Slashdot Mirror
Alexa and Google Assistant Have a Problem: People Aren't Sticking With Voice Apps They Try (recode.net)
Amazon Echo and Google Home were the breakaway hits of the holiday shopping season. But both devices -- and the voice technologies that power them -- have some major hurdles to overcome if they want to keep both consumers and software developers engaged. From a report on Recode: That's one of the big takeaways from a new report that an industry startup, VoiceLabs, released on Monday. For starters, 69 percent of the 7,000-plus Alexa "Skills" -- voice apps, if you will -- have zero or one customer review, signaling low usage. What's more, when developers for Alexa and its competitor, Google Assistant, do get someone to enable a voice app, there's only a 3 percent chance, on average, that the person will be an active user by week 2, according to the report. (There are outliers that have week 2 retention rates of more than 20 percent.) For comparison's sake, Android and iOS apps have average retention rates of 13 percent and 11 percent, respectively, one week after first use. "There are lots of [voice] apps out there, but they are zombie apps," VoiceLabs co-founder Adam Marchick said in an interview.
It's simple physics. Voice requires more energy than a few taps on the screen.
The problem with these platforms and applications is their usefulness. More specifically: they aren't very useful. These voice-powered applications will be nothing more than a novelty until they can actually do something that would take more than few minutes to do yourself. When I can say from my couch "Alexa, make me a steak, medium rare, and bring me a beer, IPA" and a robot hands me a beer in 1 minute and a plate with a hot steak 18 minutes later, I'll give a shit and I think other consumers will, too.
Do you mean to tell me that gimmicks have no longevity?
Craziness.
Mod me down with all of your hatred and your journey towards the dark side will be complete!
I think maybe this problem is due to the novelty effect where it seems really cool to try it out a few times but after a while it doesn't seem like it makes life easier. Let's say you voice activate your lights despite having a light switch. I'm going to guess most of us have the light switch memorized so we'd hit it on and off even without looking or in the dark so changing to a voice activated system would likely slow you down. If you look at systems like Nest, they roughly figure out when you're home or not and then automatically adjust the heat and cooling to suit you. I'm sure the novelty would wear off if you had to tell it every time.
What they really need is "star trek" like sliding doors when it knows what you need before you even realize it. That would be awesome.
We spent decades tweaking the graphical user interface to make it easy and efficient. We have very little interface design experience with voice.
There is also a latency issue, at least with Google (no personal experience with Amazon, but I assume the same). That processing delay may be small on average, but it is extremely annoying---most especially when the internet is less than perfect, but also when it takes a very long time for no apparent reason.
Some feedback, like status indicators for internet and background noise may help.
The interface needs to mature. I don't think I can predict what that will look like. It is already extremely accurate, probably better than a human transciptionist, so this is more of an integration issue than a technical problem.
---
According to the latest ruleset, this post should be modded as Vorpal Flamebait +5.
The issue is that most voice apps suck. For example:
Me: Set an alarm for 8 PM
Siri: Calling Dave
Currently, these so-called "intelligent assistants" are little more than toys. You play with them for a couple of days, they are kind of fun, but then the novelty of the toy wears off. They are OK when it comes to rather specific questions, and all but useless for issues that require a minimum of intelligence.I expect that, one day, they will live up to their name. As of today, they are toys.
Of all the possible uses of Siri, "Siri Stop Navigating" when you are trying to pull into a parking lot at your destination and she won't shut up about making a U-turn is about the only use that we've found yet. Voice is great for a minuscule number of real life situations.
The only thing worse than a Democrat is a Republican.
I am almost never in a place where I would feel comfortable talking aloud to me phone. Should I get up and go somewhere private to talk to my phone, or just stay here and use the screen? Hmm, that is such a tough decision.
Voice control is still too unreliable for me. About half the times I try to use it, I end up repeating myself with different variations, trying to get the exact right phrasing it wants. I usually end up having to do it manually anyway. I might as well just skip those step and go right to the finger.
When it works right the first time, it's like magic, but that is so rare.
Voice to text is handy in a bunch of situations now. Mostly when I want to send someone a message when I'm walking or typing. It's sort of neat for appointments, and I use it all the time for setting a timer. ..on my phone.
Screens and keyboards, or touch devices, are remarkably effective at conveying large amounts of information almost instantly. Voice is not an efficient medium. Do you know why people hate voicemail? Because it's slow and ineffective.
Voice apps will be the next 3DTV, which should have been clear to everyone wasn't going anywhere. If you have to put on goggles, it might as well be a VR headset..
In the meantime, I can order from Amazon from my tablet in a few seconds, deal with issues and confirmations, while I'm watching Netflix on my 65" regular 2D TV.
..don't panic
Isn't nice to know that everything you do with these things is tracked? Oh ya, and a FUCK APPLE just because.
When I am in the kitchen, I use Echo to set timers (very useful), play/change music and do news briefings and stuff like that. Using a touchscreen when you are working with meat or baking is a pain.
Reasonable enough. Other than the stock capabilities (weather, time, shopping list, timers, alarms, "what's playing at the movies?", "what's the phone number for Tire-Rama?", oodles of music sent to the theater system), the only third-party capabilities we use regularly are:
o Adjust the lighting via TP-Link smart plugs
o Adjust the heating / cooling via Sensi smart thermostat
o Check Fitbit stats / progress
Is it worth $49 or so out the door, plus hardware cost for associated devices to be able to do all this without having to otherwise go and do it? Well, it is to us.
For instance, sitting in the theater, it's either get up, make a 20 foot walk to the light switch, flip the switch, a 20 foot walk back in the dark, and sit down again, or just say "Echo, Turn off the lights." Likewise, when the show is over, it's just "Echo, Turn on the lights."
But when it'll cook a meal, see it delivered to the table, even see that the dishes are washed... yeah, that's going to be a fine day. At consumer prices, I'd hazard a guess that's still five or six years off.
I've fallen off your lawn, and I can't get up.
Speaking as an Echo owner, I've found that most of the apps I've found have been rubbish (translation: garbage). But that's not all that surprising: this is a completely new paradigm, on a par with WIMP. It's hardly surprising that
i) developers haven't yet really discovered what works well, nor have users discovered and popularised them; and
ii) the marketplace is flooded with junk, because it hasn't yet found a bedrock of solid, popular, useful stuff.
But it'll come. I keep my Echo in the kitchen, and setting a timer by voice is useful, if trivial. And I frequently use Spotify and TuneIn Radio to choose music.
I have little doubt that more will come in time. I think that voice control confirming the invention-hype-disappointment-delivery-equilibrium curve, and after a lot of users got acquired it as a Christmas gift during the hype phase, we're now in disappointment.
Similarly, people overestimate the impact of a new technology in the short run (which leads to the disappointment) but underestimate it in the long wrong. Voice will be no different.
me: Alexa, why can't you just understand my question?
alexa: Sorry, I don't understand this question. Please check out the list of questions I can understand on the App.
There, I said it.
It doesn't mean they're totally USELESS; no. For the majority of situations, they're more trouble than they're worth.
First, you have to be in exactly the right situation - there cannot be background noise or crosstalk - so essentially, a nearly SILENT room. How many of us spend a substantial amount of time in silence? I'm certainly not going to use a voice app on a bus, plane, or in public even if it was quiet, because anyone who does that is an obnoxious asshole.
Second, you have to know exactly the syntax the system is looking for. On my stupid car (BMX x5) it has voice activation but I'll be damned if I can ever remember what phrases it wants. "CALL HOME" (doesn't work, oh yeah, have to kick it to the phone menu) "PHONE" phone connected "CALL HOME" many results pick one.
Sigh. Oh, and my wife's name is Dawn, so fuck me if I don't have to sort through every damn "DON" in my phone book, distracting me away from the road while I do that - what am I *saving* using a voice app, again?
Third, you have to inevitably put up with a substantial failure rate. If I try to use a voice app for the simplest thing, dictating a slowly, clearly spoken text, I have to expect to spend the next few moments re-reading, editing, and correcting the text. If I'm trying to use it to come up with harder info - like names, in the example above - it's just a crapton easier to dial the number myself.
And I'm a Minnesotan (a region reputed to have a relatively clear style of speaking). I can't imagine how hard it must be for people with less intellgible accents.
-Styopa
I think their biggest problem is that they are completely useless devices, while at the same time being highly intrusive. I cannot think of a single compelling reason to have one. There is nothing that would cause me to remark, "Wow! I can't wait until I get my Alexa so I can do _________ ."
Proverbs 21:19
I have watched my wife and kids struggle with siri for years, and with alexa since christmas. I think a big tech fail with both is that the voice assistants do not get the context of the discussion. So, the user needs to be very precise with the instructions that are provided. For example, I hooked up alexa to wemo switches, connected it to our two christmas trees, named them white tree and green tree, and put them in a group called christmas trees.
To turn on/off, I just need to say 'Alex, turn on the christmas trees', or 'Alexa, turn off the white tree'. Then my kids end up trying:
- Alexa, turn on the trees? - fail
- Alexa, turn on the christmas tree lights? - fail
- Alexa, turn on the white christmas tree? - fail
- Alexa, is the white christmas tree on? - fail
Also, the voice assistants seem to struggle with accents. maybe the voice samples used for training are all captured in studios in southern California. My wife speaks with a slight south eastern accent, and I end up having to repeat her commands for the voice assistants like Sigourney Weaver in galaxy quest.
If you're going to buy something, whether new to you or a "refill", a device with a screen is better. Lots of information is immediately available for a quick read. This includes things like price, choice of vendors, warranty, user reviews. Even for a repeat purchase, something may have changed since the last purchase. Even for music apps, looking at your playlists for choices or perhaps looking for something new to listen to and seeing where you are in an album or play list, a screen is better. A weather app has far more information on screen than just one answer from one of these verbal only devices. The results of a search using a search app give many selections on screen to figure which might be the best result for what you want. I'm sure readers here can give many more examples where a device with a screen is far superior to one of these tube shaped so called AI based smart devices.
In a time of universal deceit, telling the truth is a revolutionary act. George Orwell
It's one thing when everybody you see is bent over their phone tapping away and ignoring everything else in the world. But it's a lot worse when they're talking at their phones.
While Amazon's Echo does seem to have gained some traction among the geek set, I'm not sure why Google Home was included here... unless the article was written by a Google fanboy.
I've seen a number of Echos in the wild, but not one Google device.
#DeleteChrome
If you want to know what the precursor to deflation in a sector, or the broader economy is, it's referred to by economists as an "overhang in productive capacity".
It's when there's more productiion than demand. And .. it's bad. Really bad.
It's an indication that an industry or a society has jumped the shark, and investment has crossed the line into mal-investment by exceeding consumer demand.
That's where we are now.
We have too many app developers. Too many coders. Too many UX designers. Too many entrepreneurs seeking first-mover advantage on new platforms.
Along comes the Echo and *bang*... there's a glut before there was ever a "thing".
Historically and mathematically speaking what comes next is a withering of said capacity. It's not going to be fun.
------ The best brain training is now totally free : )
I still see people on iphones texting and driving when they could just use "hey siri" to send a text instead. People build all this technology, but no one really teaches anyone how to use it or what it can really do and how it can help them, so it's nothing more than a novelty to the majority of users.
I find voice very useful. But the current "apps ecosystem" sucks, in addition to being a privacy nightmare. Most voice recognition should probably be on device, not in the cloud. That is, instead of some complex Rube Goldberg contraption springing into action when I say "OK Google, turn on bedroom lights", the bedroom lights should just understand voice directly.
I understand why that isn't working just yet (lightbulbs don't have enough computational power yet), but I expect it won't be long before that's how it works. That will also address privacy and security issues much better.
(Buys new lightbulb, screws it in, turns it on.)
Lightbulb: "Hello! I'm your new lightbulb. What would you like to call me?"
User: "You're one of my 'bedroom' lightbulbs."
Lightbulb: "OK, I'll respond to 'bedroom'. Would you like me to connect to the cloud?"
User: "No, thanks."
Lightbulb: "OK."
User: "Turn off bedroom lights."
(Lightbulb switches off.)
I personally have three options different voice ..Google on my phone, Alexa on my Fire TV, and Cortana on my computer and Xbox One. Three varying experiences. Cortana i barely use because I only use my Xbox for gaming which I seldom have time to do. Alexa on Fire TV i hardly use because once I've got Fire TV on most likely I'm going to netflix which doesn't support voice (lame as fuck). Google I used frequently on my phone till I migrated to a new phone in December and it completely forgot how to listen to me. I get so many errors now with Google that the only good use case is if I literally cannot type and that's rare. Not that there is an obvious solution, but there is definitely a long way before voice really is ubiquitous and easy rather than disjointed and complex. Lots of voice apps doesn't mean shit.
I used to use Google Assistant daily, and I liked it. Now I almost never do, because it is buggy, and it gets buggier with each release. It's highly accurate, *when it works at all*.
Voice recognition quality:
The voice recognition quality is stunningly accurate. It almost never gets things wrong. This is the hardest part, and they nailed it. But it seems like they had an intern write the rest of the code. Maybe it just wasn't exciting enough?
Speed:
My Galaxy S5, in 2015: "Okay google" *beep* "Send a text to..."
My Galaxy S5, in 2017: "Okay google" (45 second delay) *screen flash* "Send a text to..." (15 second delay)
This isn't just my specific phone, because my wife has the same model, but with almost no apps installed, and it performs the same way.
Bad parsing code: ... 2 paragraphs of text..." Google shows me the exact correct text I spoke, in a text box, then promptly says "Who do you want to send this text to?" Confused, I respond "Harold Smith" then it correctly finds the contact, then says "What would you like the text to be?" I say the text, it transcribes it perfectly, then says "Who do you want to send this text to?"
If you give Google assistant a command that is more that some arbitrary limit, like 256 characters or something, it gets stuck in a loop.
I say "Okay Google, send a text to Harold Smith, saying that
Bad contact lookup:
I say "Okay Google, send text to Dad" then it says "I cannot find a contact named Dad." Then I open my contact list, and there is a single entry named "Dad" with a cell phone number on it. Same spelling, same case.
No retry logic:
Sometimes it tells me something like "I'm sorry, I wasn't able to contact the server, please repeat that again." Why would I have to repeat it? Didn't it just record my voice? Other times, it actually transcribes the text, then tells me it couldn't contact the server. Ummm.... what? And it does that even if the action is local and doesn't require the server, like running an app or adding an appointment to my local calendar.
Must look down at the screen to use it:
On my iPhone, it would repeat back to me the message and prompt me to confirm. With Google Assistant, I have to look down at my phone and read it. I used my iPhone to send voice texts while on the road. I can't do that with Google Assistant since the whole point is to not have to take my eyes off the road.
Poor app integration:
After I send a text, it isn't in my text history.
i agree that most of the alexa skills are useless. they are junk. read them, they are like joke tellers or useless trivia.
but some are very useful. i think the key here is (as someone said above) we have 25+ years of GUI experience and about 2 of voice as UI experience. Give it some time.
A lot of these posts are mixing up alexa/google home and siri/phone. those are very different use cases. I don't like siri on my phone. i love alexa. we have three and use them all the time. but its about finding the right apps.
things we use them for all the time:
cooking timers
turning lights on or off
sonos integration!!! (right now this is a hacked version but native integration coming 2017).
alexa's native music selection is very slick
weather/current temp
simple calculations and conversions
adding things to amazon shopping list (she is wonderful at this, too wonderful, i think they dedicate way too many CPU cycles to this vs. other shit)
however its the right apps. lots of those skills are dumbassed.
IMHO. I've been building apps for 35 years on every platform possible (mainframe, unix boxes, mobile devices, pcs, etc). I think the voice UI is going to explode. We are going to look back in 5 years and be shocked at the use of voice as a UI. again IMHO. your mileage will vary
i think the key for the apps will be integration to other systems. it makes sense for nest to add alexa integration. for sonos. for lifx. etc.
its minimal work to enhance a product you are already heavily invested in.
i agree, most of the current apps are Hello World and silly. And i agree, there is no market in Alexa only apps.
...How many people even know that there are "apps" for these devices? It's not like they are really advertised or promoted. The Echo certainly doesn't tell you about them or ofFer suggestions. Other than a small banner on the Amazon site that can be easily missed, I would imagine few people even know that they can expand the capability of their devices.
It's a lot like the early web... There were the "official" web sites which looked good, loaded quickly, and worked. And then there were java applets, which were slow to load, buggy, and looked like high school projects.
It's pretty much the same. The features that come built-in are really good, voice recognition is fantastic, and it's overall a useful gadget to have to play music, set timers, reminders, check weather, and sports schedules.
Try to use any of the silly "skills" available, and you'll be very disappointed. The integration sucks ("tell to "), reliability sucks, and, therefore, usability sucks. Apart from having my kids ask for fart noises, there's not much out there yet.
Never underestimate the bandwidth of a 747 filled with CD-ROMs.
there are very few tasks where using a voice interface is actually faster than just punching it in with your fingers. {...}
- Setting alarms
- Setting reminders
And that's because the "time input" interface of iOS is completely stupid.
(compare with the "clock-face"-like interface. as seen on Jolla and other Mer derivatives).
- Playing a specific song or a specific album
Again, that mostly because iOS lacks a "Just type..." search interfae like Palm/HP webOS.
(in webOS, starting to type on the physical keyboard will nearly always cause a reaction.
- When in an app, it usually filters the on-screen list .
- When outside, it causes some kind of desktop search. Any object that could correspond to the typed swquence is automatically consitered.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
Try speaking in Tamil or Tamil accent to Google Assistant. Sundar Pichai is a Tamilian and they might make sure that accent works correctly when they demo it to their boss.
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
Hey, newsflash, most people text/use a chat program rather than calling each other because reading something is actually faster than listening to something. If we choose text for interacting with other human beings over voice when given equal opportunity for both then why would we do any different for interacting with a computer?
No, you most certainly didn't. You found the guy who doesn't spend even a tiny fraction of what others do on children, booze, drugs, bars, travel, going out to eat, long trips, interest, hotels, sports events, video games, software, "apps", new cars, parties, education, or junkfood — and hasn't for quite a few decades now.
Which left me way more than enough to build a very nice theater into my home, the entire interior of which I built and wired by hand, after buying the property. Even with a modest income. Also, I bought the property with the specific intent of putting a theater into it - it was an abandoned church, a classic tabula rasa. Just a huge, empty room. And I had mucho help - my SO is awesome, and very much like-minded.
We each have our priorities. Home entertainment and at-home convenience are some of mine, that's all. In fact, almost every optional expenditure I make is in pursuit of a concrete, lasting improvement to my physical circumstance. If you don't have enough left over to do what you dream of by the time you're my age (I started this particular undertaking when I was 50, I'm 60 now), then you're Doing It Wrong.
Up till now, anyway. I don't know what's going to happen to the younger people going forward. Looking a good deal more bleak than it did for me.
I've fallen off your lawn, and I can't get up.
Older gear (probably not going to get updated, either. Because we have a good viewing and listening experience already.) Discrete components; pre-pro, amps, speakers, etc. The pre-pro could be remoted, perhaps, but it's very early on the curve of network control, and I've found it's not even reliable to tell to turn on and off. Denon bought Marantz, and they have been pretty sad about proper updates to nominally update-capable components.
OTOH, if a proper STT interface ever hits the streets (and no, I don't count the Echo - the number of negative developer and privacy issues there are ridiculous) I might be motivated to undertake such a setup. Mainly change the pre-pro to one that's smart enough to reliably remote and dedicate a computer with lots of storage to the theater as an AV source. But I'm 60, and every year that passes, I'm more satisfied with what I already have, so... perhaps not.
Already pretty much ignoring the 4K thing. Aside from very low media availability at this point in time, 1080p looks great on a big screen (and your average movie director still thinks it's "artsy" to soft focus and/or use a lens with horrific DOF, either/both of which completely waste all that fine resolution goodness anyway.)
I've fallen off your lawn, and I can't get up.
Most are uninteresting, useless, or easier to do some other way. Its great to be able to come home and talk to it and have my lights go on, or tell it what song to play from Spotify, or ask it about the weather - when it actually understands. Having tried to develop an Alexa skill for my own use, its clear how the only way to get it to react to natural sentences is for the skill developer to come up with pretty much every combination and ordering of words for which the user might convey what it is that they want. And even then, with context provided, it has almost no ability to deal with words where homynyms exist. Tell it that the user will speak a number in between two other words, and when the user says "two" or "four" in between those two words, it doesn't care that a number is supposed to go there and instead insists the user meant "to", "too", "for" or "fore". My favorite is when I tried to have it recognize the word "main" it insisted I was saying "Maine". Bottom line, its no better than other voice recognition technologies which can usually form words pretty well from the sounds that are spoken, but really struggles to chose the one that makes sense when alternatives exist.
Understanding is a three edged sword. - Ambassador Kosh Naranek, Babylon 5
The first dot mainly plays music in my living room but the second I use as an alarm clock and music player in my bed room. It's really nice not to have that bright LED screen glaring at you all night and the first time you check the time in the middle of the night with out opening your eyes you'll never want to go back.
Two changes I would like to see that would make the echo's even better are named timers (chicken 20 minutes, potatoes 10 minutes, etc..) and the ability to control other echos (like Sonos speakers).
The perversity of the Universe tends towards a maximum. - O'Toole's Corollary
I think 3% after 2 weeks is BETTER than 13% retention rate after 1 week.
I believe that if you go from 100% to 13% in 1st week, and the losses continue at that rate, then retention will go from 13% to 1.7% in the 2nd week.
I just used 0.13^2 to calculate it and I don't know if that math is right. But if it is right, then the article is stating the opposite of the truth.