Slashdot Mirror
Google Executive Addresses Horrifying Reaction To Uncanny AI Tech (bloomberg.com)
The most talked-about product from Google's developer conference earlier this week -- Duplex -- has drawn concerns from many. At the conference Google previewed Duplex, an experimental service that lets its voice-based digital assistant make phone calls and write emails. In a demonstration on stage, the Google Assistant spoke with a hair salon receptionist, mimicking the "ums" and "hmms" pauses of human speech. In another demo, it chatted with a restaurant employee to book a table. But outside Google's circles, people are worried; and Google appears to be aware of the concerns. From a report: "Horrifying," Zeynep Tufekci, a professor and frequent tech company critic, wrote on Twitter about Duplex. "Silicon Valley is ethically lost, rudderless and has not learned a thing." As in previous years, the company unveiled a feature before it was ready. Google is still debating how to unleash it, and how human to make the technology, several employees said during the conference. That debate touches on a far bigger dilemma for Google: As the company races to build uncanny, human-like intelligence, it is wary of any missteps that cause people to lose trust in using its services.
Scott Huffman, an executive on Google's Assistant team, said the response to Duplex was mixed. Some people were blown away by the technical demos, while others were concerned about the implications. Huffman said he understands the concerns. Although he doesn't endorse one proposed solution to the creepy factor: Giving it an obviously robotic voice when it calls. "People will probably hang up," he said.
[...] Another Google employee working on the assistant seemed to disagree. "We don't want to pretend to be a human," designer Ryan Germick said when discussing the digital assistant at a developer session earlier on Wednesday. Germick did agree, however, that Google's aim was to make the assistant human enough to keep users engaged. The unspoken goal: Keep users asking questions and sharing information with the company -- which can use that to collect more data to improve its answers and services.
Scott Huffman, an executive on Google's Assistant team, said the response to Duplex was mixed. Some people were blown away by the technical demos, while others were concerned about the implications. Huffman said he understands the concerns. Although he doesn't endorse one proposed solution to the creepy factor: Giving it an obviously robotic voice when it calls. "People will probably hang up," he said.
[...] Another Google employee working on the assistant seemed to disagree. "We don't want to pretend to be a human," designer Ryan Germick said when discussing the digital assistant at a developer session earlier on Wednesday. Germick did agree, however, that Google's aim was to make the assistant human enough to keep users engaged. The unspoken goal: Keep users asking questions and sharing information with the company -- which can use that to collect more data to improve its answers and services.
Your programmers were so preoccupied with whether or not they could, they didn’t stop to think if they should.
Or a reverse CAPTCHA.
I'm looking up to see if I have anything free on that date. While I'm looking can you please confirm the prime factorization of 28573782909827352?
I'll see your senator, and I'll raise you two judges.
Silicon Valley is ethically lost, rudderless and has not learned a thing
Wow, exaggerate much? All that may be true, mind. But using such hyperbole when voicing concerns does nothing for the guy’s credibility. He comes across as someone who has already made up his mind about SV companies a long time ago, and sees every new issue only as something that confirms his fears, as something that’s part of a bigger plot to rape the planet and enslave humanity.
If construction was anything like programming, an incorrectly fitted lock would bring down the entire building...
Their only option is to make it mimic and impersonate human voice. If it sounds very good, and there is an announcement that it is robotic, no need, but people will hang up immediately just like any other robo-call. If it sounds obviously robotic, instant hang-up. The only way it works is if they can fake it long enough to get some information, and don't let anyone know about it.
Essentially, the only value to Google is if they trick people into using it.
Comment removed based on user account deletion
In this context, it's calling a business as a potential customer. Hanging up on what from all appearances is a potential customer is a good way not to have a job anymore. In fact, even if it is not a human but is calling to arrange the business of a real human customer, it's probably still a bad move to hang up.
XML is like violence. If it doesn't solve the problem, use more.
This is an important point. Their use case is to call and make reservations or purchase services: basically do the stuff somebody might do online but with an analog interface.
I'll bet the companies in question won't really mind selling to somebody might not contact them via other means. Basically this lets them delay on building a web page or buying into some reservation service they don't care about/don't want and yet they can still interact with the digital world.
As long as the technology is *good enough* to approximate or improve on the average stoned, drunk, spaced out, partially literate, callers they might get today then it improves their lives. And if they learn to recognize the voice they can even accelerate the interaction by getting right to the point (like they would if they were talking to a person that calls them a lot).
The issue of deception is only a part of the actually worrisome part about this technology. The real issue is that it creates an imbalance of effort, which is exactly what we have been fighting with spam emails. With either method, minimal effort on my part can cause another person to use up a significant chunk of their time. Google is displaying this technology in a situation where that would be considered acceptable, because the outcome is profitable for the person taking the appointment, but what if it wasn't? This technology could easily be used for both parasitic purposes such as sales calls, and outright hostile purposes such as tying up phone lines with seemingly benign callers. The fundamental issue is that when a human knows they are speaking to another human, they can assume that each has a similar opportunity cost for the time spent in the conversation. A machine has no such costs, and the transaction is inherently lopsided. I think this is what the push for identification stems from, the basic need for the two parties to be on (relatively) equal footing.
Because presumably places of businesses get obviously recorded spam calls like everyone else, and for the moment no one is going to assume what sounds like a recording is going to be interactive and/or on behalf of an individual customer, and will try to keep the line open/save time by hanging up before they even hear enough words to recognize it wants to make an appointment/reservation.
All in all, impressive as it sounded, this is one area I can't understand why I'd need Google to take care of it for me. In the time it takes for me to tell google to do this, I could have just called the restaurant or whatever.
XML is like violence. If it doesn't solve the problem, use more.
Forget "uh" and "umm." I want to see the AI that can competently use "Eh" like a Canadian.
Like this https://www.youtube.com/watch?...
In this case, the business does not care about 'communication', they care that the call is a legitimate request to buy goods or services. Today a robocall is 0% chance, but this would represent a probable transaction.
Sure, it's creepy. Sure, I'd rather it be blatantly obvious what it is. However, they do have a legitimate concern that an obviously artificial behavior will get terminated by someone mistaking it for spam (no, even if it is by machine, a request for a reservation is not spam).
Now it can more naturally react and say some sentence containing the phrase 'google assisstant', which would be clear as to what it is, but doesn't have to sound intentionally synthetic in the process.
XML is like violence. If it doesn't solve the problem, use more.
> Thank you for calling Bill's Hair Salon, this is Mike. That's all that's needed in a lot of cases to identify people in a small business environment.
Oh my. Someone calls a company and finds out Mike works there. I notice that the caller didn't need to say anything to get Mike to divulge that super-secret private information, so whether Google Assistant sounds human or not is irrelevant.
And, as I said, if this "Mike" is stupid enough to hand over personal information to a random caller, whether or not the voice he hears sounds human or not, then the problem is much deeper than Google Assistant having a human-sounding voice. If this "Mike" is not authorized to handle ordering and payments and I own the shop, if "Mike" hands out my credit card info in response to any request from a cold-caller for ANY reason he's fired, period, get your ass out of my shop. If "Mike" is the one who makes the payments and he hands out the card info just because someone asks for it, then once again, he's fired. Notice the common element here. Hand out private information to unknown callers, you're fired.
This has nothing to do with the type of voice - you missed the point entirely.
Of course it has to do with the type of voice. That's the HORRIFYING part of this whole process. Google's AI can insert "ums" and "uhhs" and stuff and make itself sound like it is a real human. THAT'S the horrifying "uncanny AI tech" being talked about.
It's that Google has gone one step further and is now contacting potentially non-Google users without their consent.
Google isn't contacting anyone. It's PEOPLE who use GOOGLE ASSISTANT to make the contacts. Google isn't cold calling anyone, and even if they were, then the point I made about stupid people handing out personal information to random callers still applies.
But it isn't GOOGLE that's doing this. It's Joe Smith telling his Google Assistant to make a call ON HIS BEHALF to someone that he does not need consent to call in the first place. I don't need "consent" from a barbershop to call them up to make an appointment. It's stupid to think you do.
You DO know what Google Assistant is, don't you? I guess not. It's not a Google program to randomly call people and collect data. It's that stupid, annoying piece of crap that keeps popping up when you long-press the middle (or only) button on your Android device in more recent versions of Android, offering to help you do stuff that most people can do better themselves. Mine is turned off, so instead of Google Assistant popping up directly, I get a notification that says "Your Google Assistant is ready to help you get things done. TURN ON". THAT'S what the "horrifying uncanny tech is. This Google Assistant thing is now going to be able to make phone calls that imitate a real person's mannerisms (not a specific real person, by the way). It will now be able to send emails that look like they were written by a human. That's HORRIFYING.
Yawn. If someone asks their digital assistant to do something, then it's the person who asks for it to be done that is responsible, not GOOGLE, and it is no more horrifying to have to deal with that than with a real person.
Telemarketers already have this technology. I get a lot of crap calls that start out with a perfect human voice asking if they are speaking to me, by name. Unless you typically answer "rutabaga" instead of "yes" or "no", you don't know it's automated right away. Or maybe answer "who is calling?". Once they get the initial answer they continue in a perfect human voice. It's not new. It's not even particularly horrifying. It's annoying, but it's really no worse than a recognizable robotic voice. They don't get anything more than they would otherwise.