Apple is struggling to become an AI powerhouse
washingtonpost.com
washingtonpost.com
Once you have your model as a CoreML file, it is stupidly simple to incorporate it into an app (the live demo with the flower model was very impressive), and Xcode will convert it to a machine-optimized version.
I was skeptical when I saw the announcement, but honestly it seems like a game changer -- to be able to drag-and-drop models that other people have trained into an app and use them with virtually zero boilerplate is just great.
Video here: https://developer.apple.com/videos/play/wwdc2017/703/
"Alexa, play me that song I like. And while you're at it, order me a case of my favorite beer. Thanks."
Here's a fairly innocent scenario: if Amazon Echo can learn everything about you, they may discover that you're crazy about Paris and you have good income. Then they can sell this info to every travel website who'll put you into a pricing bucket which would guarantee that you'll overpay for your next trip.
But generally speaking, we take advantage of information asymmetry all the time: buying real estate, getting a new job, planning a vacation, etc. Sometimes you have the leverage, sometimes you don't. With massive information gathering by a handful of big players you'll most likely end up on the losing side in the future, especially when dealing with businesses/governments who can afford to buy a profile on you before engaging. Imagine if your religious beliefs and either you voted for Trump were available to anyone via an API call which takes your face image as the input and costs $0.02 to make.
Now you may say that you can't trust corporations with this responsibility, but what if there WAS something we could do to enforce that responsibility? How much control I have over my information should be left up to me, and it would be great if an entity I trust could enforce that. As of this moment, we don't have that option. However, it doesn't mean that such a future is impossible. If we argue about it enough, maybe the technology and the courts can find a way to make it happen.
Data is being gleamed about you even if you aren't using their products, and it's being used by companies you haven't authorized all the time.
If you tell me that you like bananas, and I bring you bananas, I'm not being unethical. The problem with online services is that you're baring your soul to the eye of sauron without realizing what it means. The only way to win is not to play.
This is set to activate in a year, and it is a huge problem for software used in the EU because almost nobody is ready for it.
https://en.m.wikipedia.org/wiki/General_Data_Protection_Regu...
But you can have reasonable control of it and you can't with the latter.
Sure we do. But this is generally regarded as a bad thing and indicative of a failure in our commercial system to engender fair outcomes. Informational asymmetries aren't things we should be actively involved in exacerbating.
Of course we exacerbate information asymmetries. That's what we do. That's why the neanderthals are in our museums, instead of the other way around.
Saying we shouldn't is like saying birds shouldn't fly, because it's not fair to the fish.
If not, how do you know it is in fact the right thing?
If you prescribe and enforce your ethics, I think you're wandering close to another definition of politics.
My ethics? I never said mine.
And yes, there are inherent dangers (like tyranny of the majority), but there are trivial ethical issues where there is no good solution for allowing the [or any] minority to self-determine. (Like rules regarding sex with children, just to throw in the classic think of the children. But of course the age of consent is not a magical line somehow creating an overnight transformative event for every resident of that state.)
But the point remains that most people are voluntarily, enthusiastically, giving up personal data in exchange for features that they find valuable.
It's hard to see that as a privacy issue.
Being part of the world involves being known. In the proverbial "olde days," when people lived in much smaller communities, everyone knew everyone's business.
Government spying on its own citizens is more problematic. But again, being a citizen, by definition, means aspects of an individual's life are known, and are public data.
This notion of "privacy" is a special, legal, construct. It's meaning in everyday language is provisional.
What is more problematic, IMO, is the excessive use of classification by the Federal Government. This practice is vulnerable to abuse, as it enables legally enforced secrecy to cover up misconduct.
https://techcrunch.com/2015/06/06/the-online-privacy-lie-is-...
No, that's demonstrably false. Most people have no idea how exactly they are tracked and what information is being collected about them. How would they? It's not always as simple as "Uhh, so I guess if I write an email in Gmail Google can see my email?!" And even then, I think the majority don't actually think that Google does that. "I mean, email is private is it not?!" - is what some people likely think.
Do you actually think that most people know that Facebook and Google track them across the web and know most of the sites they visit when they aren't even on Google or Facebook's websites, through Google Analytics code and the Like button (without having to press on it)?
How long before the technology and bandwidth becomes advanced enough that it's actually possible to capture all conversations from mics 24/7 of the people around a device, and store it for analysis in the cloud? Do you think people expect that to happen right now, or are "enthusiastic" about it? No, they don't and aren't.
In fact some call this idea a "conspiracy theory." But ten years from now we'll probably discover Facebook or Amazon has already been doing it for a few years, and a few more years later the privacy invasion apoligists will start saying "well yeah, of course they listen to your microphone, that's why it's there! To analyze your context and give you even more targeted products and ads!"
Companies first abuse new ways to track people (I guess they would call that "innovation"), and maybe 10 years later, regulators catch-up to it and force them to pull-back their tracking a little bit, after the information has long been stored and used in their databases.
Even in a competitive market place they can collectively decide to overcharge the same customers.
Economically it would be the market clearing price so would be beneficial to the market generally. Sounds like a great thing.
In fact, this would lead to widespread pricing discrimination [2] based on what they learned (or bought) about you. Imagine seeing 20% markup on virtually everything available to you online, because every retailer can see your higher-than-average income before their page loads in your browser.
Personalized content (which everyone seems to love) is just an inch away from personalized pricing.
[1] https://en.wikipedia.org/wiki/Efficient-market_hypothesis
It's not perfectly efficient because there isn't perfect information, however it's more efficient than it otherwise would be.
More information in the system = greater efficiency.
More information does not always increase market efficiency. In fact, it can have just the opposite effect when there is information asymmetry.
https://en.wikipedia.org/wiki/Information_asymmetry
http://www.economicsonline.co.uk/Market_failures/Information...
Or let's go back to the Paris example. If everyone is paying the most they can afford, that leaves less money on the table for restaurants, cheese and wine when actually in Paris. It's 'great' if you only look at the market dynamics of flights and ignore the economic/social benefits of cheap travel.
The notion of a "public good" also isn't about how things should be treated, but more about how things necessarily are.
It sort of seems you're making an argument unrelated to economics here, even though it's using words also used in the jargon of economists.
If one is to believe that healthcare products should be subject to market forces, then their prices should be variable in the same way other products are. In which case the answer to the question is, yes someone with a "niche" health problem would pay a substantially high price.
If however one believes that healthcare should not be subject to market forces, that is, it should be available to all equally (non-excludable) and cannot incur asymmetric costs between users (non-rival) then the answer is no, it would not be good.
This is a fine opinion to have, but you do yourself no favors by misusing jargon like this.
This says nothing of the possibility that one party has a lot of information, and the other party has relatively little information, and the guy with very little information is getting screwed.
Genuinely wondering. My economics is a bit rusty so not sure if what I said is remotely valid
The consumer surplus is zero, but the producer surplus is the maximum possible value.
In other words, the monopolistic supplier would otherwise charge a higher price for everyone if it is unable to price discriminate. If price discrimination is successful, more consumers can afford the product.
Also, while we trade our privacy to Google (or Amazon, or whomever) in exchange for customization and convenience (a social contract I'm generally happy to sign off on), those companies have even more incentive to keep our data safe.
Google is one of the most valuable companies in the world precisely because it, and only it, has the AdSense knowledge (and whatever other knowledge Google collects about me) to target me.
Insurance companies go to Google and say "show our ads to people Googling insurance companies" – that's how Google makes money. It's not as though Google says "here you go State Farm, here's everyone who's been looking up car insurance." It's business model is based on proprietary customer knowledge. It can't give away this data; it's incentivized to limit it to its own ad-targeting tech.
Are there still problems with this model? Sure. If the government decides to subpoena Google on me, they'll turn over my Gmail. But is it a hell of a lot easier getting access to Google services (e.g. Google Maps knowing where I generally go and what the traffic is like) versus using, say, Duck Duck Go on a VPN (let alone Tor or Tails)? For me, and I'd assume most people, yes.
EDIT: I would also point out that we've long been facing the privacy vs. convenience issue. It used to be that merely signing up for a landline meant getting your phone # listed in the White Pages. Paying utility bills makes your name and home address a matter of public record (unless you choose to shield them via owning and paying through a corporate shell). Ditto real estate transactions involving your name/address. All public records, unless you choose to hire attorneys to setup shell corps for the sake of privacy. Not so expensive to do this now in the age of LegalZoom, etc., but this used to cost quite the pretty penny.
This debate is nothing new; it's merely evolving.
Then why aren't they doing it, and why aren't they informing us when breaches happen?
I absolutely think providing that data should be voluntary. If you want to send it to them, go ahead! But in many ways it's not. I can't remove my data from those companies, nor can I control what they do with it, and that's a serious problem.
There have been countless incidents (and these will only increase in frequency) where people's sensitive information have been stolen and used for blackmail and identity fraud. There is also the increasing use of private data by governments for example in deciding on visa entry or immigration cases. The use cases for criminals and governments are only going to increase in scope and sophistication and will be applied not just to future data collected but current data.
These are all legitimate situations which are completely unprecedented and only possible because of the increased data collection policies of site like Google or Facebook.
Would Facebook agree to make all of their employee web searches public? Would Google? How about all phone traffic? Emails?
Thought experiment: imagine a world where everyone can see what everyone else is doing all of the time.
Assume absolutely no exceptions or restrictions. You can eavesdrop on anyone in the world. Anyone can eavesdrop on you.
How many "I am fine with no privacy" advocates would be happy with this?
It's an extreme thought experiment to highlight how asymmetric the current model is. In the current model privacy is becoming a privilege that is available more and more selectively.
To eliminate the privilege, you either need user controls and permissions for specific profitable use cases, or you need full openness - which I think most people would find terrifying, for all kinds of reasons.
I would have far more respect for no privacy advocates if they made public a daily ISO of the contents of their computer.
Would they really have the same position once their identity had been stolen, their credit cards maxed out and every thing they have said taken out of context and made available to their friends, family, boss and the TSA.
Slightly tangential to this topic, your comment reminded me of this short talk titled "Your smartphone is a civil rights issue" by Christopher Soghoian. [1] It truly is a great privilege to be able to control one's privacy in today's world (to whatever extent it is possible).
[1]: https://www.ted.com/talks/christopher_soghoian_your_smartpho...
> ...
> Are there still problems with this model? Sure. If the government decides to subpoena Google on me, they'll turn over my Gmail.
It seems to me like you've read about the Snowden revelations but don't see any issues with warrantless tapping and mass surveillance. As I said in another comment, privacy is not just about you or me. It's about all humans and the rights that we have granted ourselves in many countries around the world.
Because you haven't defined privacy: The ability to control what is publicly known about oneself. I'd also add the ability to assess what is known about oneself.
You also ignore identity vis a vis privacy. A completely transparent world (a la Brin) has made identity theft a booming industry.
Lastly, everyone ignores property rights. I am my data, my data is me (see identity above). At the very least, if someone is making a buck at my expense, I want my cut.
And really, Alexa is not really that much more powerful by knowing what your favorite beer is; you already know that information. Until you consider Alexa to be an actual friend, any personal details about you that it keeps are marginally useless if it cannot function autonomously - by ordering your favorite beer without even asking. This is not an 'intelligent' behavior, it's just maintaining a list of preferences. To be honest, it sounds like you are conflating voice commands with AI proper. You can certainly have one without the other, and the former seems to be the thing you're valuing here.
I'm all for the concept of powerful AIs that know everything about me - so long as they know me as an actual person to care about rather than merely as a consumer to squeeze money out of. Which means understanding my boundaries and knowing what I consider offensive and acceptable behaviors.
What we have today are AIs that only learn about you what they find useful to their owners, not what I find useful for them to know.
We should not have to sacrifice privacy for usability. There's a reasonable tradeoff and I believe Apple is doing the right approach.
1. https://www.google.com/amp/s/www.recode.net/platform/amp/201...
The problem is that nerds are allergic to advertising because they are allergic to "being sold to" because they (well, we) like to think we're above such petty foolishness.
As such we are not the customer. Therefore nobody cares what we think.
And privacy will go away.
The problem with this is, what happens if privacy goes away too much and somebody abuses it? What happens when a malevolent dictator takes over and oh hey look at all these devices so conveniently set up to spy on all the people. How wonderful. What if a benevolent dictator takes over but their idea of good doesn't align with your idea of good? What if somebody decides that oh hey, drunkness is a problem, people do terrible things when they're drunk, let's round up everyone who buys more than 3 sixpacks of beer per week and put them in a holding facility and fix them.
You know all those futuristic utopia movies that turn out to be a distopia under the hood? We're living in one of those. The question you're answering is whether you're inside the utopia or outside. Those of us who embrace the lack of privacy and being sold to, we're on the inside. Those like Stallman are outside.
It would be ludicrous to claim that I'm above being sold to. I'm not even sure what that would mean.
This, I believe, opens the door for a company like Apple or otherwise to create a machine learning company with security at the forefront -- in essence, what if these useful machine learning models you use everyday were able to be locked down, so the personal data used by the model was only viewable by you, and NOT an employee of the machine learning company who produces said model?
(EDIT: A few word choice changes)
America and a few others are arguably outliers in the western world for failing to recognize privacy, and the right to private family life, as a fundamental right in some form of constitutional legislation.
As a highly pro-privacy person, I find it ridiculous when people build up arguments like "I guess I don't have a (sic) CIA- level job that most of these commenters have...". Seriously, privacy is not just about you or me! It's about every human in this world, and the balance of power between layperson/citizen and state/corporation. Without protections like privacy (which is one among many), we'd never enjoy all the freedoms we take for granted every single moment. I'd strongly recommend reading this short and succinct article titled "Privacy protects the bothersome" [1] by Martin Fowler.
[1]: https://www.martinfowler.com/articles/bothersome-privacy.htm...
Think about the massive data gathering we could do in the medical field for example, to test the effectiveness of drugs and treatments if you could wear a "smart" bracelet that put its data in the cloud - how many more diseases could we cure?
Not that I'd feel comfortable sharing my back account # with you...
The same story has played out dozens of times in history. Stalin, Pol Pot--tyrants kill for irrational reasons.
Store the data locally?
Wow, now I'll never have to leave my chair!
Trust is vital in devices like this.
[1] I'm becoming less of a fan, as MacOS becomes "cloud aware" and less of a Unix workstation.
Apps like Lyft, Yelp and Uber already know where you are so you can't use those.
You're left with maps and browser. People tend to use google maps and Waze on iPhone anyway so that point is gone as well. If they use Apple maps, I am not convinced that apple doesn't store location history.
So then you're left with a browser. Android lets Firefox run its own browser engine and you can install all the privacy extensions there.
On the other hand, with Android it's possible to take privacy a lot more seriously. Lineage OS lets you have much more fine-grained controls over all the apps and you can forego the whole google services privacy menace.
I'm not worried about SMS because everyone and their mother can see it now.
I don't think apps that didn't require contacts during installation could access them.
To me all Apple products are a usability nightmare. I am forced to do simple things the way Apple wants me to do them instead of the way I want to do them.
To me all Android products are a usability nightmare. I am forced to do simple things the way Google wants me to do them instead of the way I want to do them.
That line of reasoning just doesn't scale.
Like only learning to type on a Dvorak keyboard, doing things "your way" can have consequences. Rampant, arbitrary customisation of interfaces is, in my estimation, more often than not an exercise in painting yourself into a corner.
I gave up on interface consistency though because every app now has its own interface and vendors like Samsung have their own for each product.
I miss the old days of windows apps with consistent menus and obvious ways to use the programs. Now it's like playing some sort of text RPG where you have to try everything to figure out where some setting is.
The problem I have with apple is that things are not there because "art". Google is also very much guilty of it. But at least they don't copy all the apple problems and usually expose more of the internal workings of things than apple does.
Only when you use them.
You can either not install them at all (problem solved).
or only allow them to "know your location" when you explicitly launch them and only for the duration of you actively using them on screen (problem lessened).
> in iOS 11, users will be able to choose the “when in use” option at all times [when asked for permission to use their location]
Closed or in the background?
I was under the impression (in iOS 10 and earlier), that "Always" meant anytime the app was open, and "While in use" is only when the app is in the foreground. So if an app has "Always" on for its Location Services, it will stop tracking when I swipe up and quit it.
If I've been incorrect this whole time, please pass the tinfoil.
You could turn off the gps but that's too coarse: you might want it for some apps.
They take the ads, and distribute them to their users based on user taste that's their secret sauce. Giving user data to advertisers would be giving away their secrets I think.
They sell targeted ads, they do not sell customers' data to advertisers. i.e. Advertisers get to advertise on Google's platforms, but they do not get access to Customer data.
I gather there are some questions around the strength of its app sandboxing as well, but I can't speak to that at all. Perhaps someone better informed will do so.
I'm an Android dev and to me the biggest problem with Android from every aspect (development, security, monetization) is how poorly new versions of Android are supported on devices.
Apple's dig at Android 7 usage vs iOS 10 usage was spot on.
This is not true
1) iOS has had lazy permissions since the very early days e.g. when an app wants to use this microphone then ask the user right then instead of when the app is installed. This feature is still not implemented on all Android devices nor is it as seamless nor is it mandatory.
2) Apple is providing more and more frameworks e.g. CoreML which run only on the device. This means that developers don't have to run their own server side apps to process data.
How is it becoming less of a Unix workstation? It can still do all the Unixy stuff it could before.
- They're paying much less attention to the "Unixy stuff" than they used to. Old versions of common things take forever to be updated. Compatibility has always been shit, but it is getting worse. SIP means a number of things I tend to modify, can't be, without hassle[1].
- With a workstation, I expect to be able tailor it specifically to my work requirements. This is becoming increasingly difficult at the proliferation of "cloud" daemons increases and they lock down the system - the documentation for most of the daemons ranges from "sucks" to "nonexistent", so it isn't clear if any given one needed for stable operation or if it yet another thing related to trying to sell me music or some such[2]. Reverting the machine to a state where "root" means "root" is rather difficult (See SIP).
Apple's focus is on chasing the flavor-of-the-month social features and keeping up with the GOOGs. Which makes sense - the set of people buying that stuff is much larger than the workstation market. But it means they don't care so much about it, and make choices like letting the core OS get as buggy and unstable as it has become.
It is mostly fine for me. I built a nice king-hell PC running Debian that I'm very happy with. I didn't think I would, but after my last MBP died unexpectedly, I ended up buying a bottom-rung used Air because I'm hopelessly dependant on OmniFocus and a couple other OS X only productivity tools, and I sync my phone local-only. So it tends to live in the dining room, where I work on my schedule and read mail via ssh over breakfast.
So yeah, Apple still has a place for me, but it is peripheral. I spent mid-4 figures on building a PC this year, and mid-3 on that MBA. And I'm thinking about what I really need out of OmniFocus, and how much of that I can build myself.
[1] This is a good thing for many people, but I expect full control of my machines, and I do occasionally sketchy things like patching Safari. The problem with SIP is that I only need full control sporadically, and not for long, but it is a serious PITA to disable/enable. So it just stays disabled.
[2] One example of many is `parsecd`. No man page. Googling about shows that a lot of people think it to be related to location-based lookups for Siri data. Despite the fact that Siri is disabled on my box, it ran and asked for network access all the time, until I disabled it. Now, WTF is 'LaterAgent', and why is it in my process table?
The very same omnifocus problem solution:
* shorterm for migration: omnifocus on ipad
* longterm: trello - it will be slightly different, and you have to come with your own GTD system for it, BUT surprisingly the differences allow me to get fresh and new approach to GTD. it's worth.
as a bonus: you can check taskwarrior instead.
I tried every other software, but it sucked, meaning got some problems. IMO only trello or taskwarrior _might_ (still in the process) do.
LaterAgent is used for system updates. Specifically, it appears to be what handles reminding you again later when the system prompts you to install an update and you select the option to remind you again later.
So, I agree with your sentiment, but I do not agree with your assessment that Apple's stance on privacy is anything even remotely close to a significant factor in attracting customers.
Prioritizing privacy and security are one of those things that fit as being fussy about things working well, but not wanting to put the time into configuring them to be so. If that crowd likes your stuff and recommends them, then the less technically savvy folks will pick up on the preferences of that crowd and it build's Apple's brand as making premium products.
Maybe it depends on who you're dealing with, but in my experience teenagers these days are a lot savvier about these things than the typical people in my (30+) cohort are. (They kind of have to be, no? They actually have parents and teachers snooping on them in their real lives so it's a more concrete concern for them.)
This is one of the reasons kids gravitate to Snapchat more than Facebook or Instagram after all: posts on there expire. They have a 'right to be forgotten' by default that they don't get with the other services. So I'd say they still care about keeping personal information personal, it's just motivated by different concerns than it is for the 30+ demographic since they don't have as much sensitive health and financial information floating out there.
A major reason for me as to why I continue to stick with Apple is exactly that, the privacy and security focus they have.
There are a lot of people living under potentially abusive dictatorships.
I hope they publish more in that direction.
I feel the same as you do, but are we sure that the mainstream public actually WANTS privacy, especially when compared to the utility companies are providing by taking it away?
The lack of privacy has to feel tangible in consequence for consumers to care, IMO. Right now it doesn't.
Do you even know what you are talking about? On an Android phone you could install other App stores and side load apps. You can't do any of this on iOS.
I think it's a very reasonable position.
I know that this is the most common approach to the situation but the article -as many other articles that have argued this before- provides no proof other than analysts speaking in broad strokes.
I use Siri, Alexa and Assistant several times a week if not everyday and I'd hesitate a lot before saying that Siri is behind. Alexa is more responsive but it's domain is also more narrow. Siri does a pretty good job when I dictate a message or want to set a timer. It's limited, sure, but so are the other assistants I use. I might be the odd case, but speech recognition of Siri vs Assitant is pretty comparable in my experience.
Moving onto services. Google Photos is outstanding, yes. But I fail to see any other example that proves the big advantage other players have over Apple in the ML or AI field.
Am I missing something? I feel like a lot of these articles argue their view from the experience they got from Siri 5 years ago.
If I set it to, say, Portuguese or French, its capabilities are greatly diminished. Even simple tasks as "play a song by Pearl Jam" are impossible because Siri will try to find some Portuguese phrase that sounds remotely like "pearl jam". If I set it to English, then I can't use it to send a text to my wife (in Portuguese) or get directions to local streets (named in French). Also people's names.
Google Assistant at least allows me to tell it that I speak those other languages so if I speak in any of them, it will understand as such. I can say "play Pearl Jam" in English, "envie uma mensagem para minha esposa" in Portuguese or "trouve-moi le boulevard laurier" in French and Google Assistant will get what language I'm using. It's a big help. Not perfect, but at least it's usable.
EDITED: grammar
Siri is flat out dumb and frustrating to use comparatively in terms of only understanding 85 percent of my queries vs. Google Home understanding 97 percent.
It seems to me Apple has not taken all their hordes of cash and used it to up the AI game. Looks like they are solely focused on the present while competitors are building the future in which they foolishly try to catch up with.
Like last weeks WWDC they debuted the AR kit but did not bother to create an amazing/innovate AR app to showcase and get the masses excited for AR and their possible upcoming AR gear.
Why would they announce an app for supposed upcoming AR product before announcing the actual product? You announce the developer kit 6 months before that happens, get them working on apps, and then when you announce your AR product there are lots of apps, not just an in-house one, ready to go. Even if an AR specific product doesn't launch soon they've brought a good AR API to devs a few months before iOS 11 - and when it does (if developers take advantage of the API's) millions of people will be using AR daily on their iOS devices. Right now I know absolutely nobody using AR on any platform. That's going to change very quickly and an in-house app won't make a bit of difference.
It's a developer conference not a product launch, so of course it focused on the code rather than "getting the masses excited". The point is that developers can now spend the next few months creating those AR applications to impress people before iOS 11 with ARKit is launched to the public.
They could have built an AR feature into the camera app to show they are not just followers of all other companies who innovate they still are innovators. Further to showcase and get the masses excited about AR.
You're describing an interface problem. Google's AI is extremely good at recognizing speech in comparison to Siri but that's a wholly different thing to deep understanding.
Siri is a more advanced AI in many ways due to it's understanding of intents. Google's is much more command based (under the hood). However Apple's speech recognition is letting it down here. Intents are a much more sophisticated way to interact with an AI than Google's command based structure but they are correspondingly more complex to integrate and you realistically need to infer things more from the user (which Apple needs to get better at).
I don't disagree that Apple needs to step up their game with how people interact with Siri, but it's a perceptual issue with the interface not with the underlying AI.
(source - have discussed these exact issues with one of the founders of Nuance).
See http://www.businessinsider.com/siri-vs-google-assistant-cort...
Siri is dead last. Surprisingly, Cortana did quite well (I didn't realize Microsoft was catching up in this space).
Those responses require a lot more than just voice recognition - they require an understanding of context and "intent".
> A recent study […] asked the major voice assistants 5,000 general knowledge questions
General knowledge questions is just one of the many things you can use voice assistants for. It's well known that Siri is the worst at general knowledge questions, but that doesn't mean Siri is the worst digital assistant.
This is where a good UI can obscure the underlying thing though. Google is trying to find intent via the text of speech and that's one way to go about it for sure. It's totally valid. It's more limited, but it'll work great if there is a command that it can match it to.
But there has to be an underlying single command that is "turn volume up"
With an intent based system you can chain complex intents together and the result is the behavior of those intents interaction.
It's... ugh, I'm explaining this poorly.
Think of intents as building blocks to create something from whereas commands are just specific things. I wish I could explain this better.
Having said that, I don't know how you can say that Siris backend is much better from an intent perspective without being able to leverage it properly because of the shortcomings related to the UI. From what I've seen, it doesn't even handle context well. Now it sounds like Siri will be used to do proactive things which is certainly new and different from Google Assistant. Yet, I suspect that logic is just being branded as Siri because it is a push to label Siri as your intelligent assistant as opposed to the weird robot thing you can use to check the weather
Basically if Apple actually manages to get the voice recognition part down (so it parses intent/structure rather than just pure words) it'll be very powerful.
Right now Apple has a frontend that is more suitable for a command-based system and Google has a frontend that is better for am intent based system.
It's easier to improve the frontend than create a better backend though. So that's the competitive trade-off that Google has made.
I actually think it was smart of Google, and quite un-Apple like of Apple to do it the way they did. Usually they are very focused on the interface as well.
Then again, at the time, Siri was quite good at that. It's just Google focused on it (clearly).
It's a massive stretch to call voice recognition AI, and the recognition leap that happened recently is not so much better from what was built into the early OS X versions. Apple's largest AI project that I've seen so far has been music recommendations on apple music.
My big problem with Siri is that I have to time when I start talking. Also, her voice recognition is awful in noisy settings.
I have an iPhone 7
What do you mean? You can just say "Hey Siri, blah blah blah" and Siri will pick up the "blah blah blah" just fine.
If you're adding an artificial delay after "Hey Siri", waiting for her to respond before continuing, then you're just making it harder. Just speak conversationally.
I actually find this frustrating and I'm surprised Apple went this way given their usual commitment to humanistic interaction design. When I'm addressing someone in real life, I call out their names and wait for acknowledgement that they're listening before rattling off questions at them.
The fact that Siri doesn't acknowledge it's listening or even give you a visual indicator that it's attention has been caught (like a person does by making eye contact) makes interaction with it feel strange and unnatural.
It's especially weird when you're in a room with multiple "Hey Siri" capable devices. If I call out "Hey Siri" I'd like to know whether I should be speaking at my iPhone, my iPad, my girlfriends iPhone, or a hypothetical HomePod. Which ones should I direct my voice at and expect a response from? All? None? Some? Can I tell my girlfriend's iPhone to go away because I'm talking to my phone right now? I'd like to. . .
Yes it does. It makes a very distinct sound when it starts listening, and the screen itself transitions to the Siri screen and shows the animating waveform.
Which is a big problem.
I wonder if implementing the feature of calling 3rd party assistants from the flagship assistant would pay off strategically.
On one hand, you appeal to people who like other assistants better, or like different assistants for different things. On the other hand, you give up on forcing people to get used to your assistant.
Also, could an isolationist provider stop other assistants from calling their one? Technically? Legally?
It'll never happen, but I can definitely see some utility in building this in. For example: "Hey Siri, have Alexa order some paper towels from Amazon." I can see Amazon allowing this since they're going for a more generic approach. But I don't think Google, Apple, MS, or Samsung will.
I figured that was more of a hardware problem?
Sometimes I'll be watching a baseball game in my living room with the volume pretty dang loud. The sound of the crowd in the background is almost like a static noise, and then you have the yellers, as well as the commentators speaking clearly, distinctly, and loudly. Then I use Siri to set a timer for something I'm cooking, or to "Call X person on speaker." and it picks up my voice without any hesitation or anything, and gets the words exactly right.
Definitely accurate.
Deepmind / AlphaGo, Watson, and other like them are about the best thing we have nowadays when it comes to AI. Let's not even talk about True AI, which if you take any of the definitions can only become reality with so much data we're only brushing on it actually. Saying True AI won't need as much data is failing to understand how much "data" it takes to make a human brain function, and while that is a whole other can of worm, one thing is certain it's a shit load accumulated over time.
> If and when true AI happens, there's a good chance Apple will be leading that.
Apple is not even in the same arena when it comes to AI. And they shouldn't be, with the cash piles they have it's better for them to wait for somebody else to make the breakthrough and pull a Siri, buy a company that does what you want and integrate it how is needed for you product / consumers.
Also don't confuse AI/True AI with clever algorithm.
>Deepmind / AlphaGo, Watson, and other like them are about the best thing we have nowadays when it comes to AI. Let's not even talk about True AI, which if you take any of the definitions can only become reality with so much data we're only brushing on it actually. Saying True AI won't need as much data is failing to understand how much "data" it takes to make a human brain function, and while that is a whole other can of worm, one thing is certain it's a shit load accumulated over time.
I think you and I are saying the same thing. Namely, intelligence is an accumulation of years of experience, information & data (though human sensory system). But, once intelligence is established (say after 21 yrs of training in case of humans), learning is rather easy (i.e. one shot learning is sufficient to accomplish new, mundane tasks). You don't need to know someone's whole browsing history and emails to make personalized recommendations or to give intelligent answers to queries. The more intelligent the algorithm, presumably, the less personalized data it needs. So when it comes to true AI, you don't need to know as much about the user to help. This is one way to rationalize Apple's approach.
It may not be decisive, but it's worth remembering that there are benefits (other than privacy) to their approach.
Siri: Here's what I found on the web for "What cheese goes well with fruit..."
Me: Hey Google, what cheese goes well with fruit?
Google: "Edam"
Siri: Here's what I found on the web for "is obama planning a coo"
Google: According to details exposed in Western Center for Journalism’s exclusive video, not only could Obama be in bed with communist Chinese, but Obama may in fact be planning a communist coup d’etat at the end of his term in 2016!
(Definitive answering style is not always better...)
More:
https://www.washingtonpost.com/news/the-intersect/wp/2017/03...
http://www.businessinsider.com/google-home-claims-obama-plan...
Would you have felt satisfied if you asked, "What wine goes well with dinner?" and heard "Zinfandel" in response?
An actually useful AI would ask you what kind of fruit(s) you want to pair it with. Not throw out an answer at random.
I've had two failures in recent memory where Siri recognized the queries correctly, but completely misinterpreted them.
"What's <time> Eastern Time in Pacific Time?" => time in Pacific, MO.
"Increase volume by 10%" => lowered volume to almost 0.
The latter failure took me a while to figure out: Siri had interpreted my query as set volume to 10 (out of 255?). Indeed, "increase / decrease volume by X" had the same behavior.
Also, asking Siri to play a particular song whose title's in Japanese might as well be asking it to play a random song. Setting its language to Japanese helps a little, but then of course I have the inverse problem. I wish I could switch Siri's language settings verbally, or be able to say "In <language X>, <query in language X>."
Google's also been known to produce completely incorrect "authoritative" answers.
"Edam" might be authoritative, but it is not correct. The problem isn't simply that the answer is fundamentally subjective, the bigger issue is that there's not enough contextual information in the question to even arrive at an answer that one could consider subjectively correct.
I replied elsewhere, but it's like answering the question "What wine would pair well with dinner?" with "Zinfandel". Without having even a vague idea of what's on the menu, you're just picking a wine at random. Maybe it's correct, maybe it isn't. But being authoritative when there's not enough information to actually speak with authority is not the right answer here.
The thing is Siri won't give me an answer either way.
Whether or not Siri or Google Assistant correctly and authoritatively answers more (or more relevant) questions is a completely different topic than what's being discussed here.
"Hey Google, is this batshit conspiracy theory I heard about true?"
Google: "Yeah, sounds right"
I mean, to be fair, isn't that every company? Google pulls the same sort of nonsense.
Your info is outdated. At a conference last December, Apple's director of AI research announced that their AI researchers will be free to publish their findings.
https://9to5mac.com/2016/12/06/apple-ai-researchers-can-publ...
Google knows all about me and its assistant is, usually, great. Amazon has troves of data on what I buy, and I get to yell at Alexa to order more TP as soon as I see we're on the last roll.
Apple knows much less about me and, while I'm still an Apple fan and am tied to iPhones/Macs thanks to iMessage, Siri stinks as a result.
If voice assistants based on machine learning (specifically, personalized voice assistants) are the next big thing, Apple's privacy ethos will separate it from its major tech competitors – either in a great way, or a very negative way.
It will be interesting to see if Apple holds its ground on privacy with the increase of AI/ML driven features and products. Coupled with Apple's closed culture which discourages open research (although they have been improving this), their concern for privacy could put them well behind other companies in this space. Depending on how you look at privacy vs. product, this could be a good thing or bad thing.
I think it makes Apple slower at making their assistant good at parsing what you're saying and returning an answer, but that's a problem that benefits from crunching reams of data in general, not so much knowing everything about you personally.
For one thing, a virtual assistant that works even when I don't have an active internet connection seems like a perk in itself no? The Siri approach is closer to being there than the Alexa/Google approach.
While some may prefer a knowledge about you or you via people like you, as in, "Please recommend me a great movie", Siri is currently not in a stage where she may tell you stuff like "How do I mix a White Russian?" Google Assistant give me 6 steps with a photo and a follow up question "What about Screwdriver?" Siri gets utterly wrecked in the knowledge based questions.
I think the main problem is not privacy-related, but knowledge base-related. Google is building upon a fricking huge search engine via a knowledge graph and the sky is the limit for how well an AI can do. Apple is building on what, a shut down Ping social network, Apple Music listening habits, Wolfram Alpha hopes if all else fails, and a sparse Bing Search API if that failed too?
Not to mention that Apple positioned itself as a brand for people who "Think Different." I don't think people who got taken in by that messaging would be attracted to the prospect of services that can more efficiently pigeonhole them.
I periodically delete all my history from google servers, thanks to their privacy tool, and google gets just a dumb for me as siri.
If you watched the HomePod reveal you'd know that it sends your speech to the cloud for understanding. Which seems like a pretty clear admission that this can't all be done on-device easily.
I believe the restrictions imposed by keyword-spotting have more to do with the always-listening and/or power-efficient nature of the task, rather than the restrictions on the legibility of offline speech recognition.
More likely AI/ML is another trip down buzzword lane. It can hang out with IoT, VR, Big Data, and hell, even containers and microservices.
It feels like everyone is casting about for that next revolutionary technological innovation, but maybe we just need to be at a plateau for awhile.
Of course I may be completely misrepresenting this. I encourage you to go look up the relevant info from WWDC 2016 (I believe that's where they talked about it).
That said, none of those features interest me much. And I'm not sure how many average consumers they do interest. It's more of a 'oh, that's cool I guess'. I prefer my iPhone 7 still.
Anecdotally speaking, anyways.
For example, when you meet someone, they don't need to know everything about your life to be useful. The same is the case for building internet AIs. Using more pre-trained models. Asking the right questions and listening to the responses.
Sure, it's easier if you have access and are willing to use user data to build those predictive models. But there is some real value in the world for companies who respect user privacy AND build predictive models.
The author of the article clearly made a mistake saying "deep learning" is the same as "unsupervised learning".
If Apple invested in genuinely new, creative AI technologies that matched human level speech recognition and other tasks then they could preserve their purported emphasis on privacy. They would not need to collect huge amounts of personal data on most customers, unlike Google or Amazon.
But related to
> Regarding privacy, when human beings meet a new person, they do not need tons of data about that person to understand their spoken language.
I think that we understand context and symbols and syntax because we used it/learned it before.
So when I meet a new person I use previous cultural/language learnings to understand the new person. Not just dictionary.
And even if I met a person who speaks a new language, I might understand after a period the words but still I use logic and symbols to get that when that person points to ice and says something that something means ice. And this I think is possible because of some previous knowledge on how to interpret the interaction that I discovered before about this context.
I think it is the same with Assistants: there is an advantage in being able to access and process large amount of communication so that you can infer meaning or predictions.
I will concede I get moderate gag value out of Google automatically creating animated gifs from a string of photographs.
As usual, for a business trip I took this week, I went all-in with Google (location, my email forwarded to my gmail account, used Google Travell app, etc.). Very convenient for keeping my schedule, travel arrangements, etc. in order.
That said, when I returned home last night, as usual after a trip, I stopped forwarding my email to gmail, uninstalled most Google apps from my iPhone, and turned off location.
When I am at home (most of the time) I like using duck-duck-go, and just Apple's software.
EDIT: I have been working in the field of AI and machine learning for about 30 years. I think that Apple will find a sweet spot between good privacy and making Siri into a useful digital assistant.
Most creepy forms of surveillance are used for ads (but I'm not sure if it even works, I always see completely irrelevant ads) and Apple is not advertising company.
Am I missing something? Has there been any announcement that third party developers could add skills or actions across all Siri instances via the cloud?
Two) Apple is never really bleeding edge with it's products, it waits to see what the industry is doing and then subsequently tries to take something and make it better, something tough to do with AI/ML. You can't really just take someone's idea and build on it with $$, there needs to be a lot of groundwork done first to even get to that point and it looks like Apple has barely done any (looking straight at Siri for an example)