All very fancy sounding but how does it handle a Geordie accent or Brum or Glesga?
It's all fun and games until someone has an accent.
Google has teased new bot technology aimed mainly at contact centres as part of its Cloud AI week in the seemingly unending Cloud Next OnAir videofest. This includes the ability to train text-to-speech on a human voice so that you could create recordings that sound as if a specific person spoke them – a technology with obvious …
This post has been deleted by its author
Sadly I had a human fail the ‘name a stripey horse if you are not a bot’ test.
Hmm ... do I say "Zebbra" or "Zeebra", and what inference will be drawn from my pronunciation? What will it tell the questioner about my age, education, and social background?
It's not that easy after all ... best play dumb and avoid the issue ...
We get phone calls at work and home continually, they sound just like a real person, their first words are, "Do not hang up the phone this is an importan"- at that point I hang up on them every time. So I expect this new technology will become widespread to promote insurance and credit card fraud - this is a big market, you can see why Google is interested.
This post has been deleted by its author
I'm not convinced that talking bots will be much better than text/chat bots are at present, probably just more infuriating to deal with when you ask something that is not in its database. At least human script jockeys pass you on to a supervisor when they don't understand.
I updated several email addresses recently and updated my accounts with most organisations via their websites with the exception of a couple. One being my energy supplier (Shell Energy) who had no facility to update email address. I clicked "contact" expecting a means to contact a human being either by telephone or email but instead got a text based chat bot. It half grasped that I wanted to update something and gave me a link to the page where you can update your phone number but nothing else. I tried to tell it the information it gave me was incorrect but of course that went right over its head. I asked it if I could speak to a real person but apparently not. So it is effectively impossible for me to contact them or to get my email address updated. I added a few four letter words to the chat bots dictionary and gave up.
Every month Shell Energy sends me an email with the link for me to submit that months gas and electricity meter readings. That email will now bounce back to them as undeliverable and consequently I won't be sending them any more meter readings.
According to the article, Gartner says by 2023, 70% of customers will prefer a robot so you must be in s minority.
Of course it is possible that Gartner was naked from the waist down when they pulled that statistic from its storage.
The only thing that pisses me off more than a bot is when companies fob you of with a link to a help forum that is usually no help at all.
I class that as a hindrance not a help.
I have yet to see an useful ''chatbot.'' In the last few months I have had , in desperation, to revert to trying out the respective option on Ryanair and SP Energy websites due to there being no other way of finding what I wanted or indeed a sensible method of contacting a bloody human. Nothing you put in the text box made sense to the algorithm.
OK I expected it with Ryanair as I wanted money back from them so no fucking chance that was going to work but with SPE it was putting £20k plus their way for an upgraded supply so I would have thought that ''I want to give you money for something'' would have led to a pretty good response.
Previously in the last year I have had the same while trying to spend money with Samsung, Dell and ( in desperation ) Carphone warehouse. The end result of which none of them got my money as they couldn't be arsed to provide a human to sort out a few basic queries.
This is still roughly the same level it was years ago so what is the problem and where is the bottleneck ?
I never had any luck trying to buy a custom spec computer system from Dell. Years ago when I first tried I had a query but they went out of their way to avoid human interaction. Finally found an email address for them and I got a reply back from a bot with a 16 page reply with various possible answers to my question; none of which answered it and some were quite laughable. They didn't get my order. I though to myself "If this is their pre-sales support I wonder what their post-sales support is like?"
A few years later while living in France I wanted a custom built computer with UK Windows, keyboard etc. Went right through all the specification details blah, blah, blah and a couple of hours later "Sorry we can't deliver to France." Went through the process again with Dell France but couldn't have a computer with UK Windows and keyboard. I gave up on Dell in the end. They just aren't customer oriented.
This post has been deleted by its author
My interaction with google over a week.
Monday
Google : Are you going to work.
Me : Yes.
Google: Please wait.. Turn Left onto....
Google : Are you going home.
Me : Yes.
Google: Please wait.. Perform a U-turn....
Tuesday
Google : Are you going to work.
Me : Yes.
Google: Please wait.. Turn Left onto....
Google : Are you going home.
Me : Yes.
Google: Please wait.. Perform a U-turn....
Wednesday
Google : Are you going to work.
Me : Yes.
Google: Please wait.. Turn Left onto....
Google : Are you going home.
Me : Yes.
Google: Please wait.. Perform a U-turn....
Thursday
Google : Are you going to work.
Me : Yes.
Google: Please wait.. Turn Left onto....
Google : Are you going home.
Me : Yes.
Google: Please wait.. Perform a U-turn....
Friday
Google : Are you going to work.
Me : Yes.
Google: Please wait.. Turn Left onto....
Google : Are you going to the YMCA.
Me : ... FFS Google!!!!! That is -ing protected category data!!!!!
"customers will prefer to use speech interfaces "
Given that we don't have the choice, I really wonder where Google is pulling that notion from.
I don't mind being given a choice between four possibilities, as long as the last one is "TALK TO AN ACTUAL HUMAN BEING" (with a brain, preferably).
I perfectly understand that there are very common issues that can be efficiently solved by pressing button 1, 2 or 3, but if your phone talky thingy is the only choice I have to contact you, then you had damn well better have some human operators for dealing with the edge cases your bunch of stupid managers forgot to plan for.
Well there goes any form of truth. Doesn't matter about ethical review. In 5 years this will be commodity and everyone will be able to use it. I think I see a gap in the market for a type is small premises staffed by humans that are local and know you and you know them who can do things like arrange insurance or travel or banking and you won't have a problem with identity theft or false information... Can't quite think what to call this idea...
"Google's criteria for ethical use and how it will prevent misuse was not stated."
However, I have access to some documents which should make them clear. Here is their ethical review form:
Your name:
Purpose of your use of the system:
Credit card number:
How much credit would you like applied to your account (note: this credit is nonrefundable):
Urgency of your request:
And this is the misuse form letter:
We have received a notice that your usage of the system is considered misuse. Your account has been closed to prevent you from continuing to misuse. If you wish to continue to use these features, please submit a request for appeal within your account dashboard. We ask you not to create a new account as we are currently unable to process appeals between accounts for privacy reasons; information connected to one account cannot be attached to another one. [Wink wink]
I recognise as tecchies, we probably try a bit better to fix things ourselves, so by the time we're calling the telephone line, we're well past anything the automated systems can handle.
I don't like it, but I can see the usefulness in running triage on the incoming calls: It's the next step on from the automation bot that tries to point you at a KB or FAQ before letting you talk to a meatsack.
Triage is just one more step in reaching a resolution, as providers of a service, anyone should have a duty of care to resolve user's and customer's problems.
What should be in place given the enormous size and profits made by these people, is a team of adequately trained humans who, if they are unable to help you can refer you to the next stage.
Ninety percent of this is about reluctantly providing customer service when all the companies want is for you to give them money and then fuck off until the next time you want to give them money. Hence the usual 'Frequently asked questions' as the entrance to a rabbit hole full of futility.
And just why are Google developing this ? They have no intention of using it. Machine or human, you cannot get a response from Google in any case, but then most of the people I see who cannot get any help for Google problems are, of course, not customers, they are part of the product.
Chris Cosgrove
If they can get it good enough to convince someone with a big call center, the business will fire half the call center workers and put that money Google's way. Then the businesspeople who heard the demo won't ever call said call center, so they'll never hear how well or badly it does in real usage. Then Google can show the same convincing demo to other businesses and cite that one as a proven success story.
This post has been deleted by its author
- Google are useless at human interaction; they go out of their way to automate absolutely everything they can.
- And when they do, their AI sucks at it (eg YouTube moderation or voice assistant).
Given those two facts, it's time people started calling bullshit and telling them where to go, before they help wipe out even more jobs and suck even more money back up to the billionaires.
The whole automation-for-scale thing has been shown to be impossible for them to do properly, and it is rotten rent-seeking to it's core.
'We're not claiming to replace humans,' says Google, but we want to be 'close enough' that you can't tell it's a bot talking
Methinks that hurdle and threshold was passed quite some time ago, with the one being too often enough so easily mistaken for the other to not already be an ACTive tool, both virtual and practical, for more general use and greater exploitation.
Indeed, El Reg may even host evidence of such a stealthy advance and silent enhancement ....... https://forums.theregister.com/forum/all/2020/09/04/linux_kernel_flaws/#c_4103006
Google may just be late to the party and trailing way behind in trials of that particular peculiar utility/facility/discipline. The real risk that they run inviting is their being just too late to the party to be in a way major instrumental in leading its future direction .... and thus always to be following and searching for novel news about recently discovered views and readily available revolutionary abilities.
In the future, my bot will phone in, repeating the words, "I need to speak to an agent, a representative" while on speakerphone, and when a harrassed sounding voice picks up, I will pick up. If I have had to phone the company, I will have a problem sufficiently arcane that I have extirpated all other attempts to ameliorate it, and is often the result of some guddling about that the company has done previously to try to make the issue "better", entangling it further. My vocabulary is extensive, and I have ferreted out chatbots by employing it. They cannot contextualize if enough polysyllabic words are distributed throughout.