I bet the full Siri interpreter is not running all the time. The "Hey Siri" functionality is probably based on simpler sound wave matching being done by the M9 chip. Otherwise, more CPU & battery power would be needed and the iPhone 6S would need to be the size of an Android brick to hold the battery needed.
In other words, the iPhone 6s is not "constantly interpreting your speech" unless triggered first by the "Hey Siri" sound, which is not voice invariant. Hence the initial "training"
That's a disappointment, I had thought the necessity of training voice recognition systems was a thing of the past.
This is the future if you want your digital personal assistance to respond only to your voice. I'd like Apple to take this even further with better training for people with different accents or even speak impediments so the system knows your speech pattern for every phoneme.
I think you've always wanted it to respond only to your voice - the difference in iOS9 is that Siri will always be on, isn't it?
Siri in iOS 8 does not require training, you just pick it up and start using it. The last time I had to train a voice recognition system would have been at least 5 years ago, a Dragon product I believe. I thought it was unnecessary now.
It isn't to train it to understand the words - it's so Siri will not respond to any random "Hey, Siri" it hears. Cortana on Windows 10 does the same thing - if you check the option to only respond to your voice....
I'd like the ability to tell Siri how to dictate when making messages. For instance, my son is named Jax, yet Siri always writes Jacks. Or, certain towns, like Temecula when trying to get directions, Siri doesn't understand what I say but when Siri says it she pronounces it Tem e cula. They sound the same/similar but it would be great if I could tell Siri to change how it's written.
:-) :-) :-)
You should realized when you were naming your son, that a non standard spelling would cause issues in the future with voice recognition technologies, and just used standard spellings....
Siri in iOS 8 does not require training, you just pick it up and start using it. The last time I had to train a voice recognition system would have been at least 5 years ago, a Dragon product I believe. I thought it was unnecessary now.
Come on, bummed by a one minute training process? Any why do you think this is necessary? Almost everyone I'm around has an iPhone. If it's always listening, don't you think its sensible that it have a simple baseline of your voice vs four other people, any of which my say the same thing?
Heck, my Watch "heard" Schiller's "Hey Siri, What's it like being you?" question from the keynote and I got the same answer as he did...
It happens often enough when watching TV on the iPad and some parts of the dialog triggers Siri and interrupted the program. Hope this will resolve that.
Do you think it's a security ("my voice is my password") type thing? I was assuming it was just a learning thing.
If it's just a learning thing then it needn't be the future, because people are capable of meeting new people and understanding them right away with no training, so there must be some way to do it, which we will presumably learn how to program eventually.
I do believe it is more of a security measure, to avoid SIRI from being
Activated by random people by accident or intentionally.
So that vagabonds don't go around people telling their IPhone
Stupid commands just for their fun. (Like some have said here they would go
Around doing)
Funny; I told my wife this morning that Siri needed this! I had my iPhone plugged into my TV, streaming the live Apple event on Wednesday. Phil or one of the other presenters demoed Siri by saying "Hey, Siri!" and my phone went into Siri mode, interrupting the stream. If it's now an always-on/unplugged feature on some devices, you need a way to limit recognition to your voice. We already do this to annoy Android friends by randomly saying "Okay, Google" and issuing commands to their phones.
I wouldn't expect the training to totally eliminate that problem. You have to train Android before it will allow OK Google to activate the search interface, and as you've discovered, search can still be triggered by other people if you're able to imitate the phone owner's speech at a basic level (pitch, cadence, etc.).
Funny to see this story. Earlier today I posted a made-up joke (also in AppleInsider):
What's the fastest way to find out who's got an iPhone 6s in a room? Yell "Hey Siri".
I followed with the more serious question of how, in a world full of iPhones, do we avoid situations like the above? I guess this article provides the answer! I think this new training has less to do with making Siri more able to understand you, and more with identifying you!
How is going to deal with multiple users ? me and my partner both use each others siri, when for example when one of us is driving or asleep or whatever ??
This is the future if you want your digital personal assistance to respond only to your voice. I'd like Apple to take this even further with better training for people with different accents or even speak impediments so the system knows your speech pattern for every phoneme.
Comments
In other words, the iPhone 6s is not "constantly interpreting your speech" unless triggered first by the "Hey Siri" sound, which is not voice invariant. Hence the initial "training"
Quote:
That's a disappointment, I had thought the necessity of training voice recognition systems was a thing of the past.
This is the future if you want your digital personal assistance to respond only to your voice. I'd like Apple to take this even further with better training for people with different accents or even speak impediments so the system knows your speech pattern for every phoneme.
I think you've always wanted it to respond only to your voice - the difference in iOS9 is that Siri will always be on, isn't it?
Only if you want it to be and only on the iPhone 6S-series, and presumably the iPad Pro.
Siri in iOS 8 does not require training, you just pick it up and start using it. The last time I had to train a voice recognition system would have been at least 5 years ago, a Dragon product I believe. I thought it was unnecessary now.
It isn't to train it to understand the words - it's so Siri will not respond to any random "Hey, Siri" it hears. Cortana on Windows 10 does the same thing - if you check the option to only respond to your voice....
I'd like the ability to tell Siri how to dictate when making messages. For instance, my son is named Jax, yet Siri always writes Jacks. Or, certain towns, like Temecula when trying to get directions, Siri doesn't understand what I say but when Siri says it she pronounces it Tem e cula. They sound the same/similar but it would be great if I could tell Siri to change how it's written.
:-) :-) :-)
You should realized when you were naming your son, that a non standard spelling would cause issues in the future with voice recognition technologies, and just used standard spellings....
In other words, "You're holding it wrong!"
:-) :-) :-)
That's a disappointment, I had thought the necessity of training voice recognition systems was a thing of the past.
It's not for comprehension: it's for security. Using your voice print like a fingerprint so it only launches when YOU want.
Siri in iOS 8 does not require training, you just pick it up and start using it. The last time I had to train a voice recognition system would have been at least 5 years ago, a Dragon product I believe. I thought it was unnecessary now.
Come on, bummed by a one minute training process? Any why do you think this is necessary? Almost everyone I'm around has an iPhone. If it's always listening, don't you think its sensible that it have a simple baseline of your voice vs four other people, any of which my say the same thing?
Heck, my Watch "heard" Schiller's "Hey Siri, What's it like being you?" question from the keynote and I got the same answer as he did...
the difference in iOS9 is that Siri will always be on, isn't it?
Only if you want it to be and only on the iPhone 6S-series, and presumably the iPad Pro.
Huh - I thought it was software, not hardware specific...
This always on setup is now built into the A9 chip, along with the M9. On other devices 'Hey, Siri' is available only when plugged in.
I do believe it is more of a security measure, to avoid SIRI from being
Activated by random people by accident or intentionally.
So that vagabonds don't go around people telling their IPhone
Stupid commands just for their fun. (Like some have said here they would go
Around doing)
I wouldn't expect the training to totally eliminate that problem. You have to train Android before it will allow OK Google to activate the search interface, and as you've discovered, search can still be triggered by other people if you're able to imitate the phone owner's speech at a basic level (pitch, cadence, etc.).
Huh - I thought it was software, not hardware specific...
This always on setup is now built into the A9 chip, along with the M9. On other devices 'Hey, Siri' is available only when plugged in.
Yep, I see that now - thanks for straightening me out - guess I wasn't paying attention
What's the fastest way to find out who's got an iPhone 6s in a room? Yell "Hey Siri".
I followed with the more serious question of how, in a world full of iPhones, do we avoid situations like the above? I guess this article provides the answer! I think this new training has less to do with making Siri more able to understand you, and more with identifying you!
There was a lot to remember that day. The 1080p version is available as a Podcast now.
This is the future if you want your digital personal assistance to respond only to your voice. I'd like Apple to take this even further with better training for people with different accents or even speak impediments so the system knows your speech pattern for every phoneme.