Siri Creators Offer First Demo of 'Viv', the Self-Coding Voice Assistant

[aleksander]

In a Guardian article the CEO of the company say that Apple have done nothing with Siri and they effectively have sidelined it hence why they have continued without Apple. It's another Apple failure.. Buy the tech and then leave it to rot! How many times have we seen Apple buy a bit of software only to let it wither on the vine afterwards. Siri is like Mac products, it's out of date!

All of them at Apple can brag about end users and experience they are after but it's clear the internal politics and bad management ruins it and we end up having only 25% of the promised awesomeness. Siri could be twice as advanced compared to Viv if they have given it the attention it deserves (AI being the future of computing an all). I'm not holding my breath for WWDC.

 

[oliversl]

Apple seems to wan to burry every tech that Scott Forstall introduced, Siri and Maps are examples. Siri could have open API right now and Maps could have been expanding more rapidly. I can not add a business in Apple Maps outside the 1st world.

The exec team seem to be acting too much childish.

 

[CJM]

I get a bit nervous when tech demos use ideal scenarios. There is some browsing and refining involved in the hotel demo, but would you really book one that blindly? Or know exactly where you're taking your uber from and that you need to specify the number of people upfront and build a perfect query containing all of that info? Very impressive and feel like Apple could greatly benefit from their expertise, but as a 3rd party product you it'll never be as frictionless as Siri, even if more robust. My guess is that they're aiming for acquisition and negotiating their terms when brought back into whatever company acquires them, which could be Apple, or even a car maker.

Yeah, if anything, I'd like to see the assistant help the user build the query. What if you only said you wanted a taxi from such and such a place. Would it ask you for how many people?

Although I admit it becomes useful when you know exactly what you want beforehand.

 

[chfilm]

wow, this is exactly what siri shouldve become by now!
Shame on you apple! Hope they'll sell this to apple and somehow stay on board this time. Judgning by their heavy promotion of apple products in this presentation, apple seems to be a more likely candidate for a partnership than google or FB..

 

[Ant3000]

Why was the guy going up in an elevator?

 

[odds]

Why is the guy's MBP running QuickTime all the time during the demo?

Maybe so he can record his demo for later reflection. You should try reflecting on your work and experience some day, too.

 

[Razeus]

Pretty powerful. Hopefully Siri will be able to do more in iOS 10.

Doubt it. Apple is so far behind in all things tech, I don't see how they could possibly catch up. Siri STILL isn't fully functional. This blows Siri out of the water on Day 1. Heck, Cortana blew Siri out of the water Day 1. Apple has to do more than build pretty rectangles.
[doublepost=1462895652][/doublepost]

unfortunately if i have to launch Viv every time i need to make a request, this isn't going to work.

It'll still be faster than using Siri.

 

[tkukoc]

unfortunately if i have to launch Viv every time i need to make a request, this isn't going to work.

Just ask Siri to launch Viv and go from there.

 

[Jakeoster]

An example of why buying companies isn't always a great idea.

Can't buy people. If they decide to leave with their ideas and connections then leave you holding their old baggage, they will.

 

[modeyabsolom]

"Self-coding" that sounds a little scary! Maybe it'll code itself into self-awareness and take over the Internet and then the world...a la Terminator!

 

[rhoydotp]

Self-coding? It will hack computers one day

Yup! But only when it thinks that your intent is to hack, which may or may not be your real intent.

 

[macduke]

How did Apple let these guys go? I thought I read something about them clashing over the future of Siri, and this was apparently it. How could the people at Apple be so blind? Siri has barely gone anywhere since 2011. Does she understand me better? Yeah. And hardware features like "Hey Siri" have made it more useful. But her functionality hasn't improved much since that time.

You can't even ask her basic questions like "When is low tide?" She never has any context for where she is or the types of things you like to do at certain times of the day. I don't mind if Apple collects certain basic pieces of information about me to serve a better AI, as long as it syncs across my devices using end-to-end encryption through iCloud and only sends Apple anonymized general information about the types of queries I make.

For now, Siri is only good for the most basic tasks and "**************************." These guys have got it right by letting third party developers build onto Viv, using all the information she has available to figure out context and build a process to complete the task. It's very much like the process someone would use when Googling around to find a specific answer to a complex question. You get answers for each part and then use that to narrow in and find the final result.

 

[acctman]

Why is the guy's MBP running QuickTime all the time during the demo?

because you can mirror you iPhone/iPad via quicktime... where have you been :/

 

[Flight Plan]

unfortunately if i have to launch Viv every time i need to make a request, this isn't going to work.

I do this already for the Google assistant. It's worth it and it DOES work.

Me: I need directions to the nearest Smashburger.
Siri: Would you like me to search the internet?
Me: Yes.
Siri: I'm sorry "me", I can't do that.
Me: Okay Google...

 

[mildocjr]

I guess this means that the iPhone 7S is going to be huge since I doubt it will be RTM until after the iPhone 7 releases. Of course, this is assuming Apple will buy the technology, again. I sure would like to see more of what this team has in the works.

 

[Flight Plan]

Wonder if that was one of those cellphone sale calls.. 201-844-9120 so annoying.

No, it was "Windows Tekinekal Suppawt" from a heavily-accented guy named "Steve Smith" in a country not near you.

 

[tennisproha]

So this is sort of an attempt at giving AI a conscience by enabling it to assume intent. Pretty intriguing.

 

[kingtj]

While I agree that Apple seems to be suffering from internal management struggles right now (not that such a thing is exactly new to the company!) -- I'm not sure a lack of upgrading Siri is a definite indicator of bad decision-making on Apple's part?

Despite Cortana and other copy-cats coming along, people still think of Siri as THE virtual assistant, including the distinctive sound of her voice. If you're in a situation where you only have so much software development talent to go around and you have to prioritize projects -- the idea of funneling time and money into making Siri more intelligent isn't necessarily what needs to be at the top of Apple's list.

It seems to me that Siri is pretty extensible under its existing framework. I mean, any time Apple wants to give Siri more ability to answer back about something, it just needs to add the relevant queries to a database and feed it the data used to respond. That's all that's really necessary to keep Siri a relevant and useful tool.

Viv is interesting, in that it's trying to take things to the next level with ability to stack queries and figure out more about context of questions when incomplete information is given. But I see a problem with that. The more you try to fake true intelligence and ability to converse with a person, the more you broaden the expectations of the person interacting with the system. With Siri, I think people pretty quickly figure out where the boundaries are. (EG. Every time I do much beyond asking a specific question about one topic, Siri starts telling me it doesn't understand, or reverts to sending me off to web search page results. So I learn Siri is for basic, quick queries only.)

In that demo example of Viv, he asks about the weather near the Golden Gate Bridge without specifying where the bridge is. Viv is able to figure that out to give a useful response. Great - but I bet that only works because the Golden Gate Bridge is a significant enough landmark that someone thought to code in information about its location. That sets certain expectations. So when I ask about weather at the "family campground" (assuming it knows I mean the only campground in my small town), will it do the same thing?

 

[Applestock]

While I agree that Apple seems to be suffering from internal management struggles right now (not that such a thing is exactly new to the company!) -- I'm not sure a lack of upgrading Siri is a definite indicator of bad decision-making on Apple's part?

Despite Cortana and other copy-cats coming along, people still think of Siri as THE virtual assistant, including the distinctive sound of her voice. If you're in a situation where you only have so much software development talent to go around and you have to prioritize projects -- the idea of funneling time and money into making Siri more intelligent isn't necessarily what needs to be at the top of Apple's list.

It seems to me that Siri is pretty extensible under its existing framework. I mean, any time Apple wants to give Siri more ability to answer back about something, it just needs to add the relevant queries to a database and feed it the data used to respond. That's all that's really necessary to keep Siri a relevant and useful tool.

Viv is interesting, in that it's trying to take things to the next level with ability to stack queries and figure out more about context of questions when incomplete information is given. But I see a problem with that. The more you try to fake true intelligence and ability to converse with a person, the more you broaden the expectations of the person interacting with the system. With Siri, I think people pretty quickly figure out where the boundaries are. (EG. Every time I do much beyond asking a specific question about one topic, Siri starts telling me it doesn't understand, or reverts to sending me off to web search page results. So I learn Siri is for basic, quick queries only.)

In that demo example of Viv, he asks about the weather near the Golden Gate Bridge without specifying where the bridge is. Viv is able to figure that out to give a useful response. Great - but I bet that only works because the Golden Gate Bridge is a significant enough landmark that someone thought to code in information about its location. That sets certain expectations. So when I ask about weather at the "family campground" (assuming it knows I mean the only campground in my small town), will it do the same thing?

Not only that, but what happens when multiple vendors of the same vein want to plug in their service into this system? When you say, "send $20 to my friend", will it list the vendors that can do this for you and then depending on the vendor that you pick, you'll get a different experience? same goes for hotels- book me a hotel at x becomes less fluid when all of a sudden a dozen different bots try to plug in at the same time and offer you various options. It seems like the simplicity offered by this platform will only remain if there is a king maker deciding various monopoly positions.

 

[thekeyring]

Siri required a complete backend revamp when Samsung purchased the company that makes Dragon Dictate (who originally handled Siri's backend). At that time Apple set up an office right across the street from MIT focused exclusively on Siri. These kinds of things take a few years of production before going live, and we have not yet seen the end results of this team's work (much like the Map team that has been hunkered down away from publicity for the last three years).

http://appleinsider.com/articles/13...cognition-tech-team-in-boston-to-improve-siri

http://www.betaboston.com/news/2014...focused-on-improving-siri-speech-recognition/

It's iOS 10 this year, and several iOS core teams have been working on projects that haven't been released yet. It's going to be a BIG WWDC this year.

I hope you're right!

 

[eroyce]

I do this already for the Google assistant. It's worth it and it DOES work.

Me: I need directions to the nearest Smashburger.
Siri: Would you like me to search the internet?
Me: Yes.
Siri: I'm sorry "me", I can't do that.
Me: Okay Google...

I was thinking about lunch and your comment planted the idea for a burger...so I asked Siri for directions to the nearest Smashburger using exactly the phrase you listed above...and it worked perfectly. Siri understood the request, asked me to confirm I wanted the location that is closest to me, and then maps opened with turn by turn directions to get there. I'm not saying Siri is perfect, just want to give credit where credit is due. Now I'm off to eat some good food.

 

[thekeyring]

Totally agree. Makes no sense that a largely server-side service shouldn't be incrementally improving monthly in terms of responses. You'd think they could just hard code a lot of requests at this point if their sentence parsing tech isn't advanced enough.

Exactly. And Apple, who managed to shrink down iOS 9 to 1.3GB from iOS 8's 4.58GB when it put its mind to it, should be able to speed up Siri by making sure the info sent to/from the cloud was as small as possible.

 

[hagar]

What strikes me as odd is that, let's pretend Siri can be split into two halves: Speech-to-text and Text-to-Action

Even if the speech-to-text wasn't great, and you had to be really clear when you spoke to it, why isn't the Text-to-Action good? Aren't there people at Apple looking through Siri queries, and able to spend all day every day adding new functionality?

It debuted in 2011, surly even if someone was to write out the commands a user needed to say by hand they could. "Siri can't do this, but it would be cool if it could. How many ways are there to say it?"

Isn't someone ticking off a list ensuring that every function of the device can also be accomplished using Siri? Why not add feedback to the Siri UI: "Did Siri do what you wanted?" with a Yes / No. Then they can look through all the "No"s and work out what happened, how to improve the service, etc?

Yes, they hired 1 student to do this during the holidays. But mind you, he also has to enter all public transit info manually in the Apple Maps database. That's why we see so little progress

 

[colbertnation]

I'm as patient with Apple as anyone - I love their computers and their phones, and I'll probably buy a Watch when the second-gen is released.

But how they bought Siri and actually had these guys working for them and yet Apple aren't the ones making this announcement is incredible.

This presentation should have been a part of WWDC, built into iOS 10, OS X, watch OS 3.0 and the newest tvOS. I guess we'll have to wait until June 13th, but man, this will be hard to beat.

Just goes to show how inane Apple leadership has become. Their prime focus: thinness and lightness, and an increasingly harder to use/more complicated and fugly UI.

Think about it: why would the creators of Siri leave Apple, unless they thought that they had a better chance of developing an even better Siri on their own without the headache of "higher ups," and then potentially flipping it and selling it to a company that values them even higher?

 

[akfgpuppet]

Before that, please allow me to play music like this:
"Hey Siri, play 僕らは今のなかで and play this song indefinitely".

Bilingual, or even trilingual Siri is a must now. I have couple of Spanish and Japanese songs that I Siri won't play for me because the name of the song is in a diffrent language.