3) If you have Siri on even the simple commands, like calling a contract or starting a Playlist are sent offsite for processing, but if you turn Siri off the old Voice Control system apparently kicks in. Can anyone verify this at the risk of Siri losing their speech pattern DB?
When you switch Siri off you get a dialog:
"The information Siri uses to respond to your requests will be removed from Apple servers. If you want to use Siri later, it will take time to resend this information."
You can cancel, but I went ahead.
Later, I turned Siri on (I am very macho and attractive ;-)> and all was as it had been before.
"The information Siri uses to respond to your requests will be removed from Apple servers. If you want to use Siri later, it will take time to resend this information."
You can cancel, but I went ahead.
Later, I turned Siri on (I am very macho and attractive ;-)> and all was as it had been before.
...I think it's the wink and the goatee.
When you had Siri off did you test the Voice Control options?
I assumed that 63KB was only for the voice packet being sent to Apple, but it does say "per query" so its much lower than I expected.
I wonder how the other services size up to Siri on the 4S, including Dragon Dictation, the original Siri app on iOS, and doing a variety of text-based searches in Google, Wolfram-Alpha, Yelp!, etc.
I just dictated the following into Nuance iP4S email content:
"Where have all the flowers gone long time passing where have all the flowers gone long long ago"
and according to DataMan:
After Dictating, before sending:
Cell: 0.0 KB WiFi: 474.8 KB
Added To (my email) and sig (Sent from my iPhone) and sent it:
Cell: 0.5 KB WiFi: 299.5 KB
Interesting, that 0.5 KB cell on the sending of the email...
DataMan shows a map with flags (I think they are cell towers). The email involved a flag about a mile from the house.
I have Location Services and Notifications (including DataMan) enabled...
I just dictated the following into Nuance iP4S email content:
"Where have all the flowers gone long time passing where have all the flowers gone long long ago"
and according to DataMan:
After Dictating, before sending:
Cell: 0.0 KB WiFi: 474.8 KB
Added To (my email) and sig (Sent from my iPhone) and sent it:
Cell: 0.5 KB WiFi: 299.5 KB
Interesting, that 0.5 KB cell on the sending of the email...
DataMan shows a map with flags (I think they are cell towers). The email involved a flag about a mile from the house.
I have Location Services and Notifications (including DataMan) enabled...
That's interesting. Looks like it is processing the data locally with lossless compression before sending.
I wonder if that 0.5KB for the cell is just your phone pinging the towers per standard cellphone upkeep or if that is Siri using the towers for some unknown reason.
Yes... you can't go back... ...it impersonated Siri's voice, though...
I noticed in one of the iOS 5.0 Betas on my iPhone 4 that the voice had changed to what we now know as Siri's voice. I assume that all iOS 5 devices have that voice.
I don't know how they can get it as low as 63kb. I would think it would be a lot more for a request and a response. Just loading a typical Google's search page as minimal as it is uses 57kb just for the response. That doesn't take into consideration the original request which if it is compressed audio has got to be more than 63kb total.
Last I looked, there were over 3 pages on comments to the linked ARS article.
I think it was 63 KB for Siri requests that didn't fail through to a web search.
The audio recording size and compression was discussed, but it is beyond my area of expertise.
Some seemed to think that the typical request could be filtered/compressed on the iPhone before sending it to Apple's servers.
That appeared reasonable.
Others, seemed to think that after the server did the speech to text, the text was sent back to the device for parsing and decoding.
That made no sense at all -- because many decoded requests would, then, be sent back to the server to complete the requests (another handshake and data transmission).
Others, seemed to think that after the server did the speech to text, the text was sent back to the device for parsing and decoding.
That makes no sense. That's the part of Siri's brain that needs to be parsed on servers. And we can see that Siri doesn't work and even tells you it's removing your information for parsing your requests will be removed from Siri's servers when you go to turn it off.
This could have an additional effect. If Siri is doing the Google search from the Apple server, Google loses the ability to see the data of who initiated the request. No data, no ads for them. THAT'S INSANELY GREAT!!!
Not to mention that Apple could and probably is designing their own search engine. This could be part of Steve-o's win at all costs strategy.
Ask Siri to tell you a joke. BTW we are again at the start of a new revolution. I've always thought talking to a machine was not that great, even on the Mac or many years ago on PC.
My most fun AI interaction was always Dr. Sbaitso, if anyone remembers that
But tell you what, Apple has cracked it. Talk naturally to something friendly that assists you in natural day-to-day tasks.
If they keep on improving, they've got a killer new user interface. Open the API to apps and Boom! The next app gold rush will be underway. There'll even be counselling and religious apps based on Siri API interfaces.
There's a lot of work to be done though, the voice recognition needs to be a lot more adaptive, I am a native English speaker but fall somewhere right in between British, American and Australian English. The British recognition seems to work best, though sometimes it is way off base since it is command-based more so than dictation-based unless you specifically tell it you are dictating something eg. text messages.
But we are just maybe five years away, if Apple keeps going in the right direction, to how the crew talks to the Icarus AI and how it responds in the movie "Sunshine". Interesting trivia for those who enjoy that movie: Danny Boyle mentions in the commentary that the "computer voice" was recorded by an actress not as canned responses separately in a sound studio but during the actual filming of the live actors in the scene. So there is an uncanny feeling to the dialogue whereby you know the AI is, well, synthetic but the AI is also engaging and responsive in a way that Siri can be by the end of the decade.
In any case, as with most things tech, sometimes it is all about timing. Making voice recognition and AI cloud-based was always the way to go because of the level of raw data you can collect and work with. Despite CPUs being powerful enough to process tons of data, the cloud still can collect and store that data at a level beyond local storage ever could, at least for this kind of AI application. This is big.
Not to mention that Apple could and probably is designing their own search engine. This could be part of Steve-o's win at all costs strategy.
More like another example of his common-sense strategy. For many of us, Steve doesn't seem so crazy and radical, just really, really sensible sometimes... Even if he didn't think so himself most of the time.
Comments
3) If you have Siri on even the simple commands, like calling a contract or starting a Playlist are sent offsite for processing, but if you turn Siri off the old Voice Control system apparently kicks in. Can anyone verify this at the risk of Siri losing their speech pattern DB?
When you switch Siri off you get a dialog:
"The information Siri uses to respond to your requests will be removed from Apple servers. If you want to use Siri later, it will take time to resend this information."
You can cancel, but I went ahead.
Later, I turned Siri on (I am very macho and attractive ;-)> and all was as it had been before.
...I think it's the wink and the goatee.
So... If you use 1 GB of data per month on mobile web searches, doing the same with Siri could only use 90 MB of data...
... faster and no dross...
Hmm...
So you can do 1,000 Siri requests per day on your cell service without going over 2 GB, assuming you do nothing else that uses data.
When you switch Siri off you get a dialog:
"The information Siri uses to respond to your requests will be removed from Apple servers. If you want to use Siri later, it will take time to resend this information."
You can cancel, but I went ahead.
Later, I turned Siri on (I am very macho and attractive ;-)> and all was as it had been before.
...I think it's the wink and the goatee.
When you had Siri off did you test the Voice Control options?
I assumed that 63KB was only for the voice packet being sent to Apple, but it does say "per query" so its much lower than I expected.
I wonder how the other services size up to Siri on the 4S, including Dragon Dictation, the original Siri app on iOS, and doing a variety of text-based searches in Google, Wolfram-Alpha, Yelp!, etc.
I just dictated the following into Nuance iP4S email content:
"Where have all the flowers gone long time passing where have all the flowers gone long long ago"
and according to DataMan:
After Dictating, before sending:
Cell: 0.0 KB WiFi: 474.8 KB
Added To (my email) and sig (Sent from my iPhone) and sent it:
Cell: 0.5 KB WiFi: 299.5 KB
Interesting, that 0.5 KB cell on the sending of the email...
DataMan shows a map with flags (I think they are cell towers). The email involved a flag about a mile from the house.
I have Location Services and Notifications (including DataMan) enabled...
I just dictated the following into Nuance iP4S email content:
"Where have all the flowers gone long time passing where have all the flowers gone long long ago"
and according to DataMan:
After Dictating, before sending:
Cell: 0.0 KB WiFi: 474.8 KB
Added To (my email) and sig (Sent from my iPhone) and sent it:
Cell: 0.5 KB WiFi: 299.5 KB
Interesting, that 0.5 KB cell on the sending of the email...
DataMan shows a map with flags (I think they are cell towers). The email involved a flag about a mile from the house.
I have Location Services and Notifications (including DataMan) enabled...
That's interesting. Looks like it is processing the data locally with lossless compression before sending.
I wonder if that 0.5KB for the cell is just your phone pinging the towers per standard cellphone upkeep or if that is Siri using the towers for some unknown reason.
PS: In exactly 2 weeks I'll have an iPhone 4S.
When you had Siri off did you test the Voice Control options?
Yes... you can't go back... ...it impersonated Siri's voice, though...
Later:
me: "Siri, how did your lobotomy turn out?'
Siri: "Checking on that for you..."
Siri: "How about a web search for "how did your lobotomy turn out"?
me: "not now I have a headache"
Siri: "I found 11 drug stores... 6 of them are fairly close to you... yelp list"
Yes... you can't go back... ...it impersonated Siri's voice, though...
I noticed in one of the iOS 5.0 Betas on my iPhone 4 that the voice had changed to what we now know as Siri's voice. I assume that all iOS 5 devices have that voice.
I don't know how they can get it as low as 63kb. I would think it would be a lot more for a request and a response. Just loading a typical Google's search page as minimal as it is uses 57kb just for the response. That doesn't take into consideration the original request which if it is compressed audio has got to be more than 63kb total.
Last I looked, there were over 3 pages on comments to the linked ARS article.
I think it was 63 KB for Siri requests that didn't fail through to a web search.
The audio recording size and compression was discussed, but it is beyond my area of expertise.
Some seemed to think that the typical request could be filtered/compressed on the iPhone before sending it to Apple's servers.
That appeared reasonable.
Others, seemed to think that after the server did the speech to text, the text was sent back to the device for parsing and decoding.
That made no sense at all -- because many decoded requests would, then, be sent back to the server to complete the requests (another handshake and data transmission).
Others, seemed to think that after the server did the speech to text, the text was sent back to the device for parsing and decoding.
That makes no sense. That's the part of Siri's brain that needs to be parsed on servers. And we can see that Siri doesn't work and even tells you it's removing your information for parsing your requests will be removed from Siri's servers when you go to turn it off.
Some possible responses:
"Siri, how much cellular data are you using?"
Some possible responses:
"Thiiiiis much!"
"More than if you hadn't asked me this stupid question."
This could have an additional effect. If Siri is doing the Google search from the Apple server, Google loses the ability to see the data of who initiated the request. No data, no ads for them. THAT'S INSANELY GREAT!!!
Not to mention that Apple could and probably is designing their own search engine. This could be part of Steve-o's win at all costs strategy.
My most fun AI interaction was always Dr. Sbaitso, if anyone remembers that
But tell you what, Apple has cracked it. Talk naturally to something friendly that assists you in natural day-to-day tasks.
If they keep on improving, they've got a killer new user interface. Open the API to apps and Boom! The next app gold rush will be underway. There'll even be counselling and religious apps based on Siri API interfaces.
There's a lot of work to be done though, the voice recognition needs to be a lot more adaptive, I am a native English speaker but fall somewhere right in between British, American and Australian English. The British recognition seems to work best, though sometimes it is way off base since it is command-based more so than dictation-based unless you specifically tell it you are dictating something eg. text messages.
But we are just maybe five years away, if Apple keeps going in the right direction, to how the crew talks to the Icarus AI and how it responds in the movie "Sunshine". Interesting trivia for those who enjoy that movie: Danny Boyle mentions in the commentary that the "computer voice" was recorded by an actress not as canned responses separately in a sound studio but during the actual filming of the live actors in the scene. So there is an uncanny feeling to the dialogue whereby you know the AI is, well, synthetic but the AI is also engaging and responsive in a way that Siri can be by the end of the decade.
In any case, as with most things tech, sometimes it is all about timing. Making voice recognition and AI cloud-based was always the way to go because of the level of raw data you can collect and work with. Despite CPUs being powerful enough to process tons of data, the cloud still can collect and store that data at a level beyond local storage ever could, at least for this kind of AI application. This is big.
Not to mention that Apple could and probably is designing their own search engine. This could be part of Steve-o's win at all costs strategy.
More like another example of his common-sense strategy. For many of us, Steve doesn't seem so crazy and radical, just really, really sensible sometimes... Even if he didn't think so himself most of the time.