Apple WWDC: My Three Big Takeaways

Yesterday’s WWDC keynote from Apple felt like drinking from a firehose. I have sat through a lot of developer keynotes and it was, by far, the fastest paced I’ve ever seen. Because of the pace, it is easy to miss some of the more important elements so I’d like to share the three big things that struck me.

More Power to Developers

From a developer standpoint, Apple came out swinging and made sure it was clear Apple platforms are the best place for developers to invest their time and resources. Apple made it clear they have four platforms for developers to think about, each one with a different focal point. iOS, WatchOS, tvOS and macOS make up the pillars for developers. Apple has begun opening up more of each platform and allowing developers added ways to take advantage of each.

Apple affirmed their developers are a primary factor in their differentiation, even though they did not use those words exactly. Developers are central to Apple’s future and making sure they have new things to work on is essential.

tvOS’ Single Sign On

As I saw this being announced, not only did it relieve a serious pain point but it struck me that we are inching ever closer to being free from the terrible hardware companies like Comcast, Dish, Cox, DirecTV provide us. As more apps start to include live streaming and on-demand access to network shows, the closer we get to not needing our service provider’s hardware.

With single sign on, I log in once to my pay TV service and, not only do I not have to put in my credentials for every app I download, I’m presented with a list of apps I can access and use with my pay TV service. Cable bundles still make sense and we will likely pay a fee for a set list of channels or apps but the key is we are seeing the unbundling from cable companies hardware and that is a big deal.

iMessage + Devs

This one may be the biggest takeaway. iMessage is (and has been for a while) one of the most important apps not just on iOS but across the board. First, we must acknowledge much of what Apple has done has taken some cues from WeChat. Facebook and WhatsApp have done this as well. However, if any messaging app has the potential to become a platform like WeChat it is iMessage.

Apple did many things to increase the communication experience, but those are really there to keep people engaged and make iMessage a desired app to have conversations with people. The real power here is in letting developers start to integrate more of their app experiences into iMessage. Letting consumer pay friends back via iMessage, order food, share a song, and much more is coming. This will position iMessage at the center of more experiences in the same way WeChat does in China. This is a fascinating move.

Just a point on Stickers. LINE filed for its IPO and, if any company does Stickers well, it is LINE. What became public knowledge is LINE made $268 million dollars on Stickers alone in 2015. Its users send 389 Stickers per day. To put that in perspective, LINE has a little over 200 million monthly active users. iMessage likely has more than double that. There are very interesting revenue opportunities for both Apple and third parties. This may be the one area I’m interested to see what developers do and may represent the most upside for both parties economically.

A Brief Word About Privacy and AI

This is a subject I’d like to spend an entire post fleshing out. However, I felt Apple’s effort to spend the time to articulate a concept called differential privacy was very telling directionally about where they are headed with machine learning. The idea of AI and machine learning gets thrown around for a lot of things. For example, for Apple to have deep algorithms built into the Apple Watch software able to combine motion and heart rate data to know when I’m exercising and when I’m not was all generated by deep learning. However, my privacy was never invaded for Apple to take that communal/crowd-sourced learning and use it to make their software better. My privacy does not need to be trespassed for Apple’s visual processing algorithms to know a dog is a dog or Hawaii is Hawaii. But moving some of that machine learning on device allows the Photos app, for example, to recognize Hawaii but also my wife and my family when we are in Hawaii. That last part, who is my family, is the part that can and needs to be learned but also the part to stay private.

The key takeaway with the idea of differential privacy is that some elements of noise are inserted intentionally into the system so the data can not be cross-referenced to uncover my private details. This is a big thing to try and understand and, while there are many institutional research papers on the concept, Apple is applying it to machine learning and mixing both local learning with communal learning in ways I’m not sure have been tried before. But if they are going to succeed in both deep learning and still protect my privacy, it seems like this is the way to go about it.

It will be interesting to see where Apple’s deep learning efforts come to the forefront, with the Photos app and improvements with Siri in the fall. We’ll see then if they are actually behind Google or not.

Published by

Ben Bajarin

Ben Bajarin is a Principal Analyst and the head of primary research at Creative Strategies, Inc - An industry analysis, market intelligence and research firm located in Silicon Valley. His primary focus is consumer technology and market trend research and he is responsible for studying over 30 countries. Full Bio

447 thoughts on “Apple WWDC: My Three Big Takeaways”

  1. Do you really think that Google, Microsoft, Facebook, Amazon violates your privacy to be able to develop their AI machine learning capability?

    There is no such thing as AI or machine learning locally on device, your iPhone does not have enough processing power, data and algorithmic tools to do it locally, even the most powerful personal computer can’t do it locally.

    Sometimes it baffle me to see how hard you are trying to fit Apple’spin and narrative into a science just to make them look Good

    do not let your love for Apple make you lose your objectivity Ben wake up

    1. Sorry you feel that way but your view is incorrect on my goal. And your incorrect in how much privacy you give up in other views. There is a balance and that is what I’d like to see struck. You also mistake my appreciation for the type of business Apple is for love. As an analyst my goal is to look at all sides of the coin with reasoned logic.

      1. First off all Ben Facebook, Google, Amazon build a profile on their user for the advertising business only which as nothing to do with developing AI and machine learning that was my point.

        i admirer you passion and love for Apple business ben i just find it odd for someone as informed as you who read white paper on AI and machine learning system to keep on repeating that SIRI is AI or that Apple can build their AI and Machine Learning system locally on your IPhone which make no sense for someone who know that much about the sciences of AI and machine learning system and algorithmic tools.

        1. I have not come across any experts in the field that do not think the underlying tech for Siri is not AI.. Discussions with experts are much more revealing about what is actually happening here and what is not. Apple is simply taking a different approach to it. How competitive it is we will see but at the same time I’ve had crappy success with Googles imaging AI in their photos app. Sometimes it works sometimes it is terrible. So overall I think everyone here has a long way to go.

          1. If an AI expert told you that Google Photo, Google Now, SIRI Cortana are AI either he is not an expert how he simply do not know what he is talking about

            the underlying tech of the application are tools that is part of AI but that doesn’t mean they are AI, DeepMind, IBM Watson are what can be consider to be AI.

            and the Issue Ben is that just as we human learn from conversation or communication the Primarily basic of any Machine learning system is about share learning hence a system that have access to your local Data only is not a learning system because it doesn’t share and contrast it’s knowledge with other or learn from other.

          2. Yes there is a difference between AI and machine learning, deep learning neural networks, etc. I’ve been welll educated on them all on everything from custom silicon to hardware, software, etc.

            Underlying systems all play a role to make up the whole. Understanding context, sentiment in language and so much more is very hard stuff and has foundational parts of machine learning. There will be a private and a public side to this. Apple’s approach benefits from the larger shared data without knowing it was me exactly. But they still see the patterns. Again, study the approach there is likely no right or wrong here and they are all marching in the same direction.

      2. It’s still purely straight Apple PR though:

        1- the models and mechanisms Apple uses to supposedly anonymize and obfuscate individual data (which they do collect, they themselves say so) have not been published, reviewed, validated. We’re basically asked to take their intent and ability on faith.

        2- the security that protects that data is purely obscurity. No documentation, no external audit, no hacking challenge like pretty much all other actors have. Google will pay you $100,000 for successfully hacking ChromeOS, which makes a white-hat hack more rewarding than a black-hat one.

        3- talk is cheap. They promise to anonymize and secure the data. It’s a pinky promise: iCloud’s EULA: “APPLE DOES NOT REPRESENT OR GUARANTEE THAT THE SERVICE WILL BE FREE FROM LOSS, CORRUPTION, ATTACK, VIRUSES, INTERFERENCE, HACKING, OR OTHER SECURITY INTRUSION, AND APPLE DISCLAIMS ANY LIABILITY RELATING THERETO.” (their shout, not mine). Time to put their money where their very loud mouth is ? Aside from external threats, do they back up their own pinky promise they won’t themselves use non-anonymized data for anything ?

        Maybe a pinch of salt would be warranted at some point ?

        1. Already pointed it out. Hence why I said we will see if it is competitive or not at the end of the article.

          What they are doing with differential privacy at scale is new. Much of the math here has been theoretical and not much put into practice.So we will see what Apple does here with this approach of learning from crowd sourced data without a user profile.

    2. It’s in beta Kenny. So not public. But I’d think within the next month or so we’ll see if your paranoia is justified or just paranoia.

      There are a lot of people who want to draw a target on Apple’s heart. We’ll know within weeks who is correct.

      1. My point is not about whether Apple can build good services or not

        the point is not a single thing that was announced at the WWDC is AI or machine learning system,

        They were just recommendation engine, image recognition, algorithmic pattern and app integration there was nothing from the keynote that had something to do with AI earning System with intelligence that can function on their own outside human intervention

        You can make the case that some of them are part of AI or machine learning system but that does not mean they’re AI just as the tire in your Car is not a Car.

        1. I don’t think Apple mentioned AI at all during the whole presentation. So I’m a bit confused why this is an issue. And honestly does it matter if it achieves the same thing?

          As I mentioned before let’s see what the punters think before we start complaining that they aren’t doing something they didn’t talk about.

          1. except that Apple Fan are comparing it with Google advance in AI and machine learning system which i responded too because they keep on confusing Google Photos or Google Now as real AI engine system instead of focusing on Deep-mind and the application potential that could come from it versus what Apple is doing

    3. At this point, we don’t yet know how good Apple’s AI stuff is. Also, we don’t know how well their differential privacy really anonymises user data. There are two thing I am looking for though, that will tell us this in the near future.

      1. How good are Apple’s AI-ish services?
      Voice recognition, photo recognition will most likely be the services through which most people will evaluate Apple’s AI prowess. The release of iOS 10 will allow people to evaluate it against Google’s AI-ish services.

      2. Will Apple use health data (not fitness) AI?
      The privacy of health data (as opposed to fitness data) is typically required to adhere to very high standards. If Apple’s differential privacy really works, I expect Apple to move into this area in a much stronger way. It is also likely that Apple will have to be audited in some way.
      If Apple manages to move into this area, then I would consider this to be proof that Apple’s privacy approach is real.
      Consider the controversy over Google DeepMind’s access to health data. Apple’s approach could nullify these concerns.

      I think we can only really discuss the validity of Apple’s approach after seeing these things play out.

      One thing is for certain though. Apple has given us a new perspective into the privacy-AI trade off debate, and that alone is significant.

      1. Privacy and AI has nothing to do with each other,
        and Apple has no AI engine system to which one can do evaluation versus let say Deepmind, or IBM Watson

        Again image recognition, recommending engine or SIRI are not AI, they are preprogram App that use some AI tools
        they’re not self learning system that can function outside human intervention hence Apple isn’t shows us anything yet related to AI or machine learning system that i’m aware of

        1. I’m not sure why you conclude that Apple does not have an AI system. I’m sure that they use deep learning to train Siri to do better voice recognition, for example. This is certainly an example of machine learning.

          1. because AI mean Artificial intelligence hence self learning and operated system

            Deep Learning are not the same a self learning just as Google deep learning system or tools are not the same a general self learning system such a DeepMind

            SIRI can’t be consider an AI because everything about it is preprogram by human intervention, it does not learn nor evolve on itself, it doesn’t understand context, logic,human emotion nor deep intelligence recognition and pattern

            that’s why i mean by that

          2. I’m not sure what you mean there. I’m pretty sure though that you are straying away from what most people currently mean by the term “artificial intelligence”.

          3. do not confuse what Blogger consider AI to be the same as what AI expert say .

            what i meant is AI are self learning system which SIRI is not

          4. Has Google said that DeepMind is directly being used by Google Assistant and Allo? Google’s blog says there is “deeply integrated machine learning” in Allo but does that mean it is using the DeepMind system to learn directly from my every input?

          5. Tensorflow is the Google AI engine system on which Deepmind, Search is build in, and there are many aspect to it,

  2. The way they opened up Messages (and Maps, Siri, even Phone) was very interesting. Even more so when you combine it with the rumours of an Android version of iMessage.

    On the one hand, iMessage for Android will be very important if iMessage is to truly be a platform. On the other hand, opening up iMessage means that it will be harder, but not impossible to take the iMessage experience in full to Android.

    Definitely something to think about.

    1. Possibly to settle the “ownership” matter. Messages are content, and the user owns them. They can’t be locked in to an ecosystem (though I’m sure other’s will vehemently disagree with me).

    2. No iMessage for Android. Apple has said no according to Walt Mossberg.

      “When I asked a senior Apple executive why iMessage wasn’t being expanded to other platforms, he gave two answers. First, he said, Apple considers its own user base of 1 billion active devices to provide a large enough data set for any possible AI learning the company is working on. And, second, having a superior messaging platform that only worked on Apple devices would help sales of those devices — the company’s classic (and successful) rationale for years.”


      1. I’ve been looking into this.

        Looking back at what Apple has done in the past, it is often when they backtrack on past statements that they make huge progress.

        Steve Jobs said that Apple was not going to do a phone because they weren’t good at selling through orifices.

        Apple also had great success selling larger phones, after years of playing them down.

        Therefore, if anything, I would take the senior executive comments as an indication that iMessage on Android could be huge.

        Poor Mossberg is not considered by Apple PR as a journalist who should get the true story of inside Apple (none are). He is just one channel which Apple uses to send out the message that they want to send, and the Apple executive was probably very calculating when giving Mossberg that comment.

Leave a Reply

Your email address will not be published. Required fields are marked *