back to article Apple struggles with the idea of intelligent life outside Cupertino

In the age-old tech struggle between open and controlled systems, Apple has realized that when it comes to artificial intelligence, it needs to edge toward open. The computer giant has announced it will be opening up its digital assistant Siri to third-party apps and at the same time has put out an API to its artificial …

  1. Steve Davies 3 Silver badge

    control Fearkery?

    Is it not better for you, the individual for the AI to reside on YOUR device than back at the Mothership where the data can be slurped and used for all manner of reasons, viz Advertising and Tracking?

    If Apple deciding to keep Siri local you your iDevice is an example of control freakery then frankly, I'm all for it.

    They IMHO have a different and better approach to Google/MS/Amazon. And frankly, given the slurping tendencies of the competition, I'm with Apple on this one.

    Personally, I would not be seen dead talking to an AI but that is just the luddite in me. Or? Perhaps I don't want to broadcast that I'm looking for [redacted] to the people around me. I certainly don't want my searched to result in product recommendations. I'm sure it won't be long before there are divorces filed where the 'dodgy' searched are cited as reasons for getting unhitched.

    IMHO, Apple are being refreshingly different here. Good on them I say.

    1. jabuzz

      Re: control Fearkery?

      Except Siri is all done server side too. Trying to do that level of AI client side is not feasible.

  2. Steve Davies 3 Silver badge

    Control Freakery?

    Is it not better for you, the individual for the AI to reside on YOUR device than back at the Mothership where the data can be slurped and used for all manner of reasons, viz Advertising and Tracking?

    If Apple deciding to keep Siri local you your iDevice is an example of control freakery then frankly, I'm all for it.

    They IMHO have a different and better approach to Google/MS/Amazon. And frankly, given the slurping tendencies of the competition, I'm with Apple on this one.

    Personally, I would not be seen dead talking to an AI but that is just the luddite in me. Or? Perhaps I don't want to broadcast that I'm looking for [redacted] to the people around me. I certainly don't want my searched to result in product recommendations. I'm sure it won't be long before there are divorces filed where the 'dodgy' searched are cited as reasons for getting unhitched.

    IMHO, Apple are being refreshingly different here. Good on them I say.

  3. Anonymous Coward
    Anonymous Coward

    Obviously massive cloud computing power will be superior to what a single phone can do

    But will it be enough better to really enhance the end user experience, and is it worth the loss of privacy? Guess we'll find out....options are good.

  4. Anonymous Coward
    Anonymous Coward

    Author could use BNNS

    All apple is doing is offering a highly optimized set of tools to allow efficiently running neural networks on iDevices. This to prevent world+dog from needing to reinvent the wheel, or worse, reinvent it poorly and waste precious battery power.

    It is impractical to perform NN learning on devices and anyone doing it likely has dedicated high power specialized setups with reams of data to train on, so there is little benefit in Apple providing any of it.

    Nowhere does it say you must use their tools. Though you would likely be a fool not to.

    Dumb conspiracy article.

  5. Mage

    AI

    Maybe simulated "Intelligence" to a point. None of this is real AI at all.

    Certainly "Siri" should only run on iThings. Besides it's not actually a local app, is it? Surely it's just an inteface to an Apple Database.

    1. LionelB Silver badge

      Re: AI

      Maybe simulated "Intelligence" to a point. None of this is real AI at all.

      What's the difference between simulated intelligence and real AI? (Not a rhetorical question.)

      1. Mage

        Re: AI

        Current computer systems can't truly learn, they can't meaningfully create, they only encompass the domain they are programmed for. We don't even have a proper definition of intelligence.

        Which so called "AI" system like Watson, Cortana, Amazon Echo, Google, Siri etc isn't just a user interface to a human curated database. Can you as an end user (not the maintaining program team) teach one ANYTHING? Can they solve any problem?

        They are just variations of so called "Expert systems", which are programs in three parts, capturing information from a trained expert, curating the database and a pseudo subset of natural language to interrogate the database.

        They are not even very convincing simulations of intelligence.

        I was considering several AI projects based on my years of programming and AI study:

        1) To have a program to learn any arbitrary card game from a human with only defining the terms and no games or goals in advance.

        2) To have a program to recognise the music in a recording of any arbitrary music, instrumental, vocal only or mixed, then create the music score for each part. Lots of not very good musicians can do this. Having a program play notation via a pair of robot hands on any arbitrary instrument isn't AI, that's almost trivial and more an exercise in electro-mechanical engineering.

        I expect, like chess or Go by computer which isn't AI, but brute force, any achievement of this would turn out to be a narrow domain trick, and not AI. Item (2) is much easier than (1) and can be achieved without any AI.

        1. ArrZarr Silver badge

          Re: AI

          I don't believe that the AI which won the series of Go matches was using all that much brute force due to the absurd exponent at which go moves use.

        2. LionelB Silver badge

          Re: AI

          @Mage

          I think you may be a bit behind the curve here - you are talking about what is sometimes known as GOFAI ("Good Old-Fashioned AI"). Current approaches to AI are pretty far from "human curated databases" - although there are no doubt still plenty of those around.

          The current buzz is "deep learning" systems, which really can be said to be capable of learning with minimal domain-specific pre-programming - see e.g. http://www.wired.co.uk/article/google-deepmind-atari, where the system learns to play (and play very well) Atari games purely from pixel-by-pixel input and score feedback from its own actions. Similarly, the AlphaGo system http://www.nature.com/news/google-ai-algorithm-masters-ancient-game-of-go-1.19234 which recently triumphed at Go against top-level human players does not "brute force" anything - it learns through playing.

          It is no surprise that Google has been at the forefront of this impetus. While the basic ideas behind deep learning have arguably been in place since the 1980s if not earlier, it is only recently that the huge computing resources are available, which are required to exploit the capabilities of the ANNs (artificial neural networks) that enable deep learning to solve non-trivial tasks. And current systems do not even begin to approach the scale, complexity and efficiency of animal brains.

          Now I'm certainly not saying that learning (deep or otherwise) is in itself sufficient to achieve anything approaching human- (or even insect-) level intelligence. Tellingly, it turns out to be much harder to design a robot that can wander around a novel environment without bumping into things than it is to design a system that can beat expert humans at Go. Intelligence as we recognise it (and I'm not restricting that to human intelligence) is a product of the evolution, over geological time scales, of agents interacting with complex environments. While the current buzz for deep learning is, I think, encouraging, my feelings are that AI will not advance significantly without taking on board further organisational principles from this evolutionary/interactive history. In particular, I believe "disembodied" human-engineered AI (as in GOFAI) is a dead-end, and that while deep learning may turn out to be a useful building-block for more sophisticated AI, it is almost certainly not the full package.

          By the way, there's an excellent new book out about the state of play in the field:

          AI: Its nature and future, by Margaret A. Boden, Oxford University Press https://global.oup.com/academic/product/ai-9780198777984 (Maggie Boden is probably the leading expert in the philosophy of AI, and an excellent writer.)

          1. Mage

            Re: AI

            " Current approaches to AI are pretty far from "human curated databases" - although there are no doubt still plenty of those around.

            The current buzz is "deep learning" systems, which really can be said to be capable of learning with minimal domain-specific pre-programming -"

            No it's all marketing. Just "buzz"

            We have ONLY made progress in marketing, hype and jargon.

            Bad SF.

            1. LionelB Silver badge

              Re: AI

              "The current buzz is "deep learning" systems, which really can be said to be capable of learning with minimal domain-specific pre-programming -"

              No it's all marketing. Just "buzz"

              Really? I think some of the achievments of deep learning are genuinely impressive (google around a bit). Maybe I'm just easily impressed,but some of the image anlaysis, game-learning, etc. would have been unthinkable just a couple of years ago. And it is not going away soon - it works.

              Not that I think deep learning is all that "deep" - it's pretty well-understood how it works. I think a more interesting debate is to what extent it puts the "I" in "AI".

  6. Anonymous Coward
    Anonymous Coward

    Don't call Me Shirley

    Apple struggles with the idea of intelligent life?

    FTFY

    1. allthecoolshortnamesweretaken

      Re: Don't call Me Shirley

      Well, as I sometimes struggle with the idea of intelligent life inside Cupertino, I guess it's a fair cop... (Don't mention iTunes. Don't mention iTunes. Don't mention aaiiee!)

      Anyway, IMO AI still has a long, long way to go to truely qualify for the "I" in "AI". When an AI system invents something like Chess, or Go, or even Ludo - then we're talking.

      1. LionelB Silver badge

        Re: Don't call Me Shirley

        Anyway, IMO AI still has a long, long way to go to truely qualify for the "I" in "AI".

        Agreed.

        When an AI system invents something like Chess, or Go, or even Ludo - then we're talking.

        You're setting the bar way too low. I reckon that's already doable. The really hard stuff is AI that can interact, learn, survive and solve open-ended problems in the real world. Just like animals do (even quite dim ones).

  7. hellwig

    Apple the Also Ran

    As people have pointed out in Apple vs. Google arguments, Apple has a mediocre global market share at best. Just as they've always had. Sure, they have momentary spurts, the original Macintosh, iMac, iPod, iPhone, etc... but they always settle back down into a second tier player.

    This is directly because of their closed-off mindset. Sure, when the product is new, and the patents still in effect, business is booming. But eventually your competitors replicate your success (Steve Jobs would cry that they stole it), improve upon the technology, and outpace you.

    Even Microsoft is starting to embrace open-source. For a company who's desktop OS is based on open-source software, you'd think they'd also be more open, but I guess it's OK for them to use the work of others, but not the other way around.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like