I believe it. Amazon Echo is amazing with what it accurately understands, and I really don't think Amazon's speech-to-text technology is better than Apple's.
I rarely have any trouble with Siri parsing the words. I can tell because it gets printed to the screen very accurately. The trouble usually happens after that text is analyzed by the server where the actual intent is often completely misunderstood. I've never tried Echo but I find Google is much better than Siri at returning the desired result. I always give Siri the first crack because it is just a home button hold, but more often than not I end up launching Google, because much of the time Siri just can't deliver. Siri is excellent at some things though - setting calendars, reminders, alarms, call, texting and surprisingly getting accurate up to the minute sports scores.
I find Siri to be hit and miss these days. It has gotten to the point where I don't really bother any more. It used to be so much more reliable. These days it is as if Siri is drunk half the time. Here is copy of a message I dictated whilst driving, no interference, I held the phone as I normally do. The message was to my daughter about a type of dog she really likes, that I came across in the local dog park.
It was a very bossy. They are herding dog by nature today 101 Van Ness Circle very much. Like to change the ball and especially like just spoke with AutoZone dog I get to school with her for a while.
The thing is that sometimes Siri works perfectly, even with noise or dialogue in the background. I have friends who use Siri all the time and have no idea what I am complaining about. So either I am the one who's drunk, or perhaps the microphone has got some fluff in it. I don't know but it doesn't seem to be a universal problem.
The microphone has a tiny cross section and therefore a tiny surface area, resulting in a low sensitivity. Similarly, it's very difficult to build tiny speakers that are loud and have a large dynamic range. Low frequency sounds have a larger wavelength and cannot be accurately reproduced. Miniaturization has a trade off that is hard to get around.
Although there are multiple mics, I'm pretty sure Siri only uses one of them.
I'm pretty sure they use multiple mics to enable phase detection (the difference in time a sound reaches one microphone vs the other) to determine foreground vs background. Siri in the car uses only one mic which is why it's much inferior.
That might be a lot harder than you think. For one, the phone probably would not have enough storage space required for all the language files. Two, you would not get any improvements or fixes until the next OS update. With a server they can be updating it continuously. Furthermore, I think you probably underestimate the massive processing power required, not to mention the battery drain.
Android has good offline voice recognition. With no internet connection I can dictate an sms and it works fine and most other basic commands work. From what I can gather, it's only available in US English though, which is a bit annoying so I have turned the setting off as I don't really need it. Regardless, it shows the voice recognition can be done in device. Also, the updates arrive independently of the OS.
Phone mics are pretty capable these days. They have a noise floor under 30 dB SPL and can go over 120 dB SPL. Frequency response goes well below 100 Hz. The issue with pick-up room-wide is more about reverberation. This requires a lot of processing to overcome. In a car, there is a lot of low-frequency noise, which we are not sensitive to, but the mic has to deal with and not overload. Essentially the SNR can become very poor. Multiple mics with beamforming would help. But thank the Cloud for the huge processing it can add to overcome some of these limitations and use noise identification and speech context to give us more-accurate results.
I believe it. Amazon Echo is amazing with what it accurately understands, and I really don't think Amazon's speech-to-text technology is better than Apple's.
I rarely have any trouble with Siri parsing the words. I can tell because it gets printed to the screen very accurately. The trouble usually happens after that text is sent to the server where the actual intent is often completely misunderstood. I've never tried Echo but I find Google is much better than Siri at returning the desired result. I always give Siri the first crack because it is just a home button hold, but more often than not I end up launching Google, because much of the time Siri just can't deliver. Siri is excellent at some things though - setting calendars, reminders, alarms, call, texting and surprisingly getting accurate up to the minute sports scores.
Sound is send to the server and is parsed there, not on your device. Siri is completely disabled when not connected.
This is exactly what Apple should've fixed by now... on-device real-time processing of voice, even if it's the initial stages. A dedicated low-power chip could do that, much like their M-series chips for tracking motion. I call it the V-series chip, and let's hope that Apple is wise enough to implement something like this!
oh god. it's not a matter of being "wise" enough to think of something like that...ideas are the easy, free part. any idea you can think of they've already thought of. it's about engineering and the confines, limitations, and challenges of reality. you have no idea if a "dedicated low-power chip" could do what you want it to.
Better not listen to this analyst. Microphones do not filter frequencies or distinguish between foreground and background noises.
As a couple of others have pointed out, yes they do if there are more than one and good programming. I can reliably dictate a text in a busy, noisy restaurant or bar on a now three year old smartphone. It can also accurately parse voice commands and dictation off-line, no remote server required. I've even texted by voice while driving with the radio on, no Bluetooth connected, and my phone in my shirt pocket. Yes that one surprises me too as I discovered it accidentally, thinking my phone was on the open console "listening" for the activation phrase.
I think my three year phone might have been the first with a dedicated low-power "always-on" voice recognition chip, not certain. I'm pretty sure it's fairly common now tho.
and let's hope that Apple is wise enough to implement something like this!
I'm constantly amazed at the number of people who believe they've thought of something Apple hasn't, based solely on the fact that their pet idea hasn't been implemented.
Apple's secrecy is, rightly or wrongly, partly responsible. Many people who've had a peek behind the curtain have said that Apple has tried, is trying, and has not implemented for reasons mostly only Apple knows. Their priorities are often different from the general public which further affects what we see. But the idea that Apple hasn't already considered something bandied about in a forum is pretty silly. The recent Wired article on the Apple Watch show just how much detail is sweated by Apple.
As to Siri's accuracy, something in the algorithms/server farm/neural network does give Siri pause on occasion. I've found some phrases just aren't parsed accurately even when said by other people. (I can't think of one, at the moment). No matter how you say one thing, Siri thinks your saying something else. It happens just often enough that I can't always count on showing off Sirii to someone unfamiliar with her. (Yes, I said 'her'.)
I realize that it takes a lot to get what we get, so far. I'm just saying while she's accurate most of the time, sometimes she's not. But there's been constant improvement in the responses, so Apple isn't just kicking back.
I believe it. Amazon Echo is amazing with what it accurately understands, and I really don't think Amazon's speech-to-text technology is better than Apple's.
I rarely have any trouble with Siri parsing the words. I can tell because it gets printed to the screen very accurately. The trouble usually happens after that text is analyzed by the server where the actual intent is often completely misunderstood. I've never tried Echo but I find Google is much better than Siri at returning the desired result. I always give Siri the first crack because it is just a home button hold, but more often than not I end up launching Google, because much of the time Siri just can't deliver. Siri is excellent at some things though - setting calendars, reminders, alarms, call, texting and surprisingly getting accurate up to the minute sports scores.
And that's exactly how Apple is building the technology, its amazing how people still complain, like everything Apple, it starts by being excellent at a few things then build on from there, there is not one apple product that has come out and instantly did everything really well day one, that's not the Apple way, Timmy might have to publicize this so people understand the companies and alleviate frustration, people's expectation and Apple's internal culture conflicting all the time is placing unessesary strain on the company's brand and perpetuats ideas like the company can't innovate any more because people don't get How Apple Especially innovates, case of lost in translation I guess
Better not listen to this analyst. Microphones do not filter frequencies or distinguish between foreground and background noises.
As a couple of others have pointed out, yes they do if there are more than one and good programming. I can reliably dictate a text in a busy, noisy restaurant or bar on a now three year old smartphone. It can also accurately parse voice commands and dictation off-line, no remote server required. I've even texted by voice while driving with the radio on, no Bluetooth connected, and my phone in my shirt pocket. Yes that one surprises me too as I discovered it accidentally, thinking my phone was on the open console "listening" for the activation phrase.
I think my three year phone might have been the first with a dedicated low-power "always-on" voice recognition chip, not certain. I'm pretty sure it's fairly common now tho.
It he point is, as I pointed out clearly, it's not the microphone, it's the processing of the information not the recording of it.
This is exactly what Apple should've fixed by now... on-device real-time processing of voice, even if it's the initial stages. A dedicated low-power chip could do that, much like their M-series chips for tracking motion. I call it the V-series chip, and let's hope that Apple is wise enough to implement something like this!
oh god. it's not a matter of being "wise" enough to think of something like that...ideas are the easy, free part. any idea you can think of they've already thought of. it's about engineering and the confines, limitations, and challenges of reality. you have no idea if a "dedicated low-power chip" could do what you want it to.
I have a pretty good idea of what hardware in what volume can process sound in an adequate way. Its not the processing power nor the microphone holding back acceptable sound processing on the iPhone, it's the method used. Go is a perfect example, yes you can beat go champions with computer clusters, but it's by brute force, not intelligent and it will take a 'megawatt' of power instead of that of a light bulb (it's actually 3 orders of magnitude difference in energy use).
I believe it. Amazon Echo is amazing with what it accurately understands, and I really don't think Amazon's speech-to-text technology is better than Apple's.
I rarely have any trouble with Siri parsing the words. I can tell because it gets printed to the screen very accurately. The trouble usually happens after that text is analyzed by the server where the actual intent is often completely misunderstood. I've never tried Echo but I find Google is much better than Siri at returning the desired result. I always give Siri the first crack because it is just a home button hold, but more often than not I end up launching Google, because much of the time Siri just can't deliver. Siri is excellent at some things though - setting calendars, reminders, alarms, call, texting and surprisingly getting accurate up to the minute sports scores.
And that's exactly how Apple is building the technology, its amazing how people still complain, like everything Apple, it starts by being excellent at a few things then build on from there, there is not one apple product that has come out and instantly did everything really well day one, that's not the Apple way, Timmy might have to publicize this so people understand the companies and alleviate frustration, people's expectation and Apple's internal culture conflicting all the time is placing unessesary strain on the company's brand and perpetuats ideas like the company can't innovate any more because people don't get How Apple Especially innovates, case of lost in translation I guess
As a customer I don't care about such statements, Apple introduced Siri as usefull and it clearly isn't for me and most other people I read of. Its clear to me that Apple itself didn't understood the technology they bought and didn't have a correct idea about the possible growth of the product nor the difficulties expanding it to other languges (for example). Apple clearly had a lack in understanding AI and remains to do so until now.
and let's hope that Apple is wise enough to implement something like this!
I'm constantly amazed at the number of people who believe they've thought of something Apple hasn't, based solely on the fact that their pet idea hasn't been implemented.
she's accurate most of the time, sometimes she's not. But there's been constant improvement in the responses, so Apple isn't just kicking back.
I'm constantly amazed at the number of people who believe Apple is an all knowing entity. Its a fact that 'Apple' didn't think of most ideas (first).
I'm constantly amazed at the number of people who believe they've thought of something Apple hasn't, based solely on the fact that their pet idea hasn't been implemented.
she's accurate most of the time, sometimes she's not. But there's been constant improvement in the responses, so Apple isn't just kicking back.
I'm constantly amazed at the number of people who believe Apple is an all knowing entity. Its a fact that 'Apple' didn't think of most ideas (first).
It's also fact that 'no single entity' has thought of most ideas (first), so why make the statement. Did you have a point?
I believe it. Amazon Echo is amazing with what it accurately understands, and I really don't think Amazon's speech-to-text technology is better than Apple's.
I rarely have any trouble with Siri parsing the words. I can tell because it gets printed to the screen very accurately. The trouble usually happens after that text is sent to the server where the actual intent is often completely misunderstood. I've never tried Echo but I find Google is much better than Siri at returning the desired result. I always give Siri the first crack because it is just a home button hold, but more often than not I end up launching Google, because much of the time Siri just can't deliver. Siri is excellent at some things though - setting calendars, reminders, alarms, call, texting and surprisingly getting accurate up to the minute sports scores.
Sound is send to the server and is parsed there, not on your device. Siri is completely disabled when not connected.
This is exactly what Apple should've fixed by now... on-device real-time processing of voice, even if it's the initial stages. A dedicated low-power chip could do that, much like their M-series chips for tracking motion. I call it the V-series chip, and let's hope that Apple is wise enough to implement something like this!
That's my biggest pet peeve with Siri. When I'm in the car driving and I want to send a text message, Siri should be able to understand the request, allow me to dictate the message and then attempt to send it... all on-device. In my experience, 9 out of 10 times when I try to send a text from the car, Siri after a very long pause comes back and says something like: "I can't help you right now... Please try your request later". When I finally get to the point where Siri asks what I'd like the message to say, I'd dictate the message and then get dead air again ...eventually Siri comes back asks again: "What would you like the message to say?".
I have a 25 minute commute and it could take me 20 minutes just to send a text message! This all happens while I have 3 bars of signal and I can make phone calls. If I pull over and stop along the route, Siri works fine. I suspect it has something to do with connection being handed off from tower to tower that Siri doesn't like. BTW, I don't live in the sticks; I live in a suburban area with cell towers all around.
Siri works wonderfully on my Apple watch. Brilliantly on the iPhone 6 Plus. And pretty lousily on any bluetooth car system, e.g. Sony XAV-701BT. The Alpine iLX-007E combines with an axxess1 controller for steering wheel integration allows my 2006 era Mitsubishi to have great Siri integration and CarPlay. I have vol up, vol down/mute, siri, on the three buttons on the right of the steering wheel. The other side has up/forward/skip, play/pause, and down/backwards/rewind on the three left hand buttons. The microphone with the iLX-007 works very well with Siri in this Mitsubishi 380 sedan, as the interior is very quiet, little road noise. But the same setup in my 2004 Pajero (Montero in the US) doesn't work as well as I have all terrain tyres which produce road noise at speed and the microphone can't work as well. The iPhone works fine in the same car though. The iPhone has great Mic's. If I'm driving at high speed in the Pajero and need to use Siri I use Hey Siri via the watch, which always works well. The results still get read through the CarPlay system.
Never having a daily driver again without CarPlay. It's terrible not having it now.
Sri doesn't work for me in any environment that is even slightly noisy unless I am using a headset then it works great. This is especially true in a vehicle or public space so I do not even try anymore unless I am using a headset.
Comments
It was a very bossy. They are herding dog by nature today 101 Van Ness Circle very much. Like to change the ball and especially like just spoke with AutoZone dog I get to school with her for a while.
The thing is that sometimes Siri works perfectly, even with noise or dialogue in the background. I have friends who use Siri all the time and have no idea what I am complaining about. So either I am the one who's drunk, or perhaps the microphone has got some fluff in it. I don't know but it doesn't seem to be a universal problem.
The issue with pick-up room-wide is more about reverberation. This requires a lot of processing to overcome.
In a car, there is a lot of low-frequency noise, which we are not sensitive to, but the mic has to deal with and not overload. Essentially the SNR can become very poor. Multiple mics with beamforming would help.
But thank the Cloud for the huge processing it can add to overcome some of these limitations and use noise identification and speech context to give us more-accurate results.
I think my three year phone might have been the first with a dedicated low-power "always-on" voice recognition chip, not certain. I'm pretty sure it's fairly common now tho.
Apple's secrecy is, rightly or wrongly, partly responsible. Many people who've had a peek behind the curtain have said that Apple has tried, is trying, and has not implemented for reasons mostly only Apple knows. Their priorities are often different from the general public which further affects what we see. But the idea that Apple hasn't already considered something bandied about in a forum is pretty silly. The recent Wired article on the Apple Watch show just how much detail is sweated by Apple.
As to Siri's accuracy, something in the algorithms/server farm/neural network does give Siri pause on occasion. I've found some phrases just aren't parsed accurately even when said by other people. (I can't think of one, at the moment). No matter how you say one thing, Siri thinks your saying something else. It happens just often enough that I can't always count on showing off Sirii to someone unfamiliar with her. (Yes, I said 'her'.)
I realize that it takes a lot to get what we get, so far. I'm just saying while she's accurate most of the time, sometimes she's not. But there's been constant improvement in the responses, so Apple isn't just kicking back.
I have a pretty good idea of what hardware in what volume can process sound in an adequate way.
Its not the processing power nor the microphone holding back acceptable sound processing on the iPhone, it's the method used. Go is a perfect example, yes you can beat go champions with computer clusters, but it's by brute force, not intelligent and it will take a 'megawatt' of power instead of that of a light bulb (it's actually 3 orders of magnitude difference in energy use).
As a customer I don't care about such statements, Apple introduced Siri as usefull and it clearly isn't for me and most other people I read of.
Its clear to me that Apple itself didn't understood the technology they bought and didn't have a correct idea about the possible growth of the product nor the difficulties expanding it to other languges (for example).
Apple clearly had a lack in understanding AI and remains to do so until now.
Its a fact that 'Apple' didn't think of most ideas (first).