Slashdot Mirror

← Back to Stories (view on slashdot.org)

Should Calls From Google's 'Duplex' System Include Initial Warning Announcements? (vortex.com)

Posted by BeauHD on from the heads-up dept.

Yesterday at its I/O developer conference, Google debuted "Duplex," an AI system for accomplishing real world tasks over the phone. "To show off its capabilities, CEO Sundar Pichai played two recordings of Google Assistant running Duplex, scheduling a hair appointment and a dinner reservation," reports Quartz. "In each, the person picking up the phone didn't seem to realize they were talking to a computer." Slashdot reader Lauren Weinstein argues that the new system should come with some sort of warning to let the other person on the line know that they are talking with a computer: With no exceptions so far, the sense of these reactions has confirmed what I suspected -- that people are just fine with talking to automated systems so long as they are aware of the fact that they are not talking to another person. They react viscerally and negatively to the concept of machine-based systems that have the effect (whether intended or not) of fooling them into believing that a human is at the other end of the line. To use the vernacular: "Don't try to con me, bro!" Luckily, there's a relatively simple way to fix this problem at this early stage -- well before it becomes a big issue impacting many lives.

I believe that all production environment calls (essentially, calls not being made for internal test purposes) from Google's Duplex system should be required by Google to include an initial verbal warning to the called party that they have been called by an automated system, not by a human being -- the exact wording of that announcement to be determined.
UPDATE (5/10/18): Google now says Duplex will identify itself to humans.

17 of 276 comments (clear)

  1. Why? by LetterRip · · Score: 5, Insightful

    Why should they? There is no logical reason for them to do so. If the bot works as well in reality as it did in the three demos, thern there is no reason to 'warn' the person on the other end that it is a bot.

    Also if the bot can't respond it seamlessly hands off to a call service employee, so there shouldn't be any issues with the bot wasting the time of the reservation takers time.

    1. Re:Why? by The+MAZZTer · · Score: 3, Insightful

      Yeah, it is in Google's best interest to not announce it and to just pass it off as human if they can possibly do so. If the person is told something unexpected (if they don't know what Google Assistant is) they may very well get confused, or may insist on not dealing with it. Then Google Assistant has already failed ad that call. And if it fails at enough calls users will stop using this function entirely as it is unreliable. The reliability of Google Duplex requires the people it talks to to be just as reliable as the Google Assistant end. Sometimes the best way to accomplish that is to Keep It Simple. No need to communicate details that are ultimately irrelevant.

      Of course, as others have pointed out there might be legal aspects to this, such as recording laws and laws about robocalls. I can't speak to those.

    2. Re:Why? by rsmith-mac · · Score: 3, Insightful

      Why should they?

      Because getting jerked around by a computer sucks. Doubleplus so if it's one pretending to be a human.

      I legitimately feel sorry for service workers who are going to have to take orders from Duplex. It seems oddly dehumanizing to be ordered around by a machine.

    3. Re:Why? by Hallux-F-Sinister · · Score: 2, Insightful

      Have you ever called someone's residence, and had a small child answer the phone? The issue is that even if you can't tell it immediately, you're dealing with someone not smart enough to handle the phone call. Suppose you are with the power company, and you dial a household to advise them that a gas leak in their area necesitates their evacuation or to find out if they smell gas, and the person who answers is a 5-year old. Do you tell your boss that you warned them, knowing that the person you spoke with on the phone was a small child? Or do you, rather, ask kindly, "is your mommy or daddy home?" etc., to speak to someone you can reasonably presume to be competent to get everyone to safety or intelligently answer your question about whether or not there is the telltale aroma of a gas leak there?

      Until someone comes up with an AI that can beat every Turing Test thrown at it, because it actually IS intelligent, I'll thank the people programming them NOT to try to get them to pretend they're "smart," and by the way, they're not. The machine is NOT passing a Turing Test if no one is actually ADMINISTERING one. To suggest that one HAS done so is like suggesting that if someone rings up a private residence, and a legal child answers the phone, (meaning someone under the age of majority, in the US of A being under the age of 18 years,) and the person calling can't tell that he or she has NOT in fact spoken to someone who is legally an adult, that the child who answered should be CONSIDERED an adult for all purposes, and be allowed to act in his or her own stead as a legal adult, i.e., buying a car or realestate, signing up for the armed services, or entering into other legal contracts, purchasing tobacco, etc. That's just silly. Being able to fool SOME people who weren't looking out for it does NOT prove general competence, any more than some man who puts on a dress and makeup convincing a casual passer-by, or indeed even several, or many, that he is a woman MEANS that he is, and should be considered a woman henceforth.

      I have myself encountered this sort of thing dialing into Apple's tech support and run into their very human sounding AI, and it's super-frustrating because you have to convince this goddamned stupid machine to LET you talk to a person, when it's designed to try to prevent it. In some cases, it works well, because you have a simple question you can concisely state, and it's PROGRAMMED to understand: "Hi," it says, "I'm an automated system capable of understanding full sentences. Please tell me what you're calling about." If you can reply, "My Apple TV stopped working," it might reply, "you're having problems with your Apple TV. Is that right?" To this, you can say, "yes," and it says, "okay, I'll get someone to help you."

      But every now and again, your problem is not one it's programmed to grapple with, because it's uncommon, and even a fairly stupid human would get it. For example, if it asks, "Please tell me what you're calling about." and you reply, "I can't get my iPod out of my anus!" or something, I don't think it would know what to tell you. A human would probably ask you to confirm, "did you say your iPod is... I'm sorry, it sounded like you said it was stuck in your... anus?" and you say, "yes, I thought it might be funny to shove it up my ass, and now it's stuck," he or she would probably helpfully suggest going to a hospital, for help having it removed, and maybe even provide tips on how to clean it to minimize the risk of water damage resulting from the moist environment of your rectum. You think the AI could do that? In truth, having no iPod, and never having shoved one up anyone's ass, let alone my own, (nor anything ELSE, for that matter,) I can't say whether or not the AI they have answering the phone would or would not be able to help in that scenario, though I have a sneaking suspicion it would NOT.

      So if they reach a point where they are good enough to fool people who aren't looking to see if they're human, I'

      --
      Our reign has gone on long enough. Indeed. Summon the meteors.
    4. Re:Why? by thegarbz · · Score: 5, Insightful

      Because getting jerked around by a computer sucks.

      Implying you're getting jerked around. If this computer is no different than a human, then hang up on them if you're being jerked around. Or maybe they are making an appointment with you for their owner.

      Do you hate secretaries too?

    5. Re:Why? by cyba · · Score: 5, Insightful

      > I legitimately feel sorry for service workers who are going to have to take orders from Duplex.
      > It seems oddly dehumanizing to be ordered around by a machine.

      Soon all these service workers will be replaced by Duplex (or its competitor) as well, so it will be only some other AI that will get dehumanized :-/

    6. Re:Why? by AmiMoJo · · Score: 5, Insightful

      That was my reaction too, but on reflection it's actually quite complex.

      Where do you draw the line? What if it's someone with a disability using assistive technology to set up an appointment. Should they be required to disclose their disability to you in order to get permission to use a digital assistant?

      We are a long way from strong AI, but the parallels with how certain groups were treated in the past is striking. Some people expect trans people to declare themselves and their anatomy up front, for example.

      If this stuff doesn't matter then we have to ask if AI vs. human matters, or if other things like politeness and efficient and effective communication are.

      --
      const int one = 65536; (Silvermoon, Texture.cs)
      SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
    7. Re:Why? by Wrath0fb0b · · Score: 4, Insightful

      I legitimately feel sorry for service workers who are going to have to take orders from Duplex. It seems oddly dehumanizing to be ordered around by a machine.

      I spend all day being fed bug reports from a machine. And yet even though the message was delivered to my by our bug tracking software, it ultimately originated from the intent of a human being.

      I spend all day responding to emails (and posts on /.) delivered by a machine. But I assume that you are not a bot and that even though this interaction was mediated by machines, it serves our common human purpose.

      When I worked in food service, I spent all day being ordered to prepare food by tiny slips of paper with horrible handwriting on one of those turny-things. I hope that the food being made was ultimately consumed by humans though.

      At the end, you seem to be arguing that it is more dehumanizing to be relayed orders by a machine that emulates flapping meat sounds in meat-English as opposed to receiving those orders by reading off a computer screen or on a slip of paper. Perhaps you are right (after all, this is subjective) but it seems that the crux of your claim is that voice is different, not that you are 'ordered around by a machine'.

    8. Re:Why? by mysidia · · Score: 5, Insightful

      Right.... I just want a law that any chatbot used on the telephone must NOT attempt to lie or deceive and MUST answer truthfully, in particular, when asked questions about the call or itself, Must answer all such questions to the full extent without hanging up the call, attempting to misdirect, or attempting to transfer the call to another person or line, such as "Is this an automatic call?", "Do I have an account with your company?", "Is this call a solicitation or sales call?", "Are you a chatbot?", and, about the company that made or accepted the call, and about the chatbot operator's client whom they are making the call on behalf of, and the listing of any chain of 3rd parties engaging the client --- the Chatbot must provide their complete name, Addresses, and Registered agent names and addresses for All upon request.

  2. They all have the same name by lazarus · · Score: 5, Insightful

    When the bot called the hair salon it started the call by saying "Hi, I'm calling on behalf of a client and would like to book..."

    You can solve this problem by changing this to: "Hi, this is Alexa (or Google whatever) calling on behalf of a client and would like to book..."

    This will take the masses about 30 seconds to adapt to and we can dispense with all the drama. At this point there is no need for them to have different names.

    Sometime in the future when they're sentient and want to talk to each other that will have to change.

    --
    I am not interested in articles about life extension advancements.
    1. Re:They all have the same name by Dutch+Gun · · Score: 5, Insightful

      Anyone who's worked in retail would probably tell you they'd prefer the bot.

      --
      Irony: Agile development has too much intertia to be abandoned now.
    2. Re:They all have the same name by Bongo · · Score: 3, Insightful

      A business might have a policy that they neeed to talk to a real person. Automated calls could be the result of malware. Someone could DDOS a small business, filling their booking with fake entries for weeks.

  3. Self-importance alert. by ColaMan · · Score: 4, Insightful

    I believe that......

    And I don't. Seeing as we're both not Google, our opinions on this topic are pretty much moot.

    --

    You are in a twisty maze of processor lines, all alike.
    There is a lot of hype here.
  4. Hangups by sgunhouse · · Score: 3, Insightful

    I don't know about you, but if I get a call from an automated system I just hang up. If the call starts off immediately by saying it is automated I'm sure that is what will happen. The first thing that has to happen is to indicate to the recipient what the call is about; after that they can say (especially if there is a response the system doesn't understand) that they are a machine.

    1. Re:Hangups by LetterRip · · Score: 5, Insightful

      At first thought, anyway, this "Duplex" thing rather annoys me. If the "person" on whose behalf the bot is calling doesn't feel it is worth their time to speak to me directly, why should I have to waste my time talking to their bot?

      This bot makes reservations. As an employee of the company that the reservation is being made at, you wouldn't be "wasting your time" - you would be doing your job.

  5. Re:cut out the people by Dutch+Gun · · Score: 4, Insightful

    If a Duplex realizes it's talking to another Duplex, will it go into 56K modem mode and just talk electronically?

    --
    Irony: Agile development has too much intertia to be abandoned now.
  6. Observer Effect by mykro76 · · Score: 3, Insightful

    That would defeat the purpose of training the AI on genuine human conversations. If the recipient knew it was an AI calling they would be likely to change their behaviour such as talking in shorter, simpler sentences with overly exaggerated pronunciation.