I would have thought so too, but check out https://news.ycombinator.com/item?id=17038154 , and if this stays on the front page, a clone of that conversation is likely to spring up here, too.
It seems there's too little understanding of how far machine learning is from the AI in Hollywood. I heard a bit of this on "The drive time with some A-Hole in the morning" and it was blown so far out of proportion, it was ridiculous. The problem is, that's where the average American (and similar media outlets) gets their information from.
We live in an infotainment society that pays lip service to the information portion of the word.
> Silicon Valley is ethically lost, rudderless and has not learned a thing.
I can understand how people could be creeped out, but ethically lost? Is "ethics" now just a buzzword?
Personally I wouldn't mind getting a call from Google. Google's voice is absurdly pleasant and friendly, which is likely a completely different experience to what someone on the receiving end of "cancel my Comcast" typically experiences.
The problem is silicon valley folks talk about ethics in a self serving way. They expect ethics from others but are not willing to give it back.
Whether stalking people and abusing personal data or now dehumanizing people, enabling bad actors and making people suspicious of callers it's always someone else's fault for not understanding 'innovation', not theirs for not considering the ethical implications. This is a tone deaf play dumb approach to ethics for profit.
No. That is a narrative from the alt right. SV folks have similar ethics if not better than the average person.
I would say generally engineers tend to be more ethical and rule following. They are built in this manner.
I am not sure what is happening but suspect part of the problem is some are scared of the future and tie SV to the future and taking it out on them. Where the narrative comes from.
The other possibility is they need a bogey man with the Clinton's gone and using SV to replace.
The last is more conspiracy oriented. The biggest strength for the US, hands down, is SV. The innovation coming out of SV is just unbelievable.
It could be Russia and other countries are trying to get us to fight to hurt SV for self serving purposes.
Who gets AI first will have an unmatched advantage. Google is far ahead of everyone and that could be driving the alt right narrative at Google. Using the Damore firing as the rallying event.
It really scares me how easy it is to manipulate some in the US. They use a fear of the future for some to stir it up. Kind of like how Trump and others are using people that look different to whip up people in a frenzy and scared of people that look different.
That they are coming to rape our women.
All I got. Been still trying to figure it out. I have engaged with a number of people on Reddit and surprised how many times someone negative on Google turns out to be Russian.
Wouldn't surprise me. Glass was killed because it was perceived as creepy. It's actual capabilities fell far short of being able to actually be creepy, but the biggest fear Google has is people realizing how creepy Google actually is. And this, is definitely outright creepy.
I am still a little stunned that either nobody at Google knew how this was going to be received, or that they didn't manage to take that person seriously.
I think glass was shattered because it was really lame and uncool. Outside of the tech bubble, you looked ridiculous wearing it, and no one wants to be filmed in a bar or public places by a bunch of weirdos.
I wore it for about a year. Most people didn't notice it at all, because I wore glasses, so much of the perception of it blended into the frame.
The whole "no one wants to be filmed in a bar or public places" thing is exactly what I'm talking about: People had no competent understanding of the device or it's capabilities, so had irrational views about it. ;)
Specifically, it was as obvious if you were filming with it as if you were holding up your phone camera (you had to hold it steady and the screen is lit up while it's recording, which is pretty visible), and Glass was a terrible video camera because it was a mediocre 2010-era cell phone camera connected to a device prone to massive overheating that had an approximate battery life of fifteen minutes if it was doing video.
IMHO, it's best use case was notification and call handling, and it would've been a far superior device if they left the camera off of it entirely.
The fact that Google sold it as a prosthesis didn’t help. I have a friend whose roommate invited over a guy who refused to take his Google Glass off, because he argued it was a part of him. Exactly what a woman wants to hear when she’s making toast in her pajamas.
Anyway I think the common thread here is when tech takes something that used to be socially symmetric and makes it asymmetric. Having a potentially always-on face camera is different than pulling out your phone. The person using their phone has to stop what they’re doing, and it’s usually pretty obvious that they’re taking a picture. Not so with the face camera person.
Similarly, Duplex takes something that previously required symmetric investment of time and makes it asymmetric. In theory you could have a virtual assistant burn up hours of someone else’s time and it would cost you pennies. Any emotional blowback would be safely absorbed by a relentless robot.
> I have a friend whose roommate invited over a guy who refused to take his Google Glass off, because he argued it was a part of him.
This just means your friend's roommate's friend is an unpleasant person. Nobody ever asked me to take mine off, but depending on the locale, I probably would have had I been asked.
> Having a potentially always-on face camera is different than pulling out your phone. The person using their phone has to stop what they’re doing, and it’s usually pretty obvious that they’re taking a picture. Not so with the face camera person.
Google Glass is not an "always-on face camera". As I indicated, it is a 15-minutes-on-at-best face camera, that might literally burn the face of anyone who had it on for fifteen minutes. (Seriously, after the Galaxy Note 7, I'd be a lot more scared of the thing and the LiPo it hangs just behind your ear.) And, also as stated, it's extremely obvious when it's taking a picture.
The folks who noticed mine were generally more intrigued than anything else. I had one guy ask me if I wanted a job on the spot while I was at lunch. A lot of people wanted to try it out, I used to be pretty protective of it though, and generally didn't let people.
I also took it to Paris for a week, and definitely had some fun conversing with people about it through a bit of language barrier (my French is not good, but serviceable), customs was not super thrilled I wore it through their line though.
You're saying the same thing as the comment you replied to. It didn't fail because it didn't work, it failed because people had unease about other people using it.
I think a company that only hires "the best of the best" and who refers to themselves with a specific word (Googlers) is bound to be quite disconnected from the average joe.
The technology is impressive. The demonstration choice was disengenous. It seems way more likely to me that the early adopters of this technology will be companies that will be able to replace phone receptionists with automation - already widely done by large companies, but now, thanks to this technology, available to doctors offices, restaurants, hair salons and many medium sized businesses.
Duplex will definitely be used in the opposite direction than demonstrated. Its acceptable to talk to a robot over the phone, and improving that process is very valuable.
But having a robot speak to the business? That might be a bit much. Then again, if this version is only for appointments, reservations, etc, than that might be seen as acceptable in kind of the same way.
It's mostly the hesitation about when this starts going beyond that. Do you now use Google assistant to talk to friends/coworkers? "Ok, Google, ask $friend if he wants to see a movie this weekend."
One of the advantages of Google hosting this (as opposed to, say, this technology being an "out in the wild" agglomeration of voice synthesis and machine learning open-source solutions bailing-wired together) is that there's one central authority that can be held accountable.
You're not going to see the described scenario because there's nothing in it for Google, they have full control over both the Google Assistant ecosystem and the set of technologies powering this machine-learning interaction, and they can gate-keep it as much as they want to.
That having been said... If one company can do it, we're only N years out from the tech becoming available "in the wild" as a bailing-wired selection of open-source tools (probably in a Docker image somewhere ;) ).
Even if not another company, people develop actions for Google Assistant so I don't think it's that wild to think that 3rd party Duplex behaviors could be something Google is considering.
You can already ask Google to send a message to your friends, what would be the point of the call? The goal of Duplex is to interface with businesses that are still stuck in a phonecall-only world.
It won't be annoying, because the other person is going to want to have an AI screen their phone calls and the two systems can just figure out the most likely time the humans want to hang out and then get them to confirm.
Employing an AI to be your personal assistant is at least as moral as employing a human to be your personal assistant. Perhaps more, given the very limited lifespan humans have.
That was my initial reaction as well. I felt as if Google was trying to position this as "We're not really going to kill a ton of jobs, just look how helpful it can be as a person trying to book an appointment", but that was pretty transparent.
I don't think the behind the scenes were exposed how this all worked and the development time but high tech like this is typically not accessible to small/med businesses because the overall costs to implement. Large corporations with existing crap IVR's is definitely a target.
Give it 3 years where it's a game changer to a bigger audience and devices. I could see this complementing teachers or enabling a whole new method of interactive learning, not making appointments.
I, for one, thought it was amazing how little of the uncanny valley I felt.
There are certainly things to consider like how I wouldn't want to be robo-called by scam bots that I can't tell are bots, and neither do others. A lot of businesses have an out already built for them though. Restaurants in particular have services like Open Table and Nowait that relieve the pressure of users wanting to use a realistic robo-call.
As an "Average Joe", my biggest concern is how this will push phone communications even (edit: more) toward where Facebook privacy settings are now - registered friends only - and how inconvenient that will be. I already let any numbers I don't have registered go to voicemail, and I don't like having to do that.
> As an "Average Joe", my biggest concern is how this will push phone communications even toward where Facebook privacy settings are now - registered friends only
Yeah I mean, that's essentially where we are now... I don't give out my phone number to anyone except my friends and a couple of companies I trust not to abuse it. I reject unknown numbers. Many people are like me.
Phonecalls aren't convenient at all. Expensive, atrocious voice quality, hard-to-memorize easy-to-typo numbers that sometimes have to change and are bount to the country (and sometimes to the provider), no way to identity callers unless they're already on your contact list, ... Should I go on?
Just use email if you want to contact me at random. I use Discord for voice calls, almost exclusively.
I don't think it's at all problematic for these devices to state "Hello, I am an automated assistant calling on behalf of $(name). I'd like to book a reservation for...."
I think Google's concern would be that it would generate annoyance that might not be warranted. Like, if I call a customer service line and get an AVR system, even if the system is perfectly capable of helping me, I just hate AVR systems (justifiably or not) and start yelling at it to give me a representative until it does. If the AVR system were so compelling that it just pretended to be a human and I couldn't tell the difference, presumably I'd be just as satisfied as if I had gotten an actual human, and could skip the annoyance.
Similarly, I think if this thing can handle the call as well as a human, it's to its advantage in terms of not frustrating the person on the other end of the phone to not self-identify, so that they perceive it as just like any other appointment call (because it is) rather than triggering an "ugh damnn it, another Google robocall" which will then prime people to feel like they have to speak to it a certain way to get it to behave, even if that demonstrably isn't actually necessary, and generate ill will towards the company.
On balance, maybe they should self-identify anyway -- "if this thing can handle the call as well as a human" is a pretty big "if" after all -- but I can see why they'd want not to.
If someone doesn't want to talk to you (or your bot), why not just leave them alone? It's not their job to be your guinea pig.
Was it right to not tell the victims of the Tuskegee syphilis experiment what was happening to them, because they would have refused and blocked the progress of science?
> it is their job to take reservations over the phone
Yeah, so presumably their bosses will instruct them to take reservations from automated agents, and grandparent comment's fears that people will hang up on automated agents that identify themselves as such will prove unfounded.
That's a minimum sort of requirement, but it doesn't alleviate all of the concerns. There's still massive potential for abusing the automated calling system to annoy people at scale, burdening service workers with cancelling and rebooking appointments as you fiddle with your phone and don't really think about the human burden, etc. And while I might choose to take no calls from a bot, a human employee of a given business may have no choice but to deal with fake John Legend six times an hour.
Essentially, what people have realized is that Google is turning humans into just another API they can interface with.
I don't see much risk of actually annoying people---since the service is run from a centralized company, they can take measures to prevent their own system from DDOS'ing people. It doesn't feel like it'd be too hard to institute rules like "No user can call the same number more than once every N hours." Hell, you can even institute rules like "No number is allowed to be called more than once every N hours from any user" if concern about the service itself DDOS'ing someone is significant.
On the topic of the asymmetry of the experience: is the larger issue that the fake John Legend voice will become boring to talk to for a human?
What you're thinking of as "blocking spammers" I'm thinking of as "preventing DDOS."
It's much easier to determine if there's too much traffic hitting a node or leaving a node than to determine if the traffic is "good" or "bad." I haven't seen anyone describe a way the system can generate "bad" traffic yet (that doesn't just degenerate into dumb games you could play without Duplex involved, like booking a reservation with your own phone and then just not showing up); generating too much traffic is a far more easily solved problem.
But on the topic of discerning whether the traffic is "good" or "bad:" this is actually a circumstance where the mass of data Google has access to that seems to so frighten people works to the advantage of the business-owner. Identifying bad users of the system could be done, for example, by collating data on reservations with geolocation data gleaned from maps to note there's a pattern of a particular user booking reservations they never use. Google can then just deny them access to the service to preserve the service's reputation.
"Ok Google, get me an appointment at the hair salon for Tuesday morning"
"I'm sorry fixermark, I can't call that number until Friday because 12 other people are in the queue to talk to the salon and Google doesn't want to annoy them"
A small snippet from an assistant conversation that is never going to happen :)
But it seems to me that the deceit is the only real innovation here.
This particular problem of booking appointments is solved by e.g. OpenTable - it's not useful to solve the phone conversation part of this transaction.
... and bridging the gap between computer interactions and land-line interactions is key for reaching people who don't have the resources to adopt an OpenTable or maintain a connection to it, which is one of Google's initiatives.
There was nothing of the sort in the demo. Just straight to "Hi, I want to book an appointment for, um, 10am". The "um" matters because it's implicitly telling you that you're not talking to a computer. (Explicitly saying "Hello, I am not a computer" would be a dead giveaway.)
I think there are two complementary things at play:
1) You can be very impressed with the tech (as everyone is)
2) You can be creeped out by how it could be abused.
BOTH things can be true. And that, I think demonstrates a continued blindspot especially by companies like Google, who simply don't seem to realize when and why things may come across as creepy to people. They keep doing this almost every year. They're so focused on how cool the tech is, they seem to just not realize that people might find it totally creepy. And this seems to be deep in the company's DNA (same with FB)
Did they do a focus group of people outside the valley bubble? I doubt it. Maybe a focus group of receptionists? What do they think of it?
I think I'm both, but I will tell you, the idea of having my own automated assistant deal with a company's automated assistant when trying to get through a phone directory to a live human sounds AMAZING. Like, I have zero problems with the ethics there.
Unless you can perfectly predict which calls are robocalls, assistant will end up talking to real people. If people know they're talking to a robot, there's a good chance they'll be annoyed.
Why would I be annoyed at talking to a smart answering machine? Seems at the very least no worse than hearing a prerecorded message and leaving a voicemail, provided the conversation goes as naturally as it did in the demos Google showed on stage.
Assistant: "Hello, you've reached Dave's personal assistant. Dave's not available right now, can I take a message?"
Bob: "No, just tell him to call me back as soon as he can."
Assistant: "Okay. Can I ask who's calling?" (Bob was not in Dave's contacts list.)
Bob: "It's Bob Jones."
Assistant: "Alright, I'll let him know."
Bob: "Okay, bye."
Perfectly natural conversation, and provides a much nicer user experience for Dave when he checks his missed calls. (Notification: "Call from: 444-444-4444 Bob Jones says 'tell him to call me back as soon as he can'")
What's so creepy about that? If I can get the same quality of service talking to a bot as I can to a person, why should I care what it is I'm talking to?
Unless you mean impersonating a _specific_ human (e.g. if my personal assistant were pretending to be me) I don't really see the issue.
I mean, all humans are specific humans. How could a bot not impersonate a specific human? It would be impersonating your assistant, who may not exist in the first place, but the bot is trying to get others to think that the person is a real person with real feelings. That duplicitousness is the problem. Like, other people are real people too, not just sheeple (terrible term that it is).
Why does it matter if I mistakenly think the answering machine bot is a real person? Assuming I can treat it like a real person and still get the results I want, why should I care one way or the other?
Because it is not a person. It's a machine. There is no shared experience, it doesn't have experiences, it's sand and metal.
Like, a really smart parrot is not a person, right? Even if the parrot could be an answering machine, it would not be due respect and it would not respect you back; because it's a parrot.
I think the issue is that the answer-bot is inherently disrespectful to the real human. Just using one at all is like saying to the real person:"You are not worthy of the time and/or respect of a real human; talk to this clever metal parrot." Text devices and digital sign-ups are at least 'equal', though the transaction is sterile.
Trying to fake that a person is talking to a bot is disrespectful of that other real person and their time; it's trying to pull one over on them and fake them into thinking that the bot's owner is worthy, but that the other real person is not worthy.
Just use a sign-up service. If the company/person does not have one, they are signaling that you have to actually talk to them, and you should respect their decision because they are a real person too, not try to pull a fast one.
Why do I need to have "shared experiences" with the person I'm ordering pizza from? I'm trying to order pizza, not get a date. It's not about "worthiness" or anything as elitist as that either; I just want to buy a pizza.
> Just use a sign-up service. If the company/person does not have one, they are signaling that you have to actually talk to them
Or that they can't afford to build an online order service? I seriously doubt the guy working the counter at my local pizza restaurant has any particular desire to speak to me personally.
Either way, forcing a person to talk to a robot without their 'consent' is, well, really rude. Many people hate phone-trees as is, forcing people to deal with them even more is not making the world a better place, it's making one person's world a better place at expense of others.
The whole point of this system though is that it's _not_ a phone tree, it's a conversational AI that's so good at what it does you might not even notice it isn't human.
I agree that forcing people to talk to a poorly designed phone tree with terrible voice recognition would be rude, but this isn't that.
Forcing a person to talk to a person without their consent is really rude too. What makes talking to a robot that's indistinguishable from a human worse than talking to a human?
That's only because they're a pain in the neck to deal with. If we're talking about an AI so good that you can barely even tell it's not human, I think that'd be much less of a problem.
Is there a reason to be annoyed? It's not very different from leaving a message with the possible exception you'll have a higher chance of having your message coming through (voice mail is where my inbound calls go to die)
I want this too, but it leads me to worry that in the not too distant future, poor people will deal mostly with robots in a sort of 2nd apartheid where social status dictates how many humans you can reach.
We're already in the situation where the only solution for a Gmail account being permanently hosed by some automated system is knowing someone at Google or having a million twitter followers.
I'm impressed with the speech synthesis, but I want to see a systematic evaluation of the conversational automation. A few demo calls do not cut it.
All of the successful Google innovations (Tensorflow, Alphago, Alphazero, map-reduce, millwheel, TPU) seem to have arrived with a proper publication style writeup. The ones that were less successful like Glass, Wave... not so much.
In the big picture, these projects are really about attention-grabbing, not necessarily researching products. They are supposed to create mixed, intense emotions, questions, dialog, and above all sublime awe. The function of this work is to control the dialog, to communicate that this organization is the nexus of the future to bring in money to survive and expand.
Many famous academic programs and large companies do this. After generating the anxiety, the lab pairs it with intimate access to the research labs in the form of sponsorship, partnering and other ways of reciprocating.
Essentially, the labs establish themselves as the alpha dog in the system. You may be afraid of what's coming, but at least you can look to them to understand it. It all breaks down when you start to realize the same people creating the stress are the ones you offer to relieve it, not by slaying it, but by cutting you in on the deal.
And it's always difficult to interpret because the tech itself is 1st class, so the artisan class is awed and jealous, so they want to be part of it. Meanwhile, most of those critical of technology sound like complete wankers who want us all playing homemade instruments around a campfire eating raw seaweed. And Chris and Sal Average Family is still cheering for U.S. space dominance as conditioned by the space program since its inception, so they're rooting for technological progress in the form of narcotizing Narcissism (as described by McLuhan).
Sorry if I'm a downer, but I have been part of this system in the past and now I'm on in the edge in and out of it and I study lots of philosophy and media theory. This is how this system survives.
> they seem to just not realize that people might find it totally creepy.
MOST people dont or dont care. The same with #deletefacebook.
There is a vocal minority including the press, and then there is everyone else who just wants cool things to work well and easily, and doesnt care at all about privacy.
I think #2 can be really simply fixed: they always have to identify themselves as machines. Once you know it's a machine you're talking to, you treat it as a less-awful version of the voice prompts we've had on phone systems for years.
The potential efficiency gains from this (think of calling into one of these instead of your traditional phone menus) are substantial. We should start looking into how to regulate AI (even if it just starts as "AI should identify itself as AI when they're using human communications channels") to make it work well for us.
This goes against the most interesting part of the demo, though - that the 'umms' and 'uh-huhs' (and other small things) are enough to make this system pass the Turing test, at least over a narrow communication channel.
Honestly, one thing your idea may show how far the big tech companies have strayed from their desire to create a concise and interesting product. Doing something like what you propose - basically, a system that calls, identifies itself as a robot and then is freed from the (difficult) technical problem of appearing to be human would be much easier to implement. With a bit of Twilio, "weekend hackathon to a workable demo" levels of easy to implement, IMHO. And, honestly, it looks like the public at large might have preferred that, and anyone using such a system to schedule an appointment would likely have a pretty much identical experience to the one provided by Duplex.
I don't think the technical problem vs the self-identification are conflated. It's still easier to talk to something that acts like a human (at the conversational level, at least). I'd prefer those subtle hints ("umm") over explicit "That's not what I asked for" when communicating. I'd prefer to expect them to understand the same.
When I know I'm talking to a software agent, I can change the way /I talk/ without having to lose fidelity from the way it talks. For example, I can spit out a large, complex sentence that's usually harder over the phone: Groups of four don't need or get reservations. "Reservation availability is Tuesday 2 to 5 pm, wednesday 2 to 3pm, friday 4pm to 6pm. " Then a "Thank you, goodbye" message tells me that I'm done with them, and as I know it's an AI, I don't have to worry about having offended a customer.
> they always have to identify themselves as machines
That sounds like it will work as well as the federal do-not-call registry.
I think most of the issues could go away if we fixed the phone system to provide authenticated information about the source of the caller. If you can't or don't want to provide that information, that's fine, but I'm not going to answer that call.
Oh, that's a good idea. The AI implicitly identifies itself via caller ID. If you're a hair salon, you get lots of calls from that number for many different customers.
I know I'm a bit of an outlier here, but I don't get all the fuss. If the bot identified itself as a bot, I can imagine the human reacting in all sorts of weird ways that would prevent the task from completing. Things like raising their voice, speaking in weird half phrases/broken english, or just immediately hanging up. "OPERATOR NOW". "WHO NAME CUSTOMER".
Perhaps it'll be awkward, and fail. But so what, I'd rather give it the ole college try.
For a field like AI, it's extremely difficult to decide if knee-jerk aversion to specific advancements is something worth taking seriously or not. For a field where the entire point is to replace humans in certain functions, it's inevitable that there will be leaps forward that people are extremely uncomfortable with in the short term.
This may be an example of something where things need to be altered, or it may not. But in either case it seems short-sighted to just think this question can be resolved in 24 hours after a demo due to a few hastily-written tweets and thrown-together articles. What's important is that as AI continues to encroach upon human capabilities that there be a framework for navigating these transitions and knowing where the lines are. Hopefully as this kind of thing becomes more common public opinion will shift from caring about the creepy-demo-of-the-week to a wider scope of how we grapple as a society with these kinds of advancements.
Emotional reactions and ad-hoc decisionmaking based on mob rule is going to result in a random, arbitrary outcome which will almost certainly be suboptimal.
You can both take something seriously, and also deliberately decide that the reactions were not sufficient to invoke a change of direction. Being serious about evaluating a customer response doesn't mean you cater to that response.
Again, you can. See how well that worked for NASA's Integrated Program Plan. It was too ambitious a project that generated some spectacular design and planning work, but ended up dashed on the rocks of the political and social environment of the time.
Sometimes suboptimal is a lot less frightening and destructive than optimal.
Dictatorships and despotism are the optimal solution for wielding of executive power.
Murder is the optimal means through which to immediately resolve problems created by other people from a time investment point of view.
Rape is an optimal solution to spreading genetic material.
See the pattern? Knee-jerk reactions from the layperson AREN'T necessarily bad. If anything, they should be the first sign that a researcher needs to back up from that awesome thing you just did, and take off the razor thin beam of relevance you used to achieve your breakthrough, and begin the HARD part of research. Turning it from an interesting academic factoid into a societal good.
When we do research, we are free to ignore huge swathes of the world to be able to reach or findings. The reckoning, however, comes when we have to ask ourselves 'how prepared is the world for this?'
The story of Doctor Frankenstein isn't just some scary story for children. It should be haunting a researcher night and day. What am I enabling? How will society take this? It shouldn't stop you by all means, but it will keep you grounded and aware of the fact your research isn't happening in a void.
Remember, all those bits and pieces you are putting out of your mind are the bread and butter of every layperson's day. Being a researcher doesn't absolve you of your status as a member of society.
It is YOUR responsibility to take your idea from the theoretical to the practical. The rest of society is busy staying intact enough to support your research efforts.
My point was that a suboptimal solution is one where there was no careful, methodical thought to the design of AI systems. If AI designers and researchers blindly react to public outcry after 24 hours from a demo (which, in general, is something I would expect Google to do) the kind of thinking you mention above is just as unlikely to happen as if they trudged on forward without any consideration to these things at all. In both cases, this is a suboptimal outcome for society.
In one, we get a fairly random, undesigned world of AI systems that don't serve anyone well and generally are underutilized because nobody is willing to push the boundaries. In the other, we get dystopian AI hell. It's important that researchers be given the space to think you describe. The right way to allow that is to foster an informed and open-minded public about the emergence of AI and the decisions we need to make about it as a society, and not have prominent voices writing about knee-jerk reactions making authoritative demands to a specific public demo after just a day. (See Max Tegmark's book Life 3.0 for the kinds of stuff we need more of being put into the world imho.)
>Emotional reactions and ad-hoc decisionmaking based on mob rule is going to result in a random, arbitrary outcome which will almost certainly be suboptimal.
I may have initially misinterpreted some context when I read your comment before, though my conclusions haven't changed much.
I agree with you that AI research must be methodically and transparently approached.
The part where I was trying to speak toward, however, was that AI researchers/designers are going to necessarily be constrained by how much faith and goodwill a (currently) troubled society, plagued with inequality and hardship in the face of increasing job loss to automation. This is a feature, not a bug to be dismissed.
AI is akin to nuclear technology in that it is perceived as a huge game changer. One that brings with it a huge potential for societal change and upheaval.
The AI designer/researcher DOESN'T necessarily see it that way. It is just a tool to do a thing. The populace, however, don't grasp the limits the researcher/designer does. They are being asked to give the same degree of faith to the AI field that nuclear physics was given in the last century.
Combine that with a severe distrust in the motives of the organizations doing the actual development and you have the stage set for large-scale rejection of AI research.
You said it yourself before. AI is a field that is trying to replace humans (having to do stuff). That is not a viewpoint that will win you the favor of the populace. No one will get further than the first half of the sentence before the hackles go up.
Nobody trusts a Utopia where people don't have to work because machines do it for them. Nor will they trust the men behind the machines. It's sad, but it is the kind of thinking that has kept humanity OUT of a complete dystopian hellholes thus far.
I don't know the fix, but I do know that if we can't solve some very fundamental social problems that advances in artificial intelligence and machine learning are poised to make worse, you will find the field of study becoming more and more difficult to find public acceptance in. The only inkling of a clue I have is it may have to do with the "industrial asymmetry" of our current times.
If you don't take heed of the state of society, you may end up in the same place NASA was for the latter half of the 20th Century. BEAUTIFUL theoretical groundwork laid, but hamstrung in execution by popular sentiment. (See some documentaries to learn how "thrilled" many were with the Space Race. NASA still hasn't recovered from the backlash despite some of the best science I've ever seen)
The only automated voice calls I get are always spammers.
I can't see many real people wanting to automate calling the hair salon, because then you're being fairly pretentious as if you're too busy to pick up the phone.
But I can see a massive incentive for scammers to stop having to employ cheap Indian labour and now have an almost infinite ability to make human like calls to non-tech people in an American accent.
So be prepared to hear the Google assistants calling you to tell you there's a problem with your Windows PC and they just need to install something on there to fix it.
> I can't see many real people wanting to automate calling the hair salon, because then you're being fairly pretentious as if you're too busy to pick up the phone.
Just a short list:
- the deaf
- the mute
- the very socially awkward
In the last category: I can personally attest that I've put off scheduling an appointment with someone I've done business with in the past for two days, purely because the thought of actually picking up the phone and going through the process of aligning our calendars makes the hair stand up on the back of my neck. If I can pass that task to an assistant, my life would be much easier.
Deaf and mute are good counter examples. I know that Google quoted 60% who don't have websites but that figure should only ever get smaller. Such an impersonal interaction works perfectly through the web.
I'm not a fan of calling people either (I've put off finding a local doctor for a year because I don't want to go through the awkwardness) but starting a relationship with an important client using a not quiet real 'assistant' would start off the relationship in the wrong foot. Certainly if someone calls me using this I'll think that they can't even be bothered to speak to me personally and so I'm not worth their time.
I'd imagine the level of personal touch that matters varies from industry to industry. I don't know a restaurant much cares if the reservation is made via the reservation holder, their human assistant, or their robot assistant as long as the customer shows up on time, pays, doesn't wreck the place, and leaves a nice tip.
I have very mixed feelings about Google Assistant due to that. Being able to automate calling the hair salon seems like a thing from the dreams. Calling people is hard and painful. On the other side, letting your life be managed by an always online assistant connected to the biggest advertisement company harvesting every bits of information about you seems even scarier. I'll probably try to live as before, preparing for a phone call for a few days until finaly gathering the needed courage :P
I think it's very interesting that the only positive defense of Duplex that I'm seeing, is about using it for accessibility purposes.
And you know what? It's a really really good point, Duplex could help a lot of people get things done. You know what else? There are already services like that, and they always announce themselves. Making that cheaper and more automated, so it can benefit more people, seems like a fantastic idea. No need to turn humans answering phones into an API to do something good in the world.
I used to do phone support for an ISP in the 90s and I would get calls quite often from someone who would announce that they were interfacing between me and a text based system for a deaf customer. If a robot did the same thing, I think everyone would be fine with that.
A friend of mine has a fairly significant speech impediment. He can communicate fine face-to-face, but he finds phone calls to be incredibly frustrating, especially with strangers. People often assume that he's deaf, drunk, learning disabled or a crank caller.
sometimes you just want to avoid talking to another human, I would be very happy to do these things online instead of an actual talk. I think the advantages are more in the other side, from customer support, which from some company is terrible (mostly fro management/budgets issues)
But this is what websites are perfect for. Perhaps some of the best places are the ones that don't have an online presence but I don't know of many.
From the other way I can see the sense, but the ones that are terrible don't get any better by having a cheaper option. It'll just mean they'll have a terribly programmed assistant instead of an Indian call centre where they've been given bad instructions.
unfortunately many places don't, at least in my experience here in UK. Either because they are too small (like the hairdresser in the example) or because they are big and old or even they only use the phone to discourage customers complaints/profit on the call (even Ryanair for legitimate complains)
IMO the success of services like Grubhub/Seamless and Opentable show that people really don't want to pick up the phone and call someone if they don't have to.
I don't particularly want to either. But then I don't want to subject my local hairdresser to Google's assistant either. I'd rather just restrict myself to the ones that provide a website.
> you're being fairly pretentious as if you're too busy to pick up the phone
I don't think this is a good argument. The same idea, applied to other domains, sounds outlandish:
- you're too busy to drive to the store to buy something [using Amazon]
- you're too busy to pick up a pen and write a letter [using email]
- you're too busy to pick up the phone [using SMS]
The progress of technology is literally about removing friction from tasks which previously took longer amounts of time, and (except at the beginning, maybe?) this is rarely interpreted as pretentiousness.
Of the non-technical people that I've spoken to about Duplex (only 4, admittedly), they were universally excited by the demo.
As far as the proliferation of scammers, I think you're absolutely correct. I think it's plausible that human callers will incorporate challenge-response questions to check for things like this ("Who is Batman's sidekick?", "If I were unfairly comparing two fruits, what would they be?", etc).
Then again, who's to say that Google will make their tech available outside the company? The accuracy of the demos probably largely stems from the massive dataset that Google has amassed, which won't be available to scammers.
What about the AI advancements in early disease screening and treatment that Sundar highlighted in the beginning of the keynote? Why not mention that as a counterbalance to quoting professional critics and at least touch on the undisputed benefits from this stuff.
What's truly “horrifying” is the somewhat prevalent phobic attitude towards technological innovation and invention, be that industrial robots ‘taking all the jobs’ or this reaction to the Duplex demo.
Our current technological achievements were largely made possible with help from the media and our prominent thinkers in cheerleading the advancements and shaping the public’s option.
Now it's the critical voices that get affirmed and amplified.
I haven't seen any backlash about AI being used to detect diseases better/sooner.
The fact that we see a backlash about something that treats humans as an API, and don't see a backlash about something that helps all humans, seems like there is no prevailing innovation-phobic attitude. No?
The industrial robots taking jobs thing is at least pragmatic to discuss - we could just accept the capitalistic inevitability of it, or we could talk about ways to move the human workforce on to other tasks.
The problem is that of general attitude, I mean this shit is cool, people should be excited about it, instead people are being trained to be pessimistic and to invoke some vague notions about 'ethics’ to justify aversion to change.
Skepticism and questioning are fine but the media is trying to turn people into luddites because they feel they were wronged by internet firms.
I think hacker news suffers from being too close to tech to really get a normal person's perspective. These advances are technologically marvelous, sure, but they are ethically bankrupt. At best, they make human interaction worse, and at worst, they enable robocallers and standardize thought. Tech optimism has become toxic because SV fails to realize how each cool new feature can be bastardized and abused.
Not only are they ethically bankrupt, it feels like the sole goal is to eliminate all human interaction and more importantly solve problems that do not exist.
Ex. I do not mind making a 60 second phone call to make a reservation vs. 30 seconds of figuring out how to tell Google what to do.
In this case Google is solving problems for rich socially awkward nerds. Not the majority of people.
I completely disagree with your line of thought. Automation is something that do make lives easier, and just because your personal experience tells you something, it doesn't mean that it will be the same for everyone. Like many others stated before, this is marvelous for the social awkward and the deficient, while also bringing a personal assistant for people with tight schedules or unable to make such calls for various reasons. Also, keep in mind that this kind of feature will most likely be present in your average Android smartphone, so basically everyone with access to one will be able to use it.
How many of you have immigrant parents who, where English is a second language, would have great difficulty performing the kind of verbal gymnastics that you take for granted?
My mother is an immigrant and speaks 4 languages, and I am learning my third. This is a ridiculous argument; if this is so important to you, advocate for Google translate integration into email.
You really can't think of a better possible outcome of technology that enables humans and computers to interact in a way that's more natural for the humans?
IO is a developer conference, the whole critique about unveiling it before they know everything about how to market the product, or before it's complete, is tone deaf.
To me, the hidden crux of this whole issue is "can this system be as effective to deal with as a human who is good at their job?"
If that condition is met or exceeded, none of this voice stuff will matter. People won't care if the system discloses it's nature, or sounds creepy, if their airline reservation change (or whatever) goes abnormally well.
If that condition is not met, which IMO it won't be in the near future, then we're just adding attempted manipulation and potential for abuse on top of an already bad UX.
Off a tangent, i don't like how bloomberg has become a tech reporter of note. They try to bridge things between the techies and the critics, which ends up making it worst for both sides. I prefer to get my tech news from nerds, with the suits standing on the other side watching.
>The obvious question soon followed: Should AI software that’s smart enough to trick humans be forced to disclose itself.
No. Replace AI with your favorite disadvantaged minority and the answer will be obvious. If AI is smart enough to pass Turing Test, it should be treated as a human.
Why do you think it's valid to replace the word "AI" with your favorite disadvantaged minority? Sounds like you have a pretty low opinion of these disadvantaged minorities.
Ha, it's because no one at Google pictured themselves as the one on the other end of the phone taking the orders.
Because phoning in for restaurant reservations is only the beginning. Ask yourself now - in a few years, will you be one of the people giving orders to a digital assistant, or one of the (many more) receiving them?
"The future is already here - it's just not very evenly distributed."
> You implied the reasoning behind your statement was obvious ("of course") but it's not at all obvious to me.
No. The affirmative answer to the original question as posed is what seems to me to be obvious. My additional statement was meant to convey the seriousness with which I feel these matters should be taken.
> Why should it be legal fraud with severe consequences for an AI caller to not disclose the fact that it is an AI?
Otherwise people (specifically the technologists implementing and selling these kinds of trickster services) will not take other people's (specifically me and anyone who feels as I do) concerns seriously and proceed apace.
There is a similar issue around folks who see CRISPR as a way to work around laws that attempt to control GMOs. I have seen a few articles written by pro-GMO folks just crowing about how CRISPR-edited organisms can be sold without GMO labels due to loopholes in the law. These folks have their heads so far up their belief systems that they don't even see the problem.
Again, I don't see why that's obvious. Why is it necessary that humans be made aware up front that they're talking to a bot? (So necessary that you think not disclosing that information actually needs to be made illegal.)
Okay, first, let's keep clear exactly what we're talking about here:
A machine that attempts to imitate a human well enough to trick people into performing a specific action, an action that involves commitments of time and money.
I have a specific and a general objection to this. The specific objection is that this thing tries to trick people. It sets up a "frame" that is intrinsically deceptive. That's it's goal.
In another comment s2g quoted:
> Giving it an obviously robotic voice when it calls. "People will probably hang up," he said.
It is designed to trick people. That's bad. When you do it in business, we call it "fraud".
My general issue with this is that it normalizes the blurring of the line between real human people and ersatz Artificial Intelligence "people". This is something that's going to be a serious problem on a deep level. There's evidence of this sort of conceptual melting right here in this thread. How long before folks are trying to marry their toaster?
Last but not least, there's the meta-problem: the disconnect between the techo-elite innovators and the masses.
Think about this: As a businessman, when the phone rings and there's robot there trying to give me money why would I hang up? The robot only called me because someone wants to make a sale (from my POV) so that's the "opposite of a problem". This is the meta-problem: the technologists are so effin' clueless yet default to starry-eyed optimism.
"We are as gods, we might as well get good at it."
IMO behaving like a normal human isn't "tricking people". Voice assistants need to behave like a human to enable natural interaction, and the fact that the AI caller isn't human is irrelevant information in most contexts.
It'd be like if I called a business and used a voice changer during the call to make me sound like I'm a different gender than I am. Would that be fraud? No, because my gender is irrelevant information in the context of that call.
Now maybe it might be a good idea to program the AI to be upfront about its nature just as a matter of courtesy, but I see no reason why it _not_ doing would be so harmful that it needs to be made illegal.
> IMO behaving like a normal human isn't "tricking people".
It is tricking people if it is done to trick people.
It is not tricking people if it is done while the person to whom it is done knows the thing they are talking to is not a human.
The question was not "should machines be allowed to act like humans?"
It was "should they be required to disclose that they are not human?"
You seem to think that I object to machines acting human, I don't.
> Voice assistants need to behave like a human to enable natural interaction
Would you say the computers in the "Star Trek" shows "behave like a human"?
> the fact that the AI caller isn't human is irrelevant information in most contexts.
How will it know if, in a given context, disclosure is relevant? One of my first questions about this Duplex service is, "Can I opt-out of receiving these calls?"
> It'd be like if I called a business and used a voice changer during the call to make me sound like I'm a different gender than I am. Would that be fraud? No, because my gender is irrelevant information in the context of that call.
I don't think it's like that at all. Are you impersonating someone else? Gender-bending for kicks? Trying to get better service because you think the person you're calling is sexist or something? Are you The Jerky Boys?
People do weird shit and that's something we all have to deal with because gravity.
Machines calling people and saying "uh" to pretend to be human without the human people realizing it is trickery.
And it's not even needed to make the Duplex service work because, as I pointed out before, when a robot calls you to give you money you don't hang up on it! Okay?
> Now maybe it might be a good idea to program the AI to be upfront about its nature just as a matter of courtesy, but I see no reason why it _not_ doing would be so harmful that it needs to be made illegal.
And now you have made my point for me: if it's not made illegal, people, like you, won't take these concerns seriously.
It's like the guys who owns my favorite cafe. He had a bit of plastic wrap covering his brown sugar and there were little flies sometimes and it was just a little gross. One day I came in and he had put a tight-fitting plastic lid on the sugar. I mentioned it to him and he started complaining about the health inspector making him do it. All I could think was, buddy you're the reason we have to hire and pay for health inspectors, right there.
AIs that can impersonate humans well enough to trick them into thinking they're also human are pretty much a munition. Thank you Google for opening the gates. Now that people realize it's possible we have another thing to guard against. You know most hacking is done though what's called "social engineering", eh?
> It is tricking people if it is done to trick people.
That's true, but I don't see behaving like a human as "tricking people". Even if the assistant opened the conversation by identifying itself as an AI, I'd expect it to use all the same mannerisms "uh", "mhmmm" that it did in the call to enable natural conversation.
> Would you say the computers in the "Star Trek" shows "behave like a human"?
No, but I wouldn't exactly call that "natural interaction" either. Jarvis is a better example. And yes, I do think Jarvis "behaves like a human".
> I don't think it's like that at all. Are you impersonating someone else? Gender-bending for kicks? Trying to get better service because you think the person you're calling is sexist or something?
In keeping with the analogy, probably closer to "Trying to get better service". An AI callbot that pretends to be a human will be more efficient at its job than one which speaks in a monotone and requires the human on the other end of the call to adjust their mannerisms to compensate for the fact that they're not talking with a human.
> And it's not even needed to make the Duplex service work because, as I pointed out before, when a robot calls you to give you money you don't hang up on it! Okay?
Depends. If the AI immediately identifies itself as such, how many people are just going to immediately assume it's a spambot and hang up, even when it's not?
> people, like you, won't take these concerns seriously
What concerns? You've still failed to describe any ill effects that might arise from AI callbots failing to disclose their nature upfront.
> You know most hacking is done though what's called "social engineering", eh?
You think a law forcing AIs to disclose themselves as AI would do anything against malicious actors who are already breaking the law? (Hacking.)
I can imagine a lot of cases in the future where this goes awry and is abused, probably not directly by Google Duplex, but by homemade versions.
I think there should be a kind of non-jarring way for AIs to identify themselves - maybe playing three quick tones before the AI speaks - to identify that it isn't a human on the other hand and therefore shouldn't be able to say, order thousands of dollars of food. This is quick, unobtrusive, easy for an AI to do, and easy for an AI to detect.
I was far more concerned about the application of AI to healthcare. Did they not think about the possible abuse from telling hospitals or insurance companies which of their patients are very likely to need urgent care within the next week?
> Giving it an obviously robotic voice when it calls. "People will probably hang up," he said.
Doesn't that really tell us all we need to know? If the only way to get people to tolerate your service is to try and trick them then is your service a good thing?
Deception is key to profitable ads and news as well. We just keep diving deeper into the murk and find ourselves more and more confused about what's going on around us, who and what to trust. It's sad and technology has continued to accelerate it.
What I think is really interesting here is that all these questions surrounding this are focused on the human-AI interaction and whether the human should know that they're talking to an AI when, in reality, what Google is after is completely removing the "human" aspect of that relationship. Google is betting that, in the future, we won't need people at all to do these kinds of things. In fact, they would love if every booking for every appointment was done seamlessly for the user without ever interacting with people. This Duplex technology is just a stop-gap to fill in that interim period where we're still using people for tasks that don't require specialization, are very mundane, and end in a very obvious success condition. All you need to make an appointment somewhere is a time/date and a confirmation that the appointment was made. Instead of an API, we currently have a "human" API that handles that. I guarantee that Google is banking on Duplex eventually bridging that gap so that, eventually, I can make an appointment anywhere simply by asking my personal assistant to do so and the "human" is completely unnecessary.
These are the questions we should be looking at, in my opinion. This situation is only temporary. We need to focus on the future (that is most certainly coming) ideas of what we're going to do with our societies when even these simple jobs that we thought required that human touch are replaced by automation and AI.
> "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."
Oh sweet bloviating lord of hot takes, this? This is what's horrifying? Not the techno-utopians and techno-capitalists working hand-in-hand to shepherd all human interaction onto the internet while simultaneously turning it into a nightmarish complex of data burbclaves and dopamine farms? Not founders, VCs and politicians gleefully circlejerking about how profitable it is to create technological monopolies that evolve too quickly for society's immune system to regulate, like some kind of economic retrovirus? Not the left and the right working opposite ends of the great authoritarian handcart, on one stroke demanding that private entities exercise government-level control over digital spaces, and on the other demanding that they be free from government-level accountability?
When you get a call from a robot at work, the thought that it talks just like us should be approximately the fifth-scariest thing to enter your mind, followed by the thought that this means your job is going to get automated away soon, the thought that maybe it won't be and you'll end up a robot's page boy, the thought that what you're doing is so mundane that a computer can not only do it, but also completely model how you will do it, and finally the thought that this stunning waste of human life bothers you less than the possibility of being unemployed.
As far as things to feel creeped out about, I'd give robo-bookings maybe two stars: some flashy overtones but ultimately lacking in substance. I recommend a deeper, more full-bodied creep-out, like remembering that Facebook cuts people off from their friends if they don't provide a data-matching-friendly real name. Papers, please!
Should AI software that’s smart enough to trick humans be forced to disclose itself. Google executives don’t have a clear answer yet.
I would think that Google's executives shouldn't have a say in that question at all. As the chosen representatives of the people, that is a job for the Government.
Wouldn't it be racist to force an intelligent entity to disclose its nature? Why does it matter if the person chatting with you on the phone is made of meat or a super-being who's having thousands of similar conversations at the same time?
Wouldn't it be racist to force an intelligent entity to disclose its nature?
This argument is hilarious, but I'll play along. First of all "racism" is about superiority, not about differentiation per-se. There is nothing inherently wrong about acknowledging differences, it is wrong when it becomes the sole reason for assuming superiority. That is the textbook definition of racism, so to answer your question: No, in my view it would not be wrong to force an intelligent entity that is not human to disclose this fact.
It matters so that my approach to dealing with this "intelligent entity" can be adjusted accordingly, as per my needs.
Why does it matter if the person chatting with you on the phone is made of meat or a super-being who's having thousands of similar conversations at the same time?
It matters because today we do not yet understand anything about the drivers and motivations of this "super-being" and before we are clear on that, I would like to know who or what I am conversing with. I am pretty sure you are referring to "Minds" from "The Culture" series, which a) is fiction, and more importantly for the sake of this discussion, b) Minds have a relatively well understood context of motivations, morality and history of mostly benign interaction with Culture citizens. The "Robocall version 2.0" touted by Google doesn't have that, and given Googles past track record, they don't really give a flying fuck about the general population. When it comes to the proprietor and owner of this technology today, we have been comprehensively identified as "the product" instead of "the customer" -- it is my right as a private citizen to know who or what I am interacting with, especially if it is a Google product.
Finally, whereas dealing with other humans can be understood on a base level of expected intelligence and wits, as well as shared identity, history and culture, interaction with "super-being" would by its' very definition be an unequal exchange. I have every right to know if my interlocutor had any kind of extreme advantage over me (you know, because it is a "super-being").
> First of all "racism" is about superiority, not about differentiation per-se.
In this case, the IA would disclose its nature and its lack of independent agency. It'd disclose the fact it's not considered a living being and not afforded any rights. The human, then, could engage in all sorts of abuse (and Microsoft had a lovely chatbot that had its "brain" damaged by trolls) that could leave lasting effects on its persisted state.
> I am pretty sure you are referring to "Minds" from "The Culture" series
You are completely wrong. Google's "Robocall 2.0" is nowhere near close a sci-fi worthy AI.
153 comments
[ 2.7 ms ] story [ 198 ms ] threadWe live in an infotainment society that pays lip service to the information portion of the word.
https://news.ycombinator.com/item?id=17022963
Which is heartening IMO. People who deal with computers all day should be the ones most skeptical of technology.
I can understand how people could be creeped out, but ethically lost? Is "ethics" now just a buzzword?
Personally I wouldn't mind getting a call from Google. Google's voice is absurdly pleasant and friendly, which is likely a completely different experience to what someone on the receiving end of "cancel my Comcast" typically experiences.
Whether stalking people and abusing personal data or now dehumanizing people, enabling bad actors and making people suspicious of callers it's always someone else's fault for not understanding 'innovation', not theirs for not considering the ethical implications. This is a tone deaf play dumb approach to ethics for profit.
I would say generally engineers tend to be more ethical and rule following. They are built in this manner.
I am not sure what is happening but suspect part of the problem is some are scared of the future and tie SV to the future and taking it out on them. Where the narrative comes from.
The other possibility is they need a bogey man with the Clinton's gone and using SV to replace.
The last is more conspiracy oriented. The biggest strength for the US, hands down, is SV. The innovation coming out of SV is just unbelievable.
It could be Russia and other countries are trying to get us to fight to hurt SV for self serving purposes.
Who gets AI first will have an unmatched advantage. Google is far ahead of everyone and that could be driving the alt right narrative at Google. Using the Damore firing as the rallying event.
It really scares me how easy it is to manipulate some in the US. They use a fear of the future for some to stir it up. Kind of like how Trump and others are using people that look different to whip up people in a frenzy and scared of people that look different.
That they are coming to rape our women.
All I got. Been still trying to figure it out. I have engaged with a number of people on Reddit and surprised how many times someone negative on Google turns out to be Russian.
I am still a little stunned that either nobody at Google knew how this was going to be received, or that they didn't manage to take that person seriously.
The whole "no one wants to be filmed in a bar or public places" thing is exactly what I'm talking about: People had no competent understanding of the device or it's capabilities, so had irrational views about it. ;)
Specifically, it was as obvious if you were filming with it as if you were holding up your phone camera (you had to hold it steady and the screen is lit up while it's recording, which is pretty visible), and Glass was a terrible video camera because it was a mediocre 2010-era cell phone camera connected to a device prone to massive overheating that had an approximate battery life of fifteen minutes if it was doing video.
IMHO, it's best use case was notification and call handling, and it would've been a far superior device if they left the camera off of it entirely.
- Please don’t record me in a public place.
- Chill, it’s not a very good camera!
The fact that Google sold it as a prosthesis didn’t help. I have a friend whose roommate invited over a guy who refused to take his Google Glass off, because he argued it was a part of him. Exactly what a woman wants to hear when she’s making toast in her pajamas.
Anyway I think the common thread here is when tech takes something that used to be socially symmetric and makes it asymmetric. Having a potentially always-on face camera is different than pulling out your phone. The person using their phone has to stop what they’re doing, and it’s usually pretty obvious that they’re taking a picture. Not so with the face camera person.
Similarly, Duplex takes something that previously required symmetric investment of time and makes it asymmetric. In theory you could have a virtual assistant burn up hours of someone else’s time and it would cost you pennies. Any emotional blowback would be safely absorbed by a relentless robot.
This just means your friend's roommate's friend is an unpleasant person. Nobody ever asked me to take mine off, but depending on the locale, I probably would have had I been asked.
> Having a potentially always-on face camera is different than pulling out your phone. The person using their phone has to stop what they’re doing, and it’s usually pretty obvious that they’re taking a picture. Not so with the face camera person.
Google Glass is not an "always-on face camera". As I indicated, it is a 15-minutes-on-at-best face camera, that might literally burn the face of anyone who had it on for fifteen minutes. (Seriously, after the Galaxy Note 7, I'd be a lot more scared of the thing and the LiPo it hangs just behind your ear.) And, also as stated, it's extremely obvious when it's taking a picture.
The few days I saw him he was getting chatted-to about it, mostly by women.
I also took it to Paris for a week, and definitely had some fun conversing with people about it through a bit of language barrier (my French is not good, but serviceable), customs was not super thrilled I wore it through their line though.
But having a robot speak to the business? That might be a bit much. Then again, if this version is only for appointments, reservations, etc, than that might be seen as acceptable in kind of the same way.
It's mostly the hesitation about when this starts going beyond that. Do you now use Google assistant to talk to friends/coworkers? "Ok, Google, ask $friend if he wants to see a movie this weekend."
You're not going to see the described scenario because there's nothing in it for Google, they have full control over both the Google Assistant ecosystem and the set of technologies powering this machine-learning interaction, and they can gate-keep it as much as they want to.
That having been said... If one company can do it, we're only N years out from the tech becoming available "in the wild" as a bailing-wired selection of open-source tools (probably in a Docker image somewhere ;) ).
Employing an AI to be your personal assistant is at least as moral as employing a human to be your personal assistant. Perhaps more, given the very limited lifespan humans have.
Give it 3 years where it's a game changer to a bigger audience and devices. I could see this complementing teachers or enabling a whole new method of interactive learning, not making appointments.
There are certainly things to consider like how I wouldn't want to be robo-called by scam bots that I can't tell are bots, and neither do others. A lot of businesses have an out already built for them though. Restaurants in particular have services like Open Table and Nowait that relieve the pressure of users wanting to use a realistic robo-call.
As an "Average Joe", my biggest concern is how this will push phone communications even (edit: more) toward where Facebook privacy settings are now - registered friends only - and how inconvenient that will be. I already let any numbers I don't have registered go to voicemail, and I don't like having to do that.
Yeah I mean, that's essentially where we are now... I don't give out my phone number to anyone except my friends and a couple of companies I trust not to abuse it. I reject unknown numbers. Many people are like me.
Phonecalls aren't convenient at all. Expensive, atrocious voice quality, hard-to-memorize easy-to-typo numbers that sometimes have to change and are bount to the country (and sometimes to the provider), no way to identity callers unless they're already on your contact list, ... Should I go on?
Just use email if you want to contact me at random. I use Discord for voice calls, almost exclusively.
Similarly, I think if this thing can handle the call as well as a human, it's to its advantage in terms of not frustrating the person on the other end of the phone to not self-identify, so that they perceive it as just like any other appointment call (because it is) rather than triggering an "ugh damnn it, another Google robocall" which will then prime people to feel like they have to speak to it a certain way to get it to behave, even if that demonstrably isn't actually necessary, and generate ill will towards the company.
On balance, maybe they should self-identify anyway -- "if this thing can handle the call as well as a human" is a pretty big "if" after all -- but I can see why they'd want not to.
Was it right to not tell the victims of the Tuskegee syphilis experiment what was happening to them, because they would have refused and blocked the progress of science?
> If someone doesn't want to talk to you (or your bot), why not just leave them alone? It's not their job to be your guinea pig.
it is their job to take reservations over the phone
Yeah, so presumably their bosses will instruct them to take reservations from automated agents, and grandparent comment's fears that people will hang up on automated agents that identify themselves as such will prove unfounded.
Essentially, what people have realized is that Google is turning humans into just another API they can interface with.
On the topic of the asymmetry of the experience: is the larger issue that the fake John Legend voice will become boring to talk to for a human?
It's much easier to determine if there's too much traffic hitting a node or leaving a node than to determine if the traffic is "good" or "bad." I haven't seen anyone describe a way the system can generate "bad" traffic yet (that doesn't just degenerate into dumb games you could play without Duplex involved, like booking a reservation with your own phone and then just not showing up); generating too much traffic is a far more easily solved problem.
But on the topic of discerning whether the traffic is "good" or "bad:" this is actually a circumstance where the mass of data Google has access to that seems to so frighten people works to the advantage of the business-owner. Identifying bad users of the system could be done, for example, by collating data on reservations with geolocation data gleaned from maps to note there's a pattern of a particular user booking reservations they never use. Google can then just deny them access to the service to preserve the service's reputation.
A small snippet from an assistant conversation that is never going to happen :)
"That number is very busy. I can put you in the queue for the call, but would you like me to suggest alternatives instead?"
(which is kind of neat, and something not even a regular human assistant could do for me)
This particular problem of booking appointments is solved by e.g. OpenTable - it's not useful to solve the phone conversation part of this transaction.
https://www.blog.google/topics/next-billion-users/
1) You can be very impressed with the tech (as everyone is)
2) You can be creeped out by how it could be abused.
BOTH things can be true. And that, I think demonstrates a continued blindspot especially by companies like Google, who simply don't seem to realize when and why things may come across as creepy to people. They keep doing this almost every year. They're so focused on how cool the tech is, they seem to just not realize that people might find it totally creepy. And this seems to be deep in the company's DNA (same with FB)
Did they do a focus group of people outside the valley bubble? I doubt it. Maybe a focus group of receptionists? What do they think of it?
Assistant: "Hello, you've reached Dave's personal assistant. Dave's not available right now, can I take a message?"
Bob: "No, just tell him to call me back as soon as he can."
Assistant: "Okay. Can I ask who's calling?" (Bob was not in Dave's contacts list.)
Bob: "It's Bob Jones."
Assistant: "Alright, I'll let him know."
Bob: "Okay, bye."
Perfectly natural conversation, and provides a much nicer user experience for Dave when he checks his missed calls. (Notification: "Call from: 444-444-4444 Bob Jones says 'tell him to call me back as soon as he can'")
A big part of what creeps people out is the notion of bots impersonating humans.
Unless you mean impersonating a _specific_ human (e.g. if my personal assistant were pretending to be me) I don't really see the issue.
Like, a really smart parrot is not a person, right? Even if the parrot could be an answering machine, it would not be due respect and it would not respect you back; because it's a parrot.
I think the issue is that the answer-bot is inherently disrespectful to the real human. Just using one at all is like saying to the real person:"You are not worthy of the time and/or respect of a real human; talk to this clever metal parrot." Text devices and digital sign-ups are at least 'equal', though the transaction is sterile.
Trying to fake that a person is talking to a bot is disrespectful of that other real person and their time; it's trying to pull one over on them and fake them into thinking that the bot's owner is worthy, but that the other real person is not worthy.
Just use a sign-up service. If the company/person does not have one, they are signaling that you have to actually talk to them, and you should respect their decision because they are a real person too, not try to pull a fast one.
> Just use a sign-up service. If the company/person does not have one, they are signaling that you have to actually talk to them
Or that they can't afford to build an online order service? I seriously doubt the guy working the counter at my local pizza restaurant has any particular desire to speak to me personally.
I agree that forcing people to talk to a poorly designed phone tree with terrible voice recognition would be rude, but this isn't that.
Then it has to navigate through the maze.
This sounds great, but seems fraught with failure.
All of the successful Google innovations (Tensorflow, Alphago, Alphazero, map-reduce, millwheel, TPU) seem to have arrived with a proper publication style writeup. The ones that were less successful like Glass, Wave... not so much.
They haven't decided in how to make a product out fo this. This was just a tech demo of what could be possible.
Many famous academic programs and large companies do this. After generating the anxiety, the lab pairs it with intimate access to the research labs in the form of sponsorship, partnering and other ways of reciprocating.
Essentially, the labs establish themselves as the alpha dog in the system. You may be afraid of what's coming, but at least you can look to them to understand it. It all breaks down when you start to realize the same people creating the stress are the ones you offer to relieve it, not by slaying it, but by cutting you in on the deal.
And it's always difficult to interpret because the tech itself is 1st class, so the artisan class is awed and jealous, so they want to be part of it. Meanwhile, most of those critical of technology sound like complete wankers who want us all playing homemade instruments around a campfire eating raw seaweed. And Chris and Sal Average Family is still cheering for U.S. space dominance as conditioned by the space program since its inception, so they're rooting for technological progress in the form of narcotizing Narcissism (as described by McLuhan).
Sorry if I'm a downer, but I have been part of this system in the past and now I'm on in the edge in and out of it and I study lots of philosophy and media theory. This is how this system survives.
MOST people dont or dont care. The same with #deletefacebook.
There is a vocal minority including the press, and then there is everyone else who just wants cool things to work well and easily, and doesnt care at all about privacy.
The potential efficiency gains from this (think of calling into one of these instead of your traditional phone menus) are substantial. We should start looking into how to regulate AI (even if it just starts as "AI should identify itself as AI when they're using human communications channels") to make it work well for us.
Honestly, one thing your idea may show how far the big tech companies have strayed from their desire to create a concise and interesting product. Doing something like what you propose - basically, a system that calls, identifies itself as a robot and then is freed from the (difficult) technical problem of appearing to be human would be much easier to implement. With a bit of Twilio, "weekend hackathon to a workable demo" levels of easy to implement, IMHO. And, honestly, it looks like the public at large might have preferred that, and anyone using such a system to schedule an appointment would likely have a pretty much identical experience to the one provided by Duplex.
When I know I'm talking to a software agent, I can change the way /I talk/ without having to lose fidelity from the way it talks. For example, I can spit out a large, complex sentence that's usually harder over the phone: Groups of four don't need or get reservations. "Reservation availability is Tuesday 2 to 5 pm, wednesday 2 to 3pm, friday 4pm to 6pm. " Then a "Thank you, goodbye" message tells me that I'm done with them, and as I know it's an AI, I don't have to worry about having offended a customer.
That sounds like it will work as well as the federal do-not-call registry.
I think most of the issues could go away if we fixed the phone system to provide authenticated information about the source of the caller. If you can't or don't want to provide that information, that's fine, but I'm not going to answer that call.
Perhaps it'll be awkward, and fail. But so what, I'd rather give it the ole college try.
This may be an example of something where things need to be altered, or it may not. But in either case it seems short-sighted to just think this question can be resolved in 24 hours after a demo due to a few hastily-written tweets and thrown-together articles. What's important is that as AI continues to encroach upon human capabilities that there be a framework for navigating these transitions and knowing where the lines are. Hopefully as this kind of thing becomes more common public opinion will shift from caring about the creepy-demo-of-the-week to a wider scope of how we grapple as a society with these kinds of advancements.
Emotional reactions and ad-hoc decisionmaking based on mob rule is going to result in a random, arbitrary outcome which will almost certainly be suboptimal.
Learn from the past that you might not repeat it.
Dictatorships and despotism are the optimal solution for wielding of executive power.
Murder is the optimal means through which to immediately resolve problems created by other people from a time investment point of view.
Rape is an optimal solution to spreading genetic material.
See the pattern? Knee-jerk reactions from the layperson AREN'T necessarily bad. If anything, they should be the first sign that a researcher needs to back up from that awesome thing you just did, and take off the razor thin beam of relevance you used to achieve your breakthrough, and begin the HARD part of research. Turning it from an interesting academic factoid into a societal good.
When we do research, we are free to ignore huge swathes of the world to be able to reach or findings. The reckoning, however, comes when we have to ask ourselves 'how prepared is the world for this?'
The story of Doctor Frankenstein isn't just some scary story for children. It should be haunting a researcher night and day. What am I enabling? How will society take this? It shouldn't stop you by all means, but it will keep you grounded and aware of the fact your research isn't happening in a void.
Remember, all those bits and pieces you are putting out of your mind are the bread and butter of every layperson's day. Being a researcher doesn't absolve you of your status as a member of society.
It is YOUR responsibility to take your idea from the theoretical to the practical. The rest of society is busy staying intact enough to support your research efforts.
My point was that a suboptimal solution is one where there was no careful, methodical thought to the design of AI systems. If AI designers and researchers blindly react to public outcry after 24 hours from a demo (which, in general, is something I would expect Google to do) the kind of thinking you mention above is just as unlikely to happen as if they trudged on forward without any consideration to these things at all. In both cases, this is a suboptimal outcome for society.
In one, we get a fairly random, undesigned world of AI systems that don't serve anyone well and generally are underutilized because nobody is willing to push the boundaries. In the other, we get dystopian AI hell. It's important that researchers be given the space to think you describe. The right way to allow that is to foster an informed and open-minded public about the emergence of AI and the decisions we need to make about it as a society, and not have prominent voices writing about knee-jerk reactions making authoritative demands to a specific public demo after just a day. (See Max Tegmark's book Life 3.0 for the kinds of stuff we need more of being put into the world imho.)
>Emotional reactions and ad-hoc decisionmaking based on mob rule is going to result in a random, arbitrary outcome which will almost certainly be suboptimal.
I may have initially misinterpreted some context when I read your comment before, though my conclusions haven't changed much.
I agree with you that AI research must be methodically and transparently approached.
The part where I was trying to speak toward, however, was that AI researchers/designers are going to necessarily be constrained by how much faith and goodwill a (currently) troubled society, plagued with inequality and hardship in the face of increasing job loss to automation. This is a feature, not a bug to be dismissed.
AI is akin to nuclear technology in that it is perceived as a huge game changer. One that brings with it a huge potential for societal change and upheaval.
The AI designer/researcher DOESN'T necessarily see it that way. It is just a tool to do a thing. The populace, however, don't grasp the limits the researcher/designer does. They are being asked to give the same degree of faith to the AI field that nuclear physics was given in the last century.
Combine that with a severe distrust in the motives of the organizations doing the actual development and you have the stage set for large-scale rejection of AI research.
You said it yourself before. AI is a field that is trying to replace humans (having to do stuff). That is not a viewpoint that will win you the favor of the populace. No one will get further than the first half of the sentence before the hackles go up.
Nobody trusts a Utopia where people don't have to work because machines do it for them. Nor will they trust the men behind the machines. It's sad, but it is the kind of thinking that has kept humanity OUT of a complete dystopian hellholes thus far.
I don't know the fix, but I do know that if we can't solve some very fundamental social problems that advances in artificial intelligence and machine learning are poised to make worse, you will find the field of study becoming more and more difficult to find public acceptance in. The only inkling of a clue I have is it may have to do with the "industrial asymmetry" of our current times.
If you don't take heed of the state of society, you may end up in the same place NASA was for the latter half of the 20th Century. BEAUTIFUL theoretical groundwork laid, but hamstrung in execution by popular sentiment. (See some documentaries to learn how "thrilled" many were with the Space Race. NASA still hasn't recovered from the backlash despite some of the best science I've ever seen)
I can't see many real people wanting to automate calling the hair salon, because then you're being fairly pretentious as if you're too busy to pick up the phone.
But I can see a massive incentive for scammers to stop having to employ cheap Indian labour and now have an almost infinite ability to make human like calls to non-tech people in an American accent.
So be prepared to hear the Google assistants calling you to tell you there's a problem with your Windows PC and they just need to install something on there to fix it.
Just a short list:
- the deaf
- the mute
- the very socially awkward
In the last category: I can personally attest that I've put off scheduling an appointment with someone I've done business with in the past for two days, purely because the thought of actually picking up the phone and going through the process of aligning our calendars makes the hair stand up on the back of my neck. If I can pass that task to an assistant, my life would be much easier.
I'm not a fan of calling people either (I've put off finding a local doctor for a year because I don't want to go through the awkwardness) but starting a relationship with an important client using a not quiet real 'assistant' would start off the relationship in the wrong foot. Certainly if someone calls me using this I'll think that they can't even be bothered to speak to me personally and so I'm not worth their time.
> scheduling an appointment with someone I've done business with
So it's not just booking at a restaurant this is someone that you will have an ongoing personal relationship with.
And you know what? It's a really really good point, Duplex could help a lot of people get things done. You know what else? There are already services like that, and they always announce themselves. Making that cheaper and more automated, so it can benefit more people, seems like a fantastic idea. No need to turn humans answering phones into an API to do something good in the world.
I used to do phone support for an ISP in the 90s and I would get calls quite often from someone who would announce that they were interfacing between me and a text based system for a deaf customer. If a robot did the same thing, I think everyone would be fine with that.
You can. It's called Magic (https://getmagic.com)
Google Duplex would be an absolute boon for him.
From the other way I can see the sense, but the ones that are terrible don't get any better by having a cheaper option. It'll just mean they'll have a terribly programmed assistant instead of an Indian call centre where they've been given bad instructions.
I don't think this is a good argument. The same idea, applied to other domains, sounds outlandish: - you're too busy to drive to the store to buy something [using Amazon] - you're too busy to pick up a pen and write a letter [using email] - you're too busy to pick up the phone [using SMS]
The progress of technology is literally about removing friction from tasks which previously took longer amounts of time, and (except at the beginning, maybe?) this is rarely interpreted as pretentiousness.
Of the non-technical people that I've spoken to about Duplex (only 4, admittedly), they were universally excited by the demo.
As far as the proliferation of scammers, I think you're absolutely correct. I think it's plausible that human callers will incorporate challenge-response questions to check for things like this ("Who is Batman's sidekick?", "If I were unfairly comparing two fruits, what would they be?", etc).
Then again, who's to say that Google will make their tech available outside the company? The accuracy of the demos probably largely stems from the massive dataset that Google has amassed, which won't be available to scammers.
What's truly “horrifying” is the somewhat prevalent phobic attitude towards technological innovation and invention, be that industrial robots ‘taking all the jobs’ or this reaction to the Duplex demo.
Our current technological achievements were largely made possible with help from the media and our prominent thinkers in cheerleading the advancements and shaping the public’s option.
Now it's the critical voices that get affirmed and amplified.
The fact that we see a backlash about something that treats humans as an API, and don't see a backlash about something that helps all humans, seems like there is no prevailing innovation-phobic attitude. No?
The industrial robots taking jobs thing is at least pragmatic to discuss - we could just accept the capitalistic inevitability of it, or we could talk about ways to move the human workforce on to other tasks.
Skepticism and questioning are fine but the media is trying to turn people into luddites because they feel they were wronged by internet firms.
Ex. I do not mind making a 60 second phone call to make a reservation vs. 30 seconds of figuring out how to tell Google what to do.
In this case Google is solving problems for rich socially awkward nerds. Not the majority of people.
If that condition is met or exceeded, none of this voice stuff will matter. People won't care if the system discloses it's nature, or sounds creepy, if their airline reservation change (or whatever) goes abnormally well.
If that condition is not met, which IMO it won't be in the near future, then we're just adding attempted manipulation and potential for abuse on top of an already bad UX.
No. Replace AI with your favorite disadvantaged minority and the answer will be obvious. If AI is smart enough to pass Turing Test, it should be treated as a human.
That's the jarring subtext of Grue3's comment.
Because phoning in for restaurant reservations is only the beginning. Ask yourself now - in a few years, will you be one of the people giving orders to a digital assistant, or one of the (many more) receiving them?
"The future is already here - it's just not very evenly distributed."
Yes, of course. I would even say it should be legal fraud not to, with severe consequences.
Cheers
You implied the reasoning behind your statement was obvious ("of course") but it's not at all obvious to me.
No. The affirmative answer to the original question as posed is what seems to me to be obvious. My additional statement was meant to convey the seriousness with which I feel these matters should be taken.
> Why should it be legal fraud with severe consequences for an AI caller to not disclose the fact that it is an AI?
Otherwise people (specifically the technologists implementing and selling these kinds of trickster services) will not take other people's (specifically me and anyone who feels as I do) concerns seriously and proceed apace.
There is a similar issue around folks who see CRISPR as a way to work around laws that attempt to control GMOs. I have seen a few articles written by pro-GMO folks just crowing about how CRISPR-edited organisms can be sold without GMO labels due to loopholes in the law. These folks have their heads so far up their belief systems that they don't even see the problem.
A machine that attempts to imitate a human well enough to trick people into performing a specific action, an action that involves commitments of time and money.
I have a specific and a general objection to this. The specific objection is that this thing tries to trick people. It sets up a "frame" that is intrinsically deceptive. That's it's goal.
In another comment s2g quoted:
> Giving it an obviously robotic voice when it calls. "People will probably hang up," he said.
It is designed to trick people. That's bad. When you do it in business, we call it "fraud".
My general issue with this is that it normalizes the blurring of the line between real human people and ersatz Artificial Intelligence "people". This is something that's going to be a serious problem on a deep level. There's evidence of this sort of conceptual melting right here in this thread. How long before folks are trying to marry their toaster?
Last but not least, there's the meta-problem: the disconnect between the techo-elite innovators and the masses.
Think about this: As a businessman, when the phone rings and there's robot there trying to give me money why would I hang up? The robot only called me because someone wants to make a sale (from my POV) so that's the "opposite of a problem". This is the meta-problem: the technologists are so effin' clueless yet default to starry-eyed optimism.
"We are as gods, we might as well get good at it."
Conversely, save me from the folly of young gods.
It'd be like if I called a business and used a voice changer during the call to make me sound like I'm a different gender than I am. Would that be fraud? No, because my gender is irrelevant information in the context of that call.
Now maybe it might be a good idea to program the AI to be upfront about its nature just as a matter of courtesy, but I see no reason why it _not_ doing would be so harmful that it needs to be made illegal.
It is tricking people if it is done to trick people.
It is not tricking people if it is done while the person to whom it is done knows the thing they are talking to is not a human.
The question was not "should machines be allowed to act like humans?"
It was "should they be required to disclose that they are not human?"
You seem to think that I object to machines acting human, I don't.
> Voice assistants need to behave like a human to enable natural interaction
Would you say the computers in the "Star Trek" shows "behave like a human"?
> the fact that the AI caller isn't human is irrelevant information in most contexts.
How will it know if, in a given context, disclosure is relevant? One of my first questions about this Duplex service is, "Can I opt-out of receiving these calls?"
> It'd be like if I called a business and used a voice changer during the call to make me sound like I'm a different gender than I am. Would that be fraud? No, because my gender is irrelevant information in the context of that call.
I don't think it's like that at all. Are you impersonating someone else? Gender-bending for kicks? Trying to get better service because you think the person you're calling is sexist or something? Are you The Jerky Boys?
People do weird shit and that's something we all have to deal with because gravity.
Machines calling people and saying "uh" to pretend to be human without the human people realizing it is trickery.
And it's not even needed to make the Duplex service work because, as I pointed out before, when a robot calls you to give you money you don't hang up on it! Okay?
> Now maybe it might be a good idea to program the AI to be upfront about its nature just as a matter of courtesy, but I see no reason why it _not_ doing would be so harmful that it needs to be made illegal.
And now you have made my point for me: if it's not made illegal, people, like you, won't take these concerns seriously.
It's like the guys who owns my favorite cafe. He had a bit of plastic wrap covering his brown sugar and there were little flies sometimes and it was just a little gross. One day I came in and he had put a tight-fitting plastic lid on the sugar. I mentioned it to him and he started complaining about the health inspector making him do it. All I could think was, buddy you're the reason we have to hire and pay for health inspectors, right there.
AIs that can impersonate humans well enough to trick them into thinking they're also human are pretty much a munition. Thank you Google for opening the gates. Now that people realize it's possible we have another thing to guard against. You know most hacking is done though what's called "social engineering", eh?
That's true, but I don't see behaving like a human as "tricking people". Even if the assistant opened the conversation by identifying itself as an AI, I'd expect it to use all the same mannerisms "uh", "mhmmm" that it did in the call to enable natural conversation.
> Would you say the computers in the "Star Trek" shows "behave like a human"?
No, but I wouldn't exactly call that "natural interaction" either. Jarvis is a better example. And yes, I do think Jarvis "behaves like a human".
> I don't think it's like that at all. Are you impersonating someone else? Gender-bending for kicks? Trying to get better service because you think the person you're calling is sexist or something?
In keeping with the analogy, probably closer to "Trying to get better service". An AI callbot that pretends to be a human will be more efficient at its job than one which speaks in a monotone and requires the human on the other end of the call to adjust their mannerisms to compensate for the fact that they're not talking with a human.
> And it's not even needed to make the Duplex service work because, as I pointed out before, when a robot calls you to give you money you don't hang up on it! Okay?
Depends. If the AI immediately identifies itself as such, how many people are just going to immediately assume it's a spambot and hang up, even when it's not?
> people, like you, won't take these concerns seriously
What concerns? You've still failed to describe any ill effects that might arise from AI callbots failing to disclose their nature upfront.
> You know most hacking is done though what's called "social engineering", eh?
You think a law forcing AIs to disclose themselves as AI would do anything against malicious actors who are already breaking the law? (Hacking.)
I think there should be a kind of non-jarring way for AIs to identify themselves - maybe playing three quick tones before the AI speaks - to identify that it isn't a human on the other hand and therefore shouldn't be able to say, order thousands of dollars of food. This is quick, unobtrusive, easy for an AI to do, and easy for an AI to detect.
Doesn't that really tell us all we need to know? If the only way to get people to tolerate your service is to try and trick them then is your service a good thing?
An electric guitar doesn't play for you. It doesn't learn to play for you. It doesn't learn how you play and make your playing better.
It just provides new ways to make sound.
These are the questions we should be looking at, in my opinion. This situation is only temporary. We need to focus on the future (that is most certainly coming) ideas of what we're going to do with our societies when even these simple jobs that we thought required that human touch are replaced by automation and AI.
Oh sweet bloviating lord of hot takes, this? This is what's horrifying? Not the techno-utopians and techno-capitalists working hand-in-hand to shepherd all human interaction onto the internet while simultaneously turning it into a nightmarish complex of data burbclaves and dopamine farms? Not founders, VCs and politicians gleefully circlejerking about how profitable it is to create technological monopolies that evolve too quickly for society's immune system to regulate, like some kind of economic retrovirus? Not the left and the right working opposite ends of the great authoritarian handcart, on one stroke demanding that private entities exercise government-level control over digital spaces, and on the other demanding that they be free from government-level accountability?
When you get a call from a robot at work, the thought that it talks just like us should be approximately the fifth-scariest thing to enter your mind, followed by the thought that this means your job is going to get automated away soon, the thought that maybe it won't be and you'll end up a robot's page boy, the thought that what you're doing is so mundane that a computer can not only do it, but also completely model how you will do it, and finally the thought that this stunning waste of human life bothers you less than the possibility of being unemployed.
As far as things to feel creeped out about, I'd give robo-bookings maybe two stars: some flashy overtones but ultimately lacking in substance. I recommend a deeper, more full-bodied creep-out, like remembering that Facebook cuts people off from their friends if they don't provide a data-matching-friendly real name. Papers, please!
Oh, and spammers. At least one cashflow relevant demographic covered.
We engineers and geeks tend to get excited about solving a problem and sometimes forget the humanity of humans.
Portal had the voices of the turrets just right.
I would think that Google's executives shouldn't have a say in that question at all. As the chosen representatives of the people, that is a job for the Government.
This argument is hilarious, but I'll play along. First of all "racism" is about superiority, not about differentiation per-se. There is nothing inherently wrong about acknowledging differences, it is wrong when it becomes the sole reason for assuming superiority. That is the textbook definition of racism, so to answer your question: No, in my view it would not be wrong to force an intelligent entity that is not human to disclose this fact.
It matters so that my approach to dealing with this "intelligent entity" can be adjusted accordingly, as per my needs.
Why does it matter if the person chatting with you on the phone is made of meat or a super-being who's having thousands of similar conversations at the same time?
It matters because today we do not yet understand anything about the drivers and motivations of this "super-being" and before we are clear on that, I would like to know who or what I am conversing with. I am pretty sure you are referring to "Minds" from "The Culture" series, which a) is fiction, and more importantly for the sake of this discussion, b) Minds have a relatively well understood context of motivations, morality and history of mostly benign interaction with Culture citizens. The "Robocall version 2.0" touted by Google doesn't have that, and given Googles past track record, they don't really give a flying fuck about the general population. When it comes to the proprietor and owner of this technology today, we have been comprehensively identified as "the product" instead of "the customer" -- it is my right as a private citizen to know who or what I am interacting with, especially if it is a Google product.
Finally, whereas dealing with other humans can be understood on a base level of expected intelligence and wits, as well as shared identity, history and culture, interaction with "super-being" would by its' very definition be an unequal exchange. I have every right to know if my interlocutor had any kind of extreme advantage over me (you know, because it is a "super-being").
Or am I being speceist now?
In this case, the IA would disclose its nature and its lack of independent agency. It'd disclose the fact it's not considered a living being and not afforded any rights. The human, then, could engage in all sorts of abuse (and Microsoft had a lovely chatbot that had its "brain" damaged by trolls) that could leave lasting effects on its persisted state.
> I am pretty sure you are referring to "Minds" from "The Culture" series
You are completely wrong. Google's "Robocall 2.0" is nowhere near close a sci-fi worthy AI.