How Apple Approached Developing Siri's Personality

[vrDrew]

My worry is that Apple might be raising people's, who don't understand, expectations that it's something that it's not.

If you've been around as long as many of us have, we know they've been working on AI for decades and are still miles away from it in reality.

I don't think Apple is selling this as "artifical intelligence" or anything like it. People aren't stupid, and they know full well you aren't going to carry a state-of-the-art supercomputer around in your pocket.

Siri is being sold as exactly what it is: A Beta release of an interesting development in voice control of smart devices. Its not "magic" for the simple reason that most rational people know that such things don't exist.

Siri IS "magic" in the same way that a movie, or a book, a song, a bottle of wine, or a delicious meal can be "magic." It amazes and delights the participant. Thats what Apple has done.

They've taken some interesting developments in sound processing, computational algorithms, and married it with high-speed data networks, and massive cloud computing.

You couldn't have done Siri ten years ago - even if they'd had the software and the computing horsepower. You couldn't do it because there was no way you could put a client computer smart enough to process the speech into your pocket. You couldn't do it because the WiFi and broadband and 3G networks capable of shovelling all that data too and fro hadn't been deployed yet. And you couldn't do it because there wasn't a worldwide market for smartphones numbering in the hundreds of millions. In other words, the economies of scale weren't there to pay back the investment in software and servers necessary to do it. Thats what Apple's "magic" was.

Apple has once again changed the game. Instead of a few hundred researchers playing around in laboratories, the "voice control" and "voice interaction" world is going to be made up of tens of millions of normal people, going about their business. And Apple, and the Siri team, are going to get a massive amount of "real world" feedback. What sort of things do people ask the phone to do? What works really well? What fails? And they can, and WILL, keep tweaking the system. So that day by day, month by month, and release by release, Siri is going to get more "magical."

Look at the planning and forethought that went into Siri. Apple knew the sort of silly questions people were going to ask. And they made sure that Siri was ready with responses that would amaze and delight the new iPhone owner. If that sort of planning and execution isn"t "magic" - then I don't know what is.

PS: Please don't ask silly questions about bombs and highjacking. There is a name for that sort of activity - terrorism. If someone overheard you folling around like that it could very well cause panic and disruption. Siri might be smart enough to ignore that sort of foolishness. But a lot perfectly normal human beings might not find the joke so funny.

 

[Piggie]

I don't think Apple is selling this as "artifical intelligence" or anything like it. People aren't stupid, and they know full well you aren't going to carry a state-of-the-art supercomputer around in your pocket.

Siri is being sold as exactly what it is: A Beta release of an interesting development in voice control of smart devices. Its not "magic" for the simple reason that most rational people know that such things don't exist.

Siri IS "magic" in the same way that a movie, or a book, a song, a bottle of wine, or a delicious meal can be "magic." It amazes and delights the participant. Thats what Apple has done.

They've taken some interesting developments in sound processing, computational algorithms, and married it with high-speed data networks, and massive cloud computing.

You couldn't have done Siri ten years ago - even if they'd had the software and the computing horsepower. You couldn't do it because there was no way you could put a client computer smart enough to process the speech into your pocket. You couldn't do it because the WiFi and broadband and 3G networks capable of shovelling all that data too and fro hadn't been deployed yet. And you couldn't do it because there wasn't a worldwide market for smartphones numbering in the hundreds of millions. In other words, the economies of scale weren't there to pay back the investment in software and servers necessary to do it. Thats what Apple's "magic" was.

Apple has once again changed the game. Instead of a few hundred researchers playing around in laboratories, the "voice control" and "voice interaction" world is going to be made up of tens of millions of normal people, going about their business. And Apple, and the Siri team, are going to get a massive amount of "real world" feedback. What sort of things do people ask the phone to do? What works really well? What fails? And they can, and WILL, keep tweaking the system. So that day by day, month by month, and release by release, Siri is going to get more "magical."

Look at the planning and forethought that went into Siri. Apple knew the sort of silly questions people were going to ask. And they made sure that Siri was ready with responses that would amaze and delight the new iPhone owner. If that sort of planning and execution isn"t "magic" - then I don't know what is.

PS: Please don't ask silly questions about bombs and highjacking. There is a name for that sort of activity - terrorism. If someone overheard you folling around like that it could very well cause panic and disruption. Siri might be smart enough to ignore that sort of foolishness. But a lot perfectly normal human beings might not find the joke so funny.

It may be wrong of me, but I do believe there are many people stupid enough to think Siri is real AI in a phone.
Let's face it. Apple are deliberately targeting the non computer people with their iProducts. Those who don't and don't want to know how things work.

It's the way we all know Sony make the iPhones Camera, and I hear Apple saying about their new camera they have fitted to the phone, but all I can think about is, they should be saying they fitted a camera made by Sony to their phone. The majority of Apple customers probably think Apple actually make and design all the parts of the phone it's it own factories, not just fit bits make by others into it's case & tie it together with their own software.

On the Bombs remark, what I meant that in response to someone suggesting there are those who may be listening in to things that are said to Siri.

I would not want to think there is anyone who could track what people say and where they are.

 

[faroZ06]

Ah, once again Apple makes a reference to Star Trek when you ask Siri to beam you up. This looks really nice, but I've never used it. Does it listen accurately?

The thing is, I'm kinda disappointed that there are a bunch of articles on the 4S, but there are no Mac-related articles recently. I thought the 4S was a disappointment like the 3GS was... Evolutionary.

----------

It may be wrong of me, but I do believe there are many people stupid enough to think Siri is real AI in a phone.
Let's face it. Apple are deliberately targeting the non computer people with their iProducts. Those who don't and don't want to know how things work.

It's the way we all know Sony make the iPhones Camera, and I hear Apple saying about their new camera they have fitted to the phone, but all I can think about is, they should be saying they fitted a camera made by Sony to their phone. The majority of Apple customers probably think Apple actually make and design all the parts of the phone it's it own factories, not just fit bits make by others into it's case & tie it together with their own software.

On the Bombs remark, what I meant that in response to someone suggesting there are those who may be listening in to things that are said to Siri.

I would not want to think there is anyone who could track what people say and where they are.

The good thing is, Apple is the only company that is smart enough to combine all of these things well. That's their winning formula. Combine the wifi router (AP Express) with iTunes, combine Mac with iOS, combine iLife and iWork, combine phones and iPods, combine backup and wifi routers... then have a good support system for it all. Meanwhile, my Sony CD drive only works on my Mac, and Windows is made by a separate company than the hardware. This is why I support the same company making hardware and software. The previously stupid Google finally figured this out and bought Motorola.

By the way, are they going to add Siri to Mac OS? The current speech recognition is horrible and useless in Mac OS.

 

[hkenneth]

I hope one day Siri can sing like Hatsune Miku~

 

[jouster]

It may be wrong of me, but I do believe there are many people stupid enough to think Siri is real AI in a phone.
Let's face it. Apple are deliberately targeting the non computer people with their iProducts. Those who don't and don't want to know how things work.

Not sure why s/he's getting marked down. As a co-worker of many iPhone owners I can safely say that the majority will have no idea how Siri works. Many will think all of it....uh....her resides in their phone.

 

[28monkeys]

I have been really impressed with Siri. Last night Siri blew me away when I said "Add oatmeal to my groceries note". I felt certain Siri wouldn't have any idea what I meant. A second later Siri says okay and pops up my already existing grocery list in notes and adds Oatmeal on a new line at the end of the list ... I was stunned.

----------

this is the best joke:

nice !

----------

Image Image Image Image Image Image Image Image Image Image Image Image Image Image Image

Grabbed these from Neowin, as my iPhone is still "processing"

Here I wish you can paste some of the questions:

1. Where is Steve Jobs now?

2. Does God exist?

3. Can you shut down my phone?

-----

 

[MirrorShield84]

Siri Blog

These are great!! There's also a great collection of them here: http://stuffsirisays.wordpress.com/

 

[Judas1]

I like that they are giving siri a sense of humor. The best fun to be had is when she appears to recognize she's a program. Asked "Will you marry me?", she refers to her EULA. Yes, I'm a tech geek and I love that type of humor.

 

[CodeJingle]

Siri seems to completely barf on this one "Siri, how old are you?"

 

[APlotdevice]

Okay, I'm going to put this whole tricorder/communicator argument to rest.

While the iPhone is a phone, and thus a "communicator", think about all the apps, all the different sorts of things it can do. It's much more of a tricorder analog. The iPad/PADD thing is a lot more easy to compare, but in the end of the day, the Siri team got it right having the iPhone be a tricorder. How do you use your iPhone? I'm willing to bet that for most people, most of the time, they don't use it as a "communicator" at all.

Perhaps that's one case where Star Trek has worse technology than the real world (and amazing that we got here so quick). Having to carry around two different devices, or pin something to your chest is just not as good as having a smaller-than-a-tricorder device that can do all kinds of crazy stuff and communicate too. Heck, on Star Trek they always have to manually set up their tricorders, today, we can just tell ours what to do, and Siri does it for us.

The only thing the iPhone has that remotely screams "tricorder" is its cameras. Otherwise it doesn't have any sensors for collecting information around its environment. If there is any device in Star Trek it is analogous to, it would be the PADD:

PADDs were multipurpose devices that could wirelessly access information, play back media, remotely control other devices, and even send messages.

 

[Winni]

It's funny and all, but it all still feels scripted.

That's because it is. There is no actual AI behind it, at least not more AI than there was in "Eliza" back in the 1970s. The only real difference between Siri and SpeakToIt or Edwin for Android is that Siri has more scripted answers in it - but it's by no means smarter than competing products.

----------

Anyway, more to the point. I think we can all agree Siri needs the ability to launch apps and be used inside 3rd party apps. After all, that's a basic requirement.

Buy a Samsung Galaxy S2. It has that built-in (powered by Vlingo) and it actually works. Once again, the real innovation coming from Cupertino at this point is the marketing machinery that makes people believe that Apple invented the voice assistant (and not just bought a third party product and integrated it in their OS). There's nothing really new here, the only points that might go to Cupertino might be for better OS integration and more scripted, funny answers. The rest has been around for years.

----------

You are missing the whole point. Siris breakthrough is processing and understanding normal human speech patterns and not just specific commands

That's neither new nor a "breakthrough". Play with SpeakToIt and Edwin (both for Android) - they've been doing the exact same thing for years. And even the Infocom text adventures from the 1980s 'understood' natural language as well as Siri. You can also do some research on the programs "Racter" and "Shrdlu", if you want - those programs are more than 30 years old and they eat Siri's natural language capabilities for breakfast.

Don't get me wrong - Siri --IS-- a nice and cool feature. But it's not the unique world wonder that Apple's marketing machinery says it is, and I just hate seeing them getting and taking credit for stuff that others have done LONG before them.

 

[liavman]

Winni: I do not want to start a war with a raging fandroid here but Siri's difference is in its intention inference and reasoning. As I wrote earlier, 'Buy giants tickets' means different things in SF and NYC and SIRI knows that difference and they are not as hard coded as you seem to imply. Not that other programs can not do it, I do not claim any knowledge of what the other programs you mention are capable of but it is good to keep in mind SIRI is not about voice recognition and acting on recongized speech like Google does today. Voice is just the input method. The intention inference and reasoning about it are big steps from there.

 

[CJM]

Siri has a personality? I worried about geeks having unnatural relationships with their iPhones. I realise that this will be a mortal sin and that these perverts will burn in hell for all eternity, but shouldn't something be to stop this happening in the first place. There need to be laws enacted to prohibit this type of ungodly union.

Is there a gender setting for Siri so the sin is is not made even worse?

You're joking, right?

That's another thing I tried to see if it would read me my email. SIRI: "I found 24 emails. I can't read them to you. Sorry about that."

"Open App Evernote" Siri: Opens a new note and types in "Open app ever".
"Open App Angry Birds: Siri: "Ever so sorry Mike...I'm not allowed to do that."

This is surely coming though...

That's just one step away from "I can't let you do that, Mike."

 

[CJM]

Double post. Sorry.

 

[dqfayi]

What people are asking and in what way. So, this personality will also develop as time goes by, since they can tweak it.In real world, that feedback is what helps people evolve a personality.A second later Siri says okay and pops up my already existing grocery list in notes .

 

[tenshu2k]

the singularity is approaching.

 

[daviddth]

I think the geofence problem is an iOS problem, not a Siri problem. Siri does set reminders with geofences the way I would if I manually entered it, which has potential to be very useful. I'm not sure if anyone else is having the same issue with it not actually triggering once you arrive/leave a location.

I tried them a number of times and they dont work at all. It shows me at the right location and when I arrive or leave there is simply no location based reminder at all.

Plus here with minimal mobile phone towers, we have to rely more on GPS and that kills the batteries big time

 

[Thomas P.]

No, they definitely shouldn't. Despite Cleverbot's name, it's actually dumb as a brick.

Well thats quite true dear...
Although people are calling Siri and underdeveloped non mature feature but as far as i know the siri is far better than the Cleverbot(pardon me, i mean dumbbot...)

 

[JulianL]

That's because it is. There is no actual AI behind it, at least not more AI than there was in "Eliza" back in the 1970s
...
You can also do some research on the programs "Racter" and "Shrdlu", if you want - those programs are more than 30 years old and they eat Siri's natural language capabilities for breakfast.
...
Don't get me wrong - Siri --IS-- a nice and cool feature. But it's not the unique world wonder that Apple's marketing machinery says it is, and I just hate seeing them getting and taking credit for stuff that others have done LONG before them.

I agree with your final paragraph. It also saddens me somewhat to see some people believing that Apple invented loads of new stuff here. Siri is built on the back of work from a research community that I would estimate numbers in the hundreds of thousands and stretches back for decades (50 years or more). One thing that Apple has undoubtedly done though is to bring the quest towards intelligent behaviour to the attention of more of the public and to put an attempt at getting there into the hands of what will soon be millions of people. Also, by deploying a system to millions of people, the opportunity to analyse massive amounts of data and use that data to improve the system (e.g. syntactic inference) is exciting.

I disagree with some of your examples however, either in principle or because they are unfair.

Eliza is incredibly basic, it's only a few hundred lines of code and has almost no AI beyond being able to map things like "mother" and "father" to "family", doing basic pronoun transformations such as mapping "my" to "your" when rewording a user's input to form the systems subsequent output, and doing ranked keyword analysis and there it only ends up matching for a single keyword with highest rank. We don't really know what Siri is doing in detail (or at least I don't) but it is clear from examples given that it is doing more than Eliza. Siri at least has the concept of actions upon entities that are located in time and space, and the ability to use previous context to fill in the blanks and to resolve pronominal (anaphora) references; these capabilities would seem to me to refute the assertion that it has "<no> more AI than there was in "Eliza" back in the 1970s".

As for Shrdlu, I agree that it is truly impressive and demonstrates capabilities beyond those of Siri, unfortunately it is only impressive within its own blocks world. If Apple users find utility in telling their phones to put the small red pyramid on top of the big green block then it's great, but for anything else they would get something like "Input not understood" so I believe that it is unfair to compare the two. Impressive performance in a narrow domain is very different to attempting to work in a less bounded domain. As another example of impressive performance in a narrow domain just look at all the stuff that Roger Schank and his team were doing at Yale in the 1970s, impressive demos that only worked within a narrow domain and the effort of hand-crafting the knowledge base to extend the scope of coverage was so high as to make the systems impractical.

Since you bought up Weizenbaum and Eliza, it might be worth quoting the introduction from the original 1966 ACM paper since it addresses a key issue running through this whole forum thread.

From "ELIZA--A Computer Program For the Study of Natural Language Communication Between Man and Machine" by Joseph Weizenbaum published in the journal "Communications of the ACM Volume 9, Number 1 (January 1966)":

It is said that to explain is to explain away. This maxim is nowhere so well fulfilled as in the area of computer programming, especially in what is called heuristic programming and artificial intelligence. For in those realms machines are made to behave in wondrous ways, often sufficient to dazzle even the most experienced observer. But once a particular program is unmasked, once its inner workings are explained in language sufficiently plain to induce understanding, its magic crumbles away; it stands revealed as a mere collection of procedures, each quite comprehensible. The observer says to himself "I could have written that". With that thought he moves the program in question from the shelf marked "intelligent" to that reserved for curios, fit to be discussed only with people less enlightened that he.

The object of this paper is to cause just such a reevaluation of the program about to be "explained". Few programs ever needed it more.

(Copyright I assume is either the Communications for the Association of Computing Machinery or Joseph Weizenbaum and is duly acknowledged.)

- Julian

 

[tivoboy]

nice little overview

Here is a nice little overview from tuaw of siri commands/options

http://www.tuaw.com/2011/10/05/iphone-4s-what-can-you-say-to-siri/

 

[liavman]

Julian, good work. Your quote of Joseph Weizenbaum is so appropriate.

There is a gray line between Computational Linguistics and AI and one can quibble over where that line is. Siri is definitely in the realm of Computational Linguistics.

 

[TalktothePHONE]

Post Your Favorite Siri Conversations

I created a site where users can share their funny SIRI convos:

http://www.TalkToThePhone.com

 

[vrDrew]

That's because it is. There is no actual AI behind it, at least not more AI than there was in "Eliza" back in the 1970s. The only real difference between Siri and SpeakToIt or Edwin for Android is that Siri has more scripted answers in it - but it's by no means smarter than competing products. points that might go to Cupertino might be for better OS integration and more scripted, funny answers. The rest has been around for years

Sorry, but that is just nonsense. And quite frankly the sort of wildly distorted opinion that that simply discredits Fandroids everywhere.

Lets start off by observing that Siri is an order of magnitude better than Android Voice Actions. For one Siri talks back to you. Siri is available in several languages - including English, American and Australian English. Siri also interprets a whole range of different syntax: "I want some Chinese" - rather than Voice Actions, which restricts the user to a fairly limited set of specific voice commands "Map", "Navigate" etc.

There's another huge difference between Siri and Eliza. Siri is going to be used by tens of millions of normal everyday people Not a handful of researchers in a lab someplace. If you are unable to understand the significance of that sort of development, then I feel very sorry for you.

The pooh-poohing of Siri by the handful of Fandroids who haunt these forums is predictable, and more than a little sad. If you haven't got the intellectual honesty to recognize and acknowledge the accomplishment of bringing this kind of technology to a widespread audience - and moreover doing it with the wit and style that Siri does - then thats your loss.

 

[Maumelle guy]

Wirelessly posted (Mozilla/5.0 (iPhone; CPU iPhone OS 5_0 like Mac OS X) AppleWebKit/534.46 (KHTML, like Gecko) Version/5.1 Mobile/9A334 Safari/7534.48.3)

Siri is a pimp she can find escorts for u was joking around and my friend said he was horny and she said she found escorts near us

 

[kdarling]

It's interesting what the former lead iOS programmer for the original Siri app says:

It’s not really AI

"With Natural language processing in the mix it feels more human, like it understands you. This in itself adds to the mystique surrounding Artificial Intelligence. But real AI can’t fit on a phone in our world…yet.

"Siri is basically a contextual, semantic, personalized search engine. We affectionately called it a “Do” engine. A search engine can evaluate text strings and look for matching results. A “Do” engine maintains awareness of the user and everything it knows about that user and processes strings in the context of the user."

- Ed Wrenbeck, former lead developer from Siri