> Having a voice assistant means I don't have to spend a few seconds typing, at the cost of having everything I say potentially overheard by someone who I'd rather didn't hear it.

You're underplaying it in order to artificially inflate your mileage.

Having a voice assistant means not having to learn to change application contexts inside an operating system follwed by a few seconds typing. You are ignoring the hundreds/thousands of hours you've spent learning which software is appropriate for which contexts. Not to mention the latency of instantiating them, or possibly even installing them, migrating data between them, etc.

Voice assistants have contexts but they map fairly closely to natural language. And non-technical users can teach other non-technical users what those contexts are. "Here's how you set an alarm." I've never used Alexa but I'd bet it's something like, "Set an alarm for blah," or even, "Wake me up at blah." And I bet if I asked an Alexa user how to do that I'll remember because it's all natural language. That's a huge win for usability.

Now try teaching a non-technical user about cron, its flags, and its relationship to other shell commands, using natural language. In a decade that bullshit will sound more antiquated than the amplitude modulation on the voice of a Dalek.

>Having a voice assistant means not having to learn to change application contexts inside an operating system follwed by a few seconds typing. You are ignoring the hundreds/thousands of hours you've spent learning which software is appropriate for which contexts. Not to mention the latency of instantiating them, or possibly even installing them, migrating data between them, etc.

So, it's interesting. I resisted the iphone-style smartphone for a long time (I loved my nokia communicators) - I mean, I was super into wearable computers; I owned an Xybernaut at one time; I had the palm v with the cdpd modem... but I resisted the winning form factor for so long.

And... yeah, it was a mistake. Not really a huge one, 'cause the things are designed to be easy to learn over being easy to use (there's usually an either/or there. the easiest to learn process is rarely the most efficient process) but still, a cost; to this day, I'm not all that great at typing long messages on a phone keyboard, like young people seem to be. For that matter, I didn't buy a car GPS until far after they were common, and I am unusually bad at reading maps for a man who started driving before consumer GPS was common, so I would have saved many hours and more than one job interview if I had that technology sooner.

I kind of feel the same way about voice control. I mean, essentially it's a command line interface with a wonky input method, no? and it's a command line interface designed for regular people, not people who are familiar with the command line tradition. I don't think it's super useful now, and I kinda think it will be a fad, like the '80s talking cars.

But, I could be totally wrong. What am I missing out on by resisting the voice controlled garbage?

I mean, if someone does come up with a google maps level 'killer app' that makes you way more effective if you use the voice control, I'll be behind the curve.

About this:

"I mean, essentially it's a command line interface with a wonky input method, no?"

This is precisely the feedback we got from salespeople, when we were working on a Natural Language Programming interface for Salesforce. Initially, I got angry and denied the comparison. But after several people made the same comparison, I came to appreciate how true it was. Unless NLP is perfect, it is really just a Command Line Interface with an awkward input device. I talk about this a little towards the end of this story:

https://www.amazon.com/Destroy-Tech-Startup-Easy-Steps/dp/09...

NLP will only continue to improve in the years ahead. I've got a few echo's around the house, and yes the current implementation of NLP is not perfect. It requires some modification of how the user speaks, but the first time you say "Alexa what's the weather today?" and Alexa comes back with the day's forecast is a jaw dropping moment (at least for me it was). I believe voice activated devices will ultimately be placed all over the house and we'll communicate with our computers for a lot of various tasks through voice - it's just natural for humans. Think of all the mundane tasks you do typing into a computer - setting calendar appointments in outlook for example - and then think how many mouse clicks and typing are required. It's so much easier to just tell a computer using a voice command.

The privacy issue is easily solvable IMO. Google already requires a physical mute button on their devices to disable the mics in hardware not just software. It will require either user trust (admittedly in short supply with tech companies lately) or some government regulations but there is no reason even today for devices to have to "record" your conversations in order to process. As long as you require a "wakeword" to tell the device to start listening to your next commands the device doesn't need to record all audio all the time.

Have you ever written an Alexa skill? Because as soon as you try that, you realize the limits of the slot/intent system that Amazon is pursuing. You will notice this especially when it came to names and letters. We built an Alexa skill that allowed sales executives to use an Amazon Echo to ask Salesforce questions such as "Who was my best sales person last month?" That worked great, but the executives we talked to mostly wanted specifics about particular cases. And Alexa could not figure out the company names. These tripped it up badly:

Avon

IBM

Sinopec

Volkswagon

Alexa could not get these, which made the skill nearly useless.

It is much easier to simply use a keyboard.

And please consider the failed promises we've heard over the last 3 years. Remember "Invoxia will enable Alexa to tell which person is talking":

http://www.smashcompany.com/technology/invoxia-will-allow-al...

One as limited options in terms of forcing Alexa to pause in certain parts of a word:

http://www.smashcompany.com/technology/using-a-glottal-stop-...

I wrote this when I was still optimistic:

http://www.smashcompany.com/technology/how-to-build-conversa...

>You are ignoring the hundreds/thousands of hours you've spent learning which software is appropriate for which contexts.

Please elaborate. What do you mean by "thousands of hours of learning". Aren't you overplaying it to fit your narrative?

Hundreds to thousands of hours seems completely reasonable to me.

Have you ever done IT for your family or friends, or seen a coworker completely helpless in the face of an unrecognized file extension? Say they're, I don't know, an engineer or a call center operative or something. 6 hours of computer in a working day, 250 days of work a year, maybe they got out of college in 2005... more than fifteen thousand hours in front of a computer, and, well, https://www.youtube.com/watch?v=fa9DLxDtPtc. Computers are mindbogglingly complicated.

And then there's us, the people that hang out on HN. I can't even think of how many times I've spent fifteen minutes digging around in config files or menus or googling for a better tool. I've probably spent over a hundred hours just on my emacs setup. Easily hundreds, perhaps thousands, of hours of focused effort dedicated solely to high-end computer-use skills, and that's not even counting the things I use the computer for. And I'm still learning things every day.

I agree with you. I know more than one who bought a new notebook because the operating system didn't work and they don't even realize that reinstalling is an option.

Continuous reinstall used to be the way to keep stuff working. It was awful.

Everything a voice assistant does could be done with an device that has nothing but a keyboard and a text input field. People are not complaining about the great natural language comprehension these things have, they're complaining about always-on microphones in their homes.

But even with typed instead of spoken commands, wouldn't the natural language processing still done in the cloud? Admittedly, typed instructions instead of recorded sounds is less intrusive but still intrusive because it requires the internet, not processed locally.

If it was implemented using something like hitting super/the Chromebook search button and then typing your intentions, it would be a more deliberate choice to send the data than it is using a voice assistant which is constantly recording to look out for the keyword.

yes - a physical interlock would be infinitely preferable.

Except that voice assistant are not self-discoverable (in addition to being very unreliable and sluggish, which granted could be improved over the long run). And it forces me to memorize commands. Not just "switch on x". But like I want to play that song from that band I don't remember the name of but I remember it is a blue album a few songs after that radiohead song in my playlist. I don't want to have to memorise all these names.

I'm with you, until you lost me on:

> Now try teaching a non-technical user about cron, its flags, and its relationship to other shell commands, using natural language.

I'm just not sure what you mean with this statement. Are you talking about setting up cron jobs via voice? Or are you just making an analogy about the (perceived?) complexity of cron and teaching it to a non-technical user?

I would assume they are saying that it's a whole lot easier for the average person to say:

"Please remind me every day at 5pm to water the plants", or "Please play this song at medium volume every morning at 7am"

than to learn how to do the equivalent in whatever operating system their device uses, and I'd tend to agree. Cron is certainly a much higher bar than learning to use the alarm app in your phone, but the point remains, particularly in a home context.

But who's using cron to set a reminder or an alarm on their phone?! That can't be it, but if it is, then they are ironically overstating.

I mean you can use the GUI (launching apps etc.), or literally just type "remind me every day at 5pm to water the plants" into the home screen search bar (on Android at least).

The voice recognition and the assistant parts are decoupled.

You two are talking about different groups of people. crontab/Linux/Win: definitely needs learning smartphone: a little bit easier to use, open an app and everything is set visual assistant: out of box for being ready to use.

I taught one of my family members about using PC and Android, and found that almost everything needs to be explained. For example: - What's Google/Microsoft for? (Yes, even the giants in tech may never be heard for some people and it's common) - Browser? And there are many of them? (Not everyone has a computer and Internet access) - URL? What is that? I cannot even spell it! (This is more complicated and requires people to remember the addresses of tens of commonly used websites, btw, what's a bookmark!?) - register? with what? bank account or passport? (if someone doesn't have a email account, it's likely that they have no idea about the process of register, activation.) - App market? wait a second, what's an app? (not to mention built-in ones and those from 3rd parties)

The thing is Alexa, google home and many others provide a unified, easy to use substitute(well, I have to admit these devices are pretty dumb these days. So they are not 100 percent alternatives). The only requirement is being able to speak in a clear way, which is true for almost everyone.

PS: I don't have any of these devices. As a hard-core programmer, my choice is, without any doubt, *nix, shell, etc. GUI sucks ;-)

> Having a voice assistant means not having to learn to change application contexts inside an operating system follwed by a few seconds typing

Are you saying that voice UI is not its own "application context"?

It's an undiscoverable mess.

It doesn't take "thousands of hours" to learn how to launch a web browser and type some words into a search engine box.

Then why do so many people not do it?

> Having a voice assistant means not having to learn to change application contexts inside an operating system follwed by a few seconds typing. You are ignoring the hundreds/thousands of hours you've spent learning which software is appropriate for which contexts. Not to mention the latency of instantiating them, or possibly even installing them, migrating data between them, etc.

I can type everything I'd issue as a voice command into my home screen. I have to install and learn just as few/many things and commands as I'd have to using voice recognition. Op might have underplayed something, you just made things up.

How do you manage to do the typing when you're in the bath?or stepping out of the shower? or when you've left your laptop in another room? or if someone else is using your laptop for something important? The point is not that the voice commands are easy to learn or remember, it's that you have with you at all times (even if entirely naked, hey it's your house) the equipment needed to issue these commands...

>I can type everything I'd issue as a voice command into my home screen. I have to install and learn just as few/many things and commands as I'd have to using voice recognition. Op might have underplayed something, you just made things up.

You are not the market that these are targeted at then. The normal person cannot do those things.

A normal person cannot type an English sentence? Wow, I really seem to he disconnected from normal people.

I’m sympathetic to this view. I’m not sure it will stay that way. There was a time where internet access on your phone was an extravagance. It would be difficult to appreciate the utility before living with it. If what we are talking about is a difference in perceived utility, I’m not sure how big of a difference that really is.

Utility of VAs are going up. Comfort with them is a normative thing, and history tends to point to the norms changing.

As others have said here, you're overplaying one and underplaying the other:

1. having a map on your phone doesn't necessitate gps; only automatically locating yourself on that map does, which is an added convenience as opposed to the core functionality.

2. even the core functionality of a map—navigation—is a rarer one than the plethora of things one can connect voice control to

3. apart from taxi integration, probably the most common multiplier of the applicability of gps is voice: "find restaurants near me", etc.

In fairness, even with the above, your comparison is probably still correct: gps represents a bigger value add relative to the trade-off, but I just think they're much much closer than you make out.

> find restaurants near me

How does that help me at home? I don't dispute the value of voice recognition on a phone. But we're talking about smart speakers here, in your home, always on.

(BTW, on a phone I can set up common queries like "find restaurants near me" as shortcuts and access them with a single touch.)

> gps represents a bigger value add relative to the trade-off

Yep. Also, I can easily turn my GPS off, restrict apps from accessing it, etc. Home smart speakers are necessarily always on.

Aside from the usefulness: True, my GPS data can be used to trace information about me (where I eat, who I meet, which doctor or prostitute I visit etc.) however I believe that conversation I have attached home are more private and reveal notably more about me than those things derived from GPS as I might discuss the exact condition and state or explicit preferences in other regards.

Much more likely to want to discuss things in your home that are outside the Overton window and don't want recorded than accidentally go somewhere that you wouldn't want people to know about.

Is typing hard for you to do?

You totally underestimate the utility. Getting a voice assistant was life-changing for me. GPS comes nowhere close.

I hate talking to computers, so I don't see a voice assistant in my future ever. I'm in my early 50s by the way, and have been coding since I was in high school.

Maps/navigation is the only reason I even have a smartphone.

> Getting a voice assistant was life-changing for me

Can you expand on that? In what way?

Compare the workflows for leaving a reminder or note. Before:

    * Find phone/computer
    * Unlock
    * Locate self in interface and navigate to application selector
    * Invoke note-taking or reminder app
    * Type in note or reminder
    * Navigate through tagging/scheduling interface
    * Commit note/reminder

Or:

    * Locate notebook
    * Locate writing utensil
    * Open notebook to current page
    * Write note
    * Remember to check notebook every thirty minutes for the rest of your life

After:

    * Be in shouting distance of phone or assistant base station
    * "Okay Google/Alexa, note to self/set a reminder for eight pm: <blah blah blah>"

For most people the "before" is a minor inconvenience. These people don't really need organizational tools anyway. For people with memory problems the "before" is a rock wall that renders the vast majority of organizational technology functionally useless. You know how sometimes you walk into a room and realize you have no idea why you needed to be in the room? Or how thirty seconds spent answering email can end with all the the eggs you were juggling crashing to the floor and it takes you half an hour to get back "in the zone"? Imagine that, but it happens at the drop of a hat, five times an hour, in the time it takes you to say good morning to someone... or because you had to context-switch to execute the motion to unlock your phone.

There are lives that can be changed by organizational technology. But exactly this population was the one least well-served by classical note-taking and reminder apps. That "before" workflow was the six-inch drop at the end of the wheelchair ramp [1]. Voice assistants are a game-changer.

[1]: https://i.redd.it/x5qhwm2blyl01.jpg

    * Find phone/computer
    * Unlock
    * Locate self in interface and navigate to application selector
    * Invoke note-taking or reminder app
    * Type in note or reminder
    * Navigate through tagging/scheduling interface
    * Commit note/reminder

I just now took a stopwatch to time how long it took me to do all of this: 30 seconds, and 20 of those were walking to my office. That is not a negligible amount of time, but it's not a huge cost either. But...

    * Be in shouting distance of phone or assistant base station
    * "Okay Google/Alexa, note to self/set a reminder for eight pm: <blah blah blah>"

That's what happens when everything goes right. But based on observing friends who have Alexa, things only go right about two times in three. The rest of the time Alexa gets something wrong, often with comical results. The time it takes to recover from those situations can be a lot longer than 30 seconds. Often my friends just give up and drag out a laptop, or, more often, just decide that whatever it was they were trying to do really wasn't that important after all and give up.

I think the key difference is the focused attention and brain bandwidth to spend. Traditional way is way more efficient but requires more of that - you need to do a relatively long sequence of actions, coordinated to the goal. Using voice, there is no sequence, and it doesn't matter how many times you need to try and to repeat the same simple phrase - you spend no additional bandwidth with every repeat.

The voice way is like writing the simple add-mupltiply loop in C, and UI way is like writing the same in assembly (using vector extensions). Assuming you know how to do it very well, writing assembly code for that loop wouldn't be even longer to do. And you don't have a 100% chance the compiler will do the right thing with unrolling and vectorizing in this particular case, but most of the time you check the output and try another compiler options until it do the right thing, and only after a bunch of tries you finally resort to assembly. That's basic energy-conserving optimization present in people's firmware in their brain.

Yep. A garbled duplicate of a reminder or note doesn't cost anything and even alarms aren't a major issue unless they're set for the middle of the night, so if anything hiccups I just retry.

Your analogy touches on another important factor, which is that the human brain has a titanic amount of hardware dedicated to accelerating and reducing the overhead of voice. It's so good that it even participates in basic cognition, so most of the time the note you want to take is already right there and properly formatted for speech. There's a reason that "thinking out loud" is a thing. :P

Although to be fair, this is an emerging technology which should improve over time.

Anyone using a voice assistant now is almost certainly at the forefront of the technology. In five years or maybe ten they'll probably be a lot more impressive.

I've never had Google Assistant fail on something that trivial.

And as the type of person who has always struggled with calendars and notes apps, it's really valuable to me.

> * Find phone/computer * Unlock * Locate self in interface and navigate to application selector * Invoke note-taking or reminder app * Type in note or reminder * Navigate through tagging/scheduling interface * Commit note/reminder

This is just bad UI. Voice assistants and search have been used to paper over a serious decline in UX over the past decade or two. PDAs from the late 90s had much better note taking workflows.

Or (android example)...

Physical:

----

  1. Find phone (Is this really an issue? Haven't they become an extension of our body?)
  2. Unlock phone with thumb (unless of course you're already using it).
  3. Type "make a note to do the thing" into the big text box at the top of your screen.
  4. Preview / confirm it's what you wanted.
  5. Hit "Checkbox" to confirm.

By Voice:

---------

  1. No need to find the phone (assuming it close, which I'd say is a fair assumption).
  2. Say "OK Google".
  3. Say "Make a note to do the thing".
  4. Depending on current coverage, wait while it goes to the cloud to figure out what you mean (annoying).
  5. Listen while it tells you what it did.
  6. Say "Yes" to confirm.

My point is that the equivalent workflow exists via typed text. The workflow itself is not coupled to voice.

That being said, I do agree that it's very convenient to use voice for these sorts of tasks. Especially when driving.

I just walk over to my fridge and write it down with the dry-erase marker stuck to it. No issues with voice recognition misunderstanding, and it's trivial to do something more complicated like add to a specific list.

If it worked, yes, absolutely. On my Android phone, however, the note-taking process is as follows:

"OK Google!"

wait 5-10 seconds for the assistant to chime. It might work if I don't wait, but mostly doesn't. Sometimes it doesn't wake up at all.

"take a note"

wait for the assistant to ask what the note is

"Remember to put your shoes on!"

Check whether the note was taken correctly

Also, if I was actually using the phone for something else:

switch back to the app I was using when it occurred to me I should remember to put my shoes on

That's a bit odd. What version of Android are you on, does it have any of those junk manufacturer skins, and how old is the phone? If my phone is already unlocked I can spout off the hotword, invocation, and content in a single stream, the assistant pops up in an overlay and prints the note for me to check visually, and then it finishes and hides with no further intervention. Pixel 2 XL, fully patched, it's worked like that since I got the phone.

If you're having trouble with getting the assistant to wake up, and you don't see an improvement after running through the wizard you get when you say "retrain voice model" or "recognize my voice", you should also be able to get the assistant up by long-pressing the middle button in the navbar at the bottom. You additionally shouldn't need to wait for a response after saying "take a note", just dump it all in a single breath.

> when it occurred to me I should remember to put my shoes on

...Are you mocking me? I thought I'd made it pretty clear that I used that tools to help compensate for a memory problem that does, in fact, qualify as a disability. Try "okay Google, remind me at eight pm to reply to $friend about $thing". Or "okay Google, remind me every day at two in the afternoon to make a dentist appointment...".

Sorry, I definitely didn't mean to mock you! It's just that lately, I feel like I might just forget to get dressed and that's why I'd like to use voice assist to build a to-do list. Too many things going on at once.

Yeah, this is the "killer app" for voice for me. It's the only thing I use Siri for on my iPhone, although I do have that to require a long press instead of wake-on-voice.

For people that are already into the IoT space, I think it's a no-brainer. I have smart bulbs (Lifx). They look great but have little to no support for physical switches, and the app works OK until you want to have two phones controlling them. Suffice to say, voice control has made the experience 10x better especially for guests. Now, we also have a Roomba and some other smart things. Rather than get a new app for everything, it's all controlled by once interface.

> ather than get a new app for everything, it's all controlled by once interface.

Makes sense.

But why does that interface have to be the clumsy, always listening crap that we see today?

And why can no one except Apple[0][1] give any guarantees with regards to what they use my data for?

[0]: No, I'm not a Apple fanboy. I just can't stand their UX, seriously, which is sad since I value their current stanace on privacy.

[1]: And for what it's worth, Apple seems happy with selling out if the alternative is leaving the Chinese market behind as documented elsewhere in this thread. Although I'll admit that from what I read they where up front with their Chinese users about the change.

The simple fact I can call people while my hands are on the wheel is a non-trivial improvement in my life.

Basic voice dialing has been a thing for at least 15 years. Its progressively gotten better, and alexa et al. are no leap forward. The old sytems, while perhaps flawed, at least did not send recordings of you anywhere.

I'd be fascinated to hear how.

I can imagine things a digital assistant could do that I'd value quite a lot, but none of them are done by what's currently on the market. (And most of them would involve heavy smart-home integration, which makes their privacy failings even worse.) Meanwhile, the offerings I've actually seen so far look like their maximum utility would be converting tablespoons to cups while my hands are messy from cooking.

Yes, hence: YMMV. (Which, in case you don't know, is an acronym that stands for "Your Mileage May Vary", which in turn is an idiomatic expression meaning: a reasonable person could disagree.)