Tim Cook: Apple will show off 'cool ideas' for Siri in coming months

Cook's remarks on Siri and how Apple has no intention to create no original content are the two most direct things in the interview about the future of Apple.

He was talking about YouTube-like services when he was talking about "original content".

Walt directly asked him about Apple producing their own original content like Netflix and MS, for the Xbox, are doing. That isn't about simply streaming videos from YouTube which the AppleTV already does or having a YouTube account which Apple already has.

• http://www.worldtvpc.com/blog/microsoft-hiring-nbc-staff-xbox-original-tv-content/
• http://www.hollywoodreporter.com/news/netflix-ted-sarandos-original-content-309275

Start by adding SIRI to iPad, please!

Based on his comments Apple won't be producing or investing in any original content any time soon.

They added the first part of Siri to the iPad, the voice dictation, but I doubt they will add the rest with iOS 6 if they didn't add it with the iPad (3). I also doubt they will add anything more than that to OS X Mountain Lion. I hope they add it latter on but for now it seems like an iPhone-only feature for the full Siri engine.

 

That was my take on it as well. And I think that it is the right way to go. Apple should stick to being the distributor paid on the back end not the producer when it comes to music, movies and tv. If someone like one of these cancelled tv shows wants to keep going on their own dime and 'air' via online venues like iTunes, or create a show straight to online,  I think they would be okay with that but I don't think they would or should front any of the production costs out of Apple funds. Same with being a book publisher in the sense of fronting money to authors for writing, or 'third party' app development funds. 

 

Maybe it will be on more Apple products once it's out of beta. I do find myself thinking about holding off on an iPad until Siri is there for sure.

 

I also want to have more Siri "personalities" to choose from. I'd love a Barbara Eden personality and she wouldn't even have to call me "master."  

I have to think they've thought of it. I know I've wanted them to consider it for a long time but I can see an issue with the largest company in the world that has a knack of becoming the dominant force and scaring their partners simply because they so damn efficient and successful if Apple planned to release original content. I can see a lot of content owners not wanting to play ball with Apple even if cost them profit simply because they think it will cost them even more profit in the future.

Remember the original Apple TV concept? An unfinished product oddly announced and demoed by Apple without even having an established name? I think that was done to show the TV and film content holders that they had a secure and safe way to distribute their content. As we know that didn't work out for them as the Apple TV struggled with just the Disney umbrella as they very slowly gained traction with them. I think they were scared because of the control they saw shifting from the music industry they helped save. It was just too much for them to jump in without excessive consideration and diaper changing.

I hope assume you mean the young Barbara Eden.

Hopefully this hints at Maps/Locations integration in international markets.

 

I'd really like to be able to request local restaurant reviews, directions etc.

Double down? Not good enough. TRIPLE DOWN.

The problems with Siri are fundamental AI problems. Tim is wrong to think it will gradually improve over time. It will stay much the same for many years, and then someone will make a breakthrough and it will suddenly be awesome.

Improvements in AI will definitely come over time but I think the biggest problem with Siri is its complete reliance on the network. Sure, you can't do search or natural language processing if there's no network but there are plenty of things that could be done locally on the phone a la Voice Control. Until then Siri will remain just a novelty. I'm hopeful that Apple will implement something like this:

 

http://jeff-with-a-g.blogspot.com.au/2012/05/siri-ously-apple-could-you-fix-this.html

 

If Netflix thinks he's going to be the next Paramount Studios they'd better close up shop now. Same goes for Microsoft.

I'd like local processing for all of Siri but that isn't possible right now for several reasons.

I'd like for the voice commands to be more intelligent locally so it knows which commands it can do locally and which commands it has to send to Siri's servers for first and second tier analysis but that's isn't feasible for several reasons if they want Siri to be great.

If you don't like Siri you can always turn it off and your pre-iPhone 4S voice commands will be back again. You''ll get the same limited scope commands but they will local and therefore not reliant on the network. It is, however, far from being a novelty.

There may be dramatic changes over time with specific innovations but it will improve gradually because one of the biggest hurdles is getting any system to understand your speech and convert it to text with the proper word use. This can only be done by analyzing a vast number of samples and has no barring on improving the "personality" of the Ai or its algorithm.

If it can't understand what you mean — something we can't even get right on this forum when writing in English — it's certainly not going to be easy for a computer to convert speech from any number of languages to text in the right context and then comprehend that context for a specific person of any age, gender, culture or whatever other demographic that might fit into to make sure what they are thinking is what Siri interprets.

We've already seen Siri improve since its debut in October which is expected since these changes can be added to the server at any time they wish.

Me: "Set an alarm for Thursday at noon
Siri: "Sorry, I can't set an alarm for more than one day ahead"

Me: "Set timer for one and a half minutes"
Siri:: "I can't set the timer for a specific time. Would you like an alarm instead?"

Me: "Add something to my calendar"
Siri: "What time is your appointment?"
Me: "Noon"
Siri: Instead of then asking me what I want to put on the calendar, Siri just puts the text "something" on my calendar.

The first one looks like a limitation in the complexity of the Siri's AI to set a more complex calendar entry. The second and third look like a limitation in the contextual understanding.

I'm not sure why people expect a digital beta service to have more comprehension than a human being but it's unrealistic. This is where this tech needs time to learn. Take your "something" comment as an example. You made an ambiguous comment where "something" was the subject but you want Siri to know that you don't want the word "something" added but for it ask you if you another question about that something is. Why not tell Siri what that something is when you made the request?

Can the service get more intelligent? Absolutely, but users need to be intelligent enough to know what that the limitations of all technology are.

 

 

As a young lad, all of 10 years old, I used to watch I Dream of Jeannie and wonder why I liked hearing her call Major Tony Nelson by the name of "Master".  It took me about another 5 years to really discover why.

 

She was hot, and seeing the naked midriff every week was the topper.

 

 

Sorry for the off-topic interjection, carry on....

And yet you can say "Play something by The Beatles," and it works.

 

It's seems like the system is just a collection of special cases, with no general understanding. That is one reason it has to be server-side, there have to be people in the loop to help it understand new things. If Tim intends to expand the domains it operates over (which it what it sounds like) it will be interesting to see how many domains it can do before it collapses under the weight of complexity.

The initial command is "play" which automatically makes it an iPod command. The difference is that you can't choose to play a particular song by any artist at this point (unless they've updated it). That is a more complex command. You either play an artist which is indicated by the "by" signifier after "play", you play an album or you play a playlist.

The only subjects it's looking for are the artist or album, yet with a calendar entry you are asking it to do something considerably more complex. From your PoV as a human the phrasing is related but it's not a human. It will always be inferior to the way we use language.

What happens when you ask Siri, "Play some album by The Beatles." Does it then ask you which album you wish to play, does it say it can't find "some", or does it choose the closest thing to the "some" waveform?

I'm tempted to say that Siri is in fact not based on AI at all, but on a huge library of manually submitted inqueries and actions.

I get the feeling Apple is getting this huge, HUGE list of inqueries that Siri can't, or isn't allowed to process yet, that they save in a database. Until the inqueries are manually solved with proper instructions and dialogue by Apple employees, Siri won't do anything. What's so AI about that? I hope they improve on all this.

It's absolutely based on AI. There are clearly many algorithms in place that aren't simply reading a database of text strings and then spitting out a static reply. There are also canned responses for certain, common inquires that help give Siri it's personality but that doesn't mean it's not based on AI. Even the Kinect is based on AI.

"Siri, turn on my Wifi!"
"Siri, turn on my 3G!"
"Siri, send SMS/EMAIL to people X when I arrive at the local Y"
"Siri, send SMS/EMAIL to people X at time HH:MM"
"Siri, connect to WiFi called My_Wifi_Acess_Point!"
"Siri, unlock my iPhone/iPad!" (voice recognition?)
"Siri, lock my iPhone/iPad!"
and so on...

I wonder if they'll add the idea I emailed him quite a while back. Integrate Shazam with Siri so you just have to ask "who is this", "what artist is this", "listen to this" etc.

It'll tell you who and give you a link to buy it from iTunes increasing the sales opportunity.

 

Overlord - "lock my iPhone/iPad" are you sure! That would require you to press and hold the home button and then asking Siri to lock your iOS device. The alternative is to just press the lock button, job done.

There are a lot of settings that I wish Siri could access but I'd settle for display brightness for right now.

 

You absolutely can choose a particular song by any artist. I just did a few tests (including tests with songs that have the same name by different artists) and it worked perfectly every time.

Can you type out the query and result? The commands or Siri for a song I'm seeing in reviews are for a song name in the iPod DB without being able to first separate out by artist or album first. As I said, they could have added this as this is the benefit of server-side and proof of gradual updates, but people still shouldn't expect miracles as once you state a key term the method of contextual understanding will alter as we can see in the iPod and Calabder apps.
Edited by SolipsismX - 5/30/12 at 8:14am

Just curious... why do people defend Siri by saying its in Beta.  Isn't the fact it's in Beta 8 months after its launch starting to be worrisome on its own?  This is technology- it wasn't developed 8 months ago, it was developed years ago- and has been in a public beta for 8 months (an eternity in technology), with every iPhone commercial, pimping its "ability".

 

Please- save the "beta" talk... I see the enormous potential in Siri as most of you do, but Apple needs to get on their game about it or stop promoting it.  I feel it was released far too early, but their backs were against the wall to release a new iPhone and had to do something different to appease people.  The 4S by itself was enough and they should have waited.  Just my meager opinion.

 

Gmail was in beta for, what, four to six years. Think "worrisome" thoughts about everyone equally.

 

 

They're on their game. You act as though Siri is a failure.

 

Nope.

This is more aVoice command beef than Siri. I think it kind of sucks. At least the more basic commands.

Just as a test I asked it 5 times in a row to "play Green Day" and 3/5 it played Beck, one time it played The Urge and the last time it told me, "no match found". Yesterday I asked it about 5 times to play Metalica in it went back to Beck. VComm loves Beck. Must be aphabetically the first artis in the list and plays as a default.

Now, I just realized that "play artist" doesn't work well but if you tell it "play something by artist" or play X track on artist's album", it works like a charm. Also telling it to play specific tracks is useless, but asking it "play something else by artist" works great.

So in sum, hate voice com. Could be because I'm using a 3GS to test it out but I'm just shocked that Jobs let this out the gate in the sad state its in. I don't suppose the iPhone 4 (non-S) would be any better. Must be the horrible mic.
Edited by antkm1 - 5/30/12 at 8:52am

 

First off- it was a little over 2 years (Feb 07'-July 09').  But more importantly, didn't your mom teach you that because one company does something everyone doesn't have to?  Seriously- you're gonna compare Apple with Google as if they should do things the same?  This is Apple- I, for one, hold them to a higher standard than Google (as do the other stockholders it seems).

 

 

 

Can't really debate if its a success or a failure as that's relative.  But there is no question It could have been much better.  There are plenty of complaints about Siri- you follow this forum, you should know better than most.  And these are coming from Apple fans.

 

Yup.

 

Gmail came out in 2004.

 

 

No, as she's not a businesswoman. She did, however, teach me to treat all others equally.

 

 

Who… are shorting the stock because of Siri? I don't understand the implication.

 

 

Sure, some people are having problems. I understand that completely. Apparently this instantaneously means "Apple should give up" to you.

 

 

Go ahead and try to imply that Siri was the only change in the iPhone 4S. Go ahead. 

Your primary point makes absolutely no sense. There is no possible way that a service that requires voice samples to understand languages, dialects, phonemes, pronunciations, other parts of language, various demographically based terminology and other cultural context (making this list very brief) can get this sort of understanding from being in a lab and coded by a team of engineers. I am absolutely astonished that anyone would actually think it can be done as easily as adjusting the color on an app icon.

Furthermore, Google had their speech-recognition system in place for 3.5 years before shutting it down. It's believed, but not verified, that Google shut it down once they felt they had enough samples to use but that happened in 2010 and we still haven't seen any proof that is the case. Apple had their iWork.com beta running for 3.5 years, also, before shutting it down. That is even less complex than capturing voice samples for a very long period and could be coded in a back room but a small team yet shows that a beta for 8 months isn't unrealistic or a very long time in tech.

Now I do agree that Apple's Siri ads aren't representative of the real world. I have no problem with them shortening sequences but i do expect that phrases stated by the actors are actually phrases that Siri picked up during the making of the ad. Of course, professional video isn't shot that way as a general rule but in this case I think it's important for Apple to only use clips that do represent successful Siri queries. I'd also like them not to use terms like 'gazpacho' when trying to show Siri off until its success rate across many accents et al. reaches a set point.

I do think the John Malkovich ads are great because they use a single word that is easy for Siri to parse. Notice how humans typically need more words to parse context but computers are more likely get confused about the context. This, at least, should be a clue to people.


PS: This site asks a simple question and gets back a fairly simple yet thorough answer from many users that should help you and others understand why it can't just be a simple hardcoding to a few keywords.

• http://stackoverflow.com/questions/8382023/why-is-speech-recognition-difficult

 

Public beta for gmail was 2007.  Public beta for Siri was when the iPhone 4S was released.  I try to compare apples to apples.

 

 

So you expect Samsung and Google to make as intelligent of decisions as Apple?  Naive... but ok.

 

I was saying that both myself and stockholders value Apple as more valuable and worthy of investment than Google (which is factual- thats the implication).

 

 

 

 

Yeah, they certainly shouldn't be held to as high a standard as Apple has set… :this_is_where_the_confused_gif_would_go_if_we_had_a_working_one.gif:

 

 

Good. Thanks for clarifying your meaning on that. No problem, then.

 

I started by asking the example I think someone listed. "Play Let It Be by The Beatles" and she understood that. Then I thought I'd try with a song that has the same name by different artists. I know I have a few different songs called "Beautiful" in my library, so I tried just asking "Play Beautiful" to see what would happen; Siri responded by playing an album I have that's named Beautiful. I then tried "Play Beautiful by Jonah 33" and she played just that song. Knew exactly what I was talking about.