goes on.
This might in fact be the one service that is a bad enough experience, that I would gladly welcome our new robotic overlords.
I think this does not solve the underlying problem. Sorry if this is out of scope but lets talk about the recent blender on YouTube thing. Often, it is not that the person answering the phone is an idiot but rather the person answering the phone does not have the authority to do anything. I've received bad bills from T-Mobile where the process goes like:
1. see bad bill in the mail 2. Call T-Mobile and wait 3. Get a person and explain the situation 4. Person looks and agrees that line item shouldn't be there and removes it. Person also says they put a note to prevent this from happening again. 5. Rinse and repeat.
I refuse to believe that multiple people over the course of a year are all incompetent. I must conclude there was either something intentional (bill stuffing?) or the staff was not empowered to fix the situation. Will we empower our new robotic overlords to "fix" the situation? How far will business capitulate on business rules? What happens when something is not right and needs an override? Previously on HN https://news.ycombinator.com/item?id=17350645
Those are your words, homie. The parent simply called out call-centers for being ineffective, which I would agree, given my own experience with them. I would contend that the workers are generally skilled and agreeable, rather the hop-skip-dance that needs to take place every time a call-center is used, is painful and ancient. Imagine a situation where your own digital PA already knows your account number, address, phone, etc., allowing you to simply provide a courteous hello before asking your question.
Why? You presumably exchange text with chatbots all the time, and get mail/slack etc updates from your CI bot as well. Do they bother you?
you get phone calls / talk to people in shops who claim to have your interests in mind when they steer you to one product over another. Does that disturb you?
However, the reason these are not comparable is proof of work/investment. Presumably Duplex can or will soon be able to make millions of calls for any purpose whatsoever. Unlike a human being with a foreign accent perhaps or other alerting auditory factors Duplex massively reduces the cost of experimental manipulation.
One of the most successful hacking techniques is phishing. Its success is directly tied to how compelling it is at counterfeiting a legit communication.
And this does not even touch on larger institutional deception.
Being worried about Duplex in its current form is like being worried about Siri becoming conscious. Your concerns about malicious AI voices are valid, but that eventual technology will be nothing like Duplex in its current form, and it's still a long way off
The outcomes I speak of, which are nothing compared to all potential exploits, are possible with the existing Duplex technology, in that they do not happen it is only in Google's good gatekeeping. That is too thin a standard for me.
I suppose Duplex could maliciously attempt to book a bunch of fake reservations, but the technology behind it really can't do much more than that
Also, I'm not really sure what you mean by "Google's good gatekeeping". 3rd parties have no control over Duplex, other than initiating specific tasks through Google Assistant. If Duplex does anything malicious, it would be because Google went rogue, and there would obviously be massive legal and PR consequences for that.
$DIETY save us all.
https://en.wikipedia.org/wiki/Tellme_Networks
Just a voice saying "Hi, I'm a bot, can you describe your issue?", followed by "Okay, sounds like a whizmo issue, can I transfer you to our whizmo department?" seems potentially both faster and more accurate than a huge options tree. That's a win even if the AI can't directly resolve your issue.