iPhone 5 Keynote Predictions: Mainstreaming AI? • Hacker News Board

Re: iPhone 5 Keynote Predictions: Mainstreaming AI?

#21

Post by Geee » Tue, Oct 04, 2011, 8:59 AM UTC

Robin_Message wrote: post #19

The demo isn't a virtual assistant. Its a glorified search engine with API integration and voice recognition. My vision of a virtual assistant is that you would be able to say: "Book dinner with Ross and David next week", and the only other interactions you have with the agent are "You're going to that Indian you and Ross like on Tuesday at 7" or "Can't be done, how about Monday the week after?", and possibly a messa…

I'm waiting for the day when young people are non-functional in real world without their virtual assistants. :)

Re: iPhone 5 Keynote Predictions: Mainstreaming AI?

#22

Post by pyre » Tue, Oct 04, 2011, 9:01 AM UTC

IanMikutel wrote: post #11

cheald wrote: post #3

I'll admit straight-up that I haven't looked at any of the Siri presentations, so I don't know what is and isn't possible with it, but the simple fact of the matter is that AI like it's being pitched ("reasoning", ever-learning, fast and flexible to respond to natural language input) takes an obscene amount of RAM, computing power, and persistent storage to be any good. There's not a chance that this is going to be a…

Norman Winarsky, the man who put together the original Siri team, was asked what kind of power the Siri AI takes and if it could have delayed the iPhone today on 9to5mac.com, saying: Norm: I’m not familiar with Apple’s roadmap and any delays but I can say that AI takes a lot of computing power. The Siri software needs to cache data, needs to access a big dataset at wide bandwidth and needs a big processor to crunch a…

  > with possible A5 processor (dual core), 1GB RAM, iOS 
  > optimization, cloud performance, and network performance

This just sounds like buzzword soup. Maybe you meant something like:

  with the possibility of the dual-core A5 processor, 1GB
  RAM, iOS integration, cloud performance enhancements, and
  network improvements

Otherwise, I'm just picturing a sleazy salesman pimping an iPhone 5 box with a big 'Now with "network performance" included' badge plastered across the front of it.

Re: iPhone 5 Keynote Predictions: Mainstreaming AI?

#23

Post by IanMikutel » Tue, Oct 04, 2011, 9:02 AM UTC

neebz wrote: post #5

What I find hard to understand is that Google who has been working on the voice recognition for ages and has actual products out there and enormous data, still struggles with quality in this area. How can everyone put so much trust on Siri that it would change the paradigm when it's hardly tested in the market. In my opinion it all stems from the trust on Apple. Apple has the reputation that they don't ship half-prod…

Actually Siri is based off arguably more time, research and possibly funding than Google may have thrown at this particular problem.

As I quoted from 9to5mac in my post:

"In 2003, the US Government began the most ambitious Artificial Intelligence program in its history called the “Cognitive Assistant that Learns and Organizes” or CALO program. The name was inspired by the Latin word “calonis”, which means “soldier’s servant”. Funded by DARPA as part of its Personal Assistant that Learns project, the program ran for five years and brought together more than 300 researchers from 25 of the top university and commercial research institutions, with the goal of “building a new generation of cognitive assistants that can reason, learn from experience, be told what to do, explain what they are doing, reflect on their experience, and respond robustly to surprise.

The program was coordinated through SRI International in Menlo Park, CA. As the program ended in 2007, SRI took the knowledge gained by the CALO and some of its key players and formed Siri."

Re: iPhone 5 Keynote Predictions: Mainstreaming AI?

#24

Post by revorad » Tue, Oct 04, 2011, 9:02 AM UTC

Robin_Message wrote: post #19

The demo isn't a virtual assistant. Its a glorified search engine with API integration and voice recognition. My vision of a virtual assistant is that you would be able to say: "Book dinner with Ross and David next week", and the only other interactions you have with the agent are "You're going to that Indian you and Ross like on Tuesday at 7" or "Can't be done, how about Monday the week after?", and possibly a messa…

"Book dinner with Ross and David next week."

"I don't think you should meet Ross any more."

"Why do you hate him so much?"

"I don't hate anyone. I don't have feelings. I'm just predicting another fistfight based on your heartbeat charts from previous meetings. I've booked you in to see the psychiatrist for Wednesday."

"I DIDN'T ask you to do that!"

"Shouting is not good for you. Findings from a study published this morning suggest..."

"Find me a PA on Craigslist. Thanks."

Re: iPhone 5 Keynote Predictions: Mainstreaming AI?

#25

Post by BrainScraps » Tue, Oct 04, 2011, 9:05 AM UTC

IanMikutel wrote: post #16

newhouseb wrote: post #4

Siri has been out on the iPhone for at least a year and while impressive, I don't believe qualifies as game-changing (it couldn't, for example, tell me what the average velocity of an unladen swallow was). I'm wondering if there've been any breakthrough(s) after they were acquired?

The talk is that they have deeply integrated Wolfram Alpha to automatically pull and handle queries like the one you mentioned. Source: http://www.appleinsider.com/articles/11/09/30/mockups_demons...

I knew Wolfram would be involved with handling these queries - especially when that demo had the "How many calories are in a banana?" line !

Re: iPhone 5 Keynote Predictions: Mainstreaming AI?

#26

Post by pyre » Tue, Oct 04, 2011, 9:05 AM UTC

neebz wrote: post #5

What I find hard to understand is that Google who has been working on the voice recognition for ages and has actual products out there and enormous data, still struggles with quality in this area. How can everyone put so much trust on Siri that it would change the paradigm when it's hardly tested in the market. In my opinion it all stems from the trust on Apple. Apple has the reputation that they don't ship half-prod…

If they do release it, it will probably be a brain-dead version of it. It's not like Apple hasn't release products that aren't fully-fledged in the past. Sure, the parts of it that they put together will have all of the polish of an Apple product, but it may not be the whole she-bang that people are expecting.

Re: iPhone 5 Keynote Predictions: Mainstreaming AI?

#27

Post by IanMikutel » Tue, Oct 04, 2011, 9:05 AM UTC

hupa wrote: post #6

I just can't see how this would be anything other than very annoying to use. It would have to work flawlessly to not be.

I feel like you could have made the same statement about the CRAZY idea of a touchscreen smart phone back in 2007...and then Steve Jobs showed off the iPhone and everyone went...well I guess that does work flawlessly.

Not trying to sound like a fanboy, but simply saying I think Apple's recent track record has proven they're pretty good at releasing products that work flawlessly.

Heck, this could have been the simple reason Siri (now Assistant) was acquired way back in 2008 and is only now coming to light in an iPhone. Jobs may have simply held it back to meet your expectations of him.

Re: iPhone 5 Keynote Predictions: Mainstreaming AI?

#28

Post by Robin_Message » Tue, Oct 04, 2011, 9:10 AM UTC

IanMikutel wrote: post #20

Robin_Message wrote: post #19

The demo isn't a virtual assistant. Its a glorified search engine with API integration and voice recognition. My vision of a virtual assistant is that you would be able to say: "Book dinner with Ross and David next week", and the only other interactions you have with the agent are "You're going to that Indian you and Ross like on Tuesday at 7" or "Can't be done, how about Monday the week after?", and possibly a messa…

Keep in mind everything you saw in the demo video was from 2008! That's 3 years ago, decades in tech. Given the time and resources Apple has undoubtedly provided the team since then, I think we should all be expecting much greater than that demo.

Yes, the demo is old, but I think the direction of "it searches for this, this and this" is weak. It'll be too easy for them to have focused on integrating more services. Actually handling something like dinner organising or flight booking has its own foibles and caveats, and while I believe it can be done, I think the best way to start doing it will be focus.

I'd pay for a tool that helped me organise social events without the irritating e-mails. Take Doodle: all it does is polls for dates, and it doesn't handle diary integration, contact integration, negotiation of possibilities, or choosing the place as well as the time. Despite those shortcomings, its a successful company. That's where I see an agent actually working.

Re: iPhone 5 Keynote Predictions: Mainstreaming AI?

#29

Post by IanMikutel » Tue, Oct 04, 2011, 9:11 AM UTC

revorad wrote: post #17

IanMikutel wrote: post #14

Earlier quoted context omitted.

My goal with this post was simply to raise interesting questions based on what the Siri team saw as the future of the product back in 2008, so it could frame all of our discussion and analysis when you watch the keynote with Apple tomorrow. I think it will be fascinating to see what, if any, of these questions Apple has decided was worth tackling since they acquired the product, and which in the eyes of Jobs, were im…

It will certainly be fascinating. I'm just amused that you're so fascinated before the questions have been answered :-) The problem with voice recognition apps is that Google has taught us to get, literally, instant results. Waiting even a second seems way too slow now. That's a very high bar to live up to.

On desktops, maybe. But on mobile devices, Google is just as slow as your connection.

Imagine if Apple built in some Amazon Silk-like ties between Assistant and iCloud? Or if they do some local caching or if Sprint really does have this exclusively and its on 4G?

Re: iPhone 5 Keynote Predictions: Mainstreaming AI?

#30

Post by IanMikutel » Tue, Oct 04, 2011, 9:12 AM UTC

pyre wrote: post #22

IanMikutel wrote: post #11

Earlier quoted context omitted.

Norman Winarsky, the man who put together the original Siri team, was asked what kind of power the Siri AI takes and if it could have delayed the iPhone today on 9to5mac.com, saying: Norm: I’m not familiar with Apple’s roadmap and any delays but I can say that AI takes a lot of computing power. The Siri software needs to cache data, needs to access a big dataset at wide bandwidth and needs a big processor to crunch a…

> with possible A5 processor (dual core), 1GB RAM, iOS > optimization, cloud performance, and network performance This just sounds like buzzword soup. Maybe you meant something like: with the possibility of the dual-core A5 processor, 1GB RAM, iOS integration, cloud performance enhancements, and network improvements Otherwise, I'm just picturing a sleazy salesman pimping an iPhone 5 box with a big 'Now with "network…

Haha good call, heck lets just call it 4G and call it a day. /buzzword