Analyst: Siri more than just speech recognition(forbes.com)
forbes.com
Analyst: Siri more than just speech recognition
http://www.forbes.com/sites/briancaulfield/2011/10/07/apples-next-big-thing-is-already-here-siri-more-than-just-speech-recognition-analyst-says/
24 comments
There's a lot of editorializing from the Android camp that seems to think Siri is just Apple playing catch-up to Google's voice search.
Guess we'll see who's right in five days.
Guess we'll see who's right in five days.
It's not "just" playing catch up. It's partly playing catchup, and partly expanding on it. And to be fair, the credit should go to the Siri team more so than to Apple as a whole. They innovated and Apple bought them and added polish.
So why are some Android users less than impressed at the gushing over Siri? Because we've been able to do many of the things you see in the video demonstrations for a long time now. Some examples:
"Navigate to <local restaurant name>"
"Text <person> <some message>"
"What is the weather in New York?"
"What is the time in Paris?"
Siri integrates more features in a polished way and also adds context to the queries. By context, I mean that the interpretation of the next request can depend on the previous requests.
While little has been released on the actual technical constructs of these systems, I would imagine that they work somewhat similar to this:
1 Translate voice to list of text interpretations. 2 Guess an intent based on each text interpretation, combining the translation and interpretation score. Pick the action of highest score.
The intents need to be hand coded. Siri likely modifies the second step by allowing intent inference to be based on the last command or couple of commands in addition to the current request. That is the innovation, and I use the word innovation lightly as I'm sure the idea has been thought of in academic circles.
A lot of people seem to think "you can throw anything at it" and I seriously doubt that is true. The library of intents might be large enough to fool some people, though. It isn't revolutionary, it's a well executed incremental improvement on what we already have. It's worth mentioning that there are also other applications such as Vlingo that try to expand the capabilities already present on current day systems.
None-the-less, Apple's excellent marketing team seems to have convinced people that this is revolutionary and that Apple should be credited with yet another huge world changing innovation. The media needs it's heros and there can't be heros unless the valuable contributions of the many people whose work led to this point are forgotten, marginalized, or simply reattributed.
So why are some Android users less than impressed at the gushing over Siri? Because we've been able to do many of the things you see in the video demonstrations for a long time now. Some examples:
"Navigate to <local restaurant name>"
"Text <person> <some message>"
"What is the weather in New York?"
"What is the time in Paris?"
Siri integrates more features in a polished way and also adds context to the queries. By context, I mean that the interpretation of the next request can depend on the previous requests.
While little has been released on the actual technical constructs of these systems, I would imagine that they work somewhat similar to this:
1 Translate voice to list of text interpretations. 2 Guess an intent based on each text interpretation, combining the translation and interpretation score. Pick the action of highest score.
The intents need to be hand coded. Siri likely modifies the second step by allowing intent inference to be based on the last command or couple of commands in addition to the current request. That is the innovation, and I use the word innovation lightly as I'm sure the idea has been thought of in academic circles.
A lot of people seem to think "you can throw anything at it" and I seriously doubt that is true. The library of intents might be large enough to fool some people, though. It isn't revolutionary, it's a well executed incremental improvement on what we already have. It's worth mentioning that there are also other applications such as Vlingo that try to expand the capabilities already present on current day systems.
None-the-less, Apple's excellent marketing team seems to have convinced people that this is revolutionary and that Apple should be credited with yet another huge world changing innovation. The media needs it's heros and there can't be heros unless the valuable contributions of the many people whose work led to this point are forgotten, marginalized, or simply reattributed.
That is the innovation, and I use the word innovation lightly as I'm sure the idea has been thought of in academic circles.
Just a quibble. Innovation is, according to the dictionary, the "introduction of new things or methods". This is quite different from invention, which is "an act or instance of creating or producing by exercise of the imagination, especially in art, music, etc".
Invention can be done in academic circles, but innovation needs someone to actually roll up their sleeves and introduce the damn thing - in this occurrence, introduce it to the vast, chaotic, challenging consumer electronics market.
Apple certainly takes no credit for inventing voice recognition, or any of the concepts used in Siri - in fact, Apple invents very little. However, they should get full credit on the innovation side. As an easy example, tablets were invented for decades before the iPad, but they were not effectively introduced to the consumer market until the iPad. That's innovation - turning ideas into realities - as opposed to invention - coming up with new ideas. Both are essential.
Just a quibble. Innovation is, according to the dictionary, the "introduction of new things or methods". This is quite different from invention, which is "an act or instance of creating or producing by exercise of the imagination, especially in art, music, etc".
Invention can be done in academic circles, but innovation needs someone to actually roll up their sleeves and introduce the damn thing - in this occurrence, introduce it to the vast, chaotic, challenging consumer electronics market.
Apple certainly takes no credit for inventing voice recognition, or any of the concepts used in Siri - in fact, Apple invents very little. However, they should get full credit on the innovation side. As an easy example, tablets were invented for decades before the iPad, but they were not effectively introduced to the consumer market until the iPad. That's innovation - turning ideas into realities - as opposed to invention - coming up with new ideas. Both are essential.
I would not consider this a "catch up", no one knows who develop this tech first. Google probably heard about Apple's plan and is able to bake it into Android first. Look at Google TV. Google get it into the market first, but it was a half cooked product.
Is Android half cooked product?
> And they needed an analyst to find that out?
What do analysts even do? Sounds like they just make random predictions about businesses which they don't even understand. If they turn out to be right, they take the credit. If they turn out to be wrong, they just sweep it under the rug and hope nobody notices.
What do analysts even do? Sounds like they just make random predictions about businesses which they don't even understand. If they turn out to be right, they take the credit. If they turn out to be wrong, they just sweep it under the rug and hope nobody notices.
Think of them as datapoints in a large survey. If you run the Teacher's Union Pension Fund with $10B, you can't just trust your internal team or one or two large advisors. You want to get as much information as possible about all the sectors of the economy, locally and globally. So instead of hiring 1000 people to do the research for you, you pay 10-20 different companies to access their newsletters, briefs, and suggestions, each of which is written by an analyst. If 75% of analysts say "Buy APPL" while 5% say "Sell APPL", rest being neutral, you can get an idea of what the general market sentiment is.
To put it simply, instead of asking Gruber what he thinks of Apple management, you ask HN, Slashdot, and Ars readers. Certainly none of us has connections like Gruber but it is "safer" in the sense that it is easier to justify to a board/shareholders why you lost money on Apple.
To put it simply, instead of asking Gruber what he thinks of Apple management, you ask HN, Slashdot, and Ars readers. Certainly none of us has connections like Gruber but it is "safer" in the sense that it is easier to justify to a board/shareholders why you lost money on Apple.
Does this necessarily improve your accuracy though?
http://www.bloomberg.com/news/2011-01-19/apple-s-underdog-an... http://tech.fortune.cnn.com/2011/01/19/apples-blow-out-quart...
This is one of the reasons I like to follow Apple to much. The majority of people apply the same old presumptions and rules to them and yet Apple consistently proves them wrong. Rather than adjusting their understanding of the market, most people simply repeat their mistakes. It seems the people who realise something else is going on is in the minority. What gives?
I am definitely a fan of their products. But I try hard not to let that influence whether I think Apple will succeed or not.
http://www.bloomberg.com/news/2011-01-19/apple-s-underdog-an... http://tech.fortune.cnn.com/2011/01/19/apples-blow-out-quart...
This is one of the reasons I like to follow Apple to much. The majority of people apply the same old presumptions and rules to them and yet Apple consistently proves them wrong. Rather than adjusting their understanding of the market, most people simply repeat their mistakes. It seems the people who realise something else is going on is in the minority. What gives?
I am definitely a fan of their products. But I try hard not to let that influence whether I think Apple will succeed or not.
I'm really interested to see if Apple's attempt is better than Google's. It looks so from demonstrations, but then, Voice Actions looks really great in demos too and I can almost never get it to do exactly what I want.
Just by looking where Siri comes from it should be clear it's better. The biggest difference is that Siri doesn't have commands 'hard-coded', you can throw anything at her. Thus, the whole design of Siri is completely different than in voice control systems. Only thing that's comparable is the speech-to-text accuracy.
>you can throw anything at her
The word 'her' surprised me - 'it', surely?
But I guess this just goes to show, once again, how easily anthropomorphisation comes to people; if the experience is good enough that customers start thinking about the software as 'her' rather than 'it', maybe Apple will have a hit on their hands.
The word 'her' surprised me - 'it', surely?
But I guess this just goes to show, once again, how easily anthropomorphisation comes to people; if the experience is good enough that customers start thinking about the software as 'her' rather than 'it', maybe Apple will have a hit on their hands.
Well, that was intentional. I'd like to think Siri as a person; she is supposed to be intelligent after all. :)
I'm actually only concerned about the speech-to-text accuracy. It's the foundation on top of which all this AI stuff is built. If it's not better than Google's, the AI fancy pants stuff is just going to be another layer of complexity that will screw things up.
I'd be more excited about Siri if I had a better way of talking to it. I wish Apple would design a really great A2DP bluetooth headset, now that'd I'd be excited about.
My Backbeat 903s have never really fit my ears right, so I'm looking for alternatives. Any suggestions?
My Backbeat 903s have never really fit my ears right, so I'm looking for alternatives. Any suggestions?
Try Jawbone's Era:
http://jawbone.com/headsets/era/overview
http://jawbone.com/headsets/era/overview
My Motorola s305's are pretty decent but are very sweat/moisture phobic - the buttons stop working if it gets wet enough. Also they don't do well in wind at all (impossible to use voice commands while on a bike for example).
But for ~$30 they are pretty decent... I'd love to see alternative A2DP headsets.
But for ~$30 they are pretty decent... I'd love to see alternative A2DP headsets.
This.
If Siri works as well as advertised, it will be a triumph for natural language processing as a field, not just Apple.
Getting that sort of parsing and semantic analysis into a form where it can be done by a mobile device must have taken a lot of work, and I'm sure that was only the start — Apple must have had to go back through every single data source and API in iOS to add semantic tags for Siri to latch onto.
If Siri works as well as advertised, it will be a triumph for natural language processing as a field, not just Apple.
Getting that sort of parsing and semantic analysis into a form where it can be done by a mobile device must have taken a lot of work, and I'm sure that was only the start — Apple must have had to go back through every single data source and API in iOS to add semantic tags for Siri to latch onto.
Isn't that done by a backend that's not on the phone? I know at least dictation is done remotely...
I don’t think it’d be fast enough. Even in damn-near-perfect signal conditions I have a RTT over 100 ms when I’m on AT&T 3G. That and how lame would it be if Siri ceased functioning entirely (not just the maps/other network-requiring bits) when you had no coverage?
Given that the original product that Apple bought worked that way, I'd find it hard to believe that the newly integrated one changes things that significantly. If they did get it working locally it would be very interesting.
What I am wondering, is when it will be available for people to hook their apps into!
"Can you see if there are any open reservations slots at Prime Italia tonight?"
"Can you tell me how many incomplete stories there are on the iCalledit project in Pivotal Tracker?"
"Can you see if there are any open reservations slots at Prime Italia tonight?"
"Can you tell me how many incomplete stories there are on the iCalledit project in Pivotal Tracker?"
Does anyone have any predictions on when android will have something comparable. I really just want an easy way to make calendar reminders.
Try the "Quick Event" app.
I thought Apple was quite clear about that in their presentation, there was no way any alert listener could have missed that.
Siri has the potential to be awesome because of that if it works well. We will find out soon. I’m curious.