Apple’s AI Advantage

hellotalk-horizontal_1000x198

My thanks to HelloTalk for sponsoring this week! HelloTalk helps you learn over 100 different languages by connecting you to native speakers of the language you want to learn. Read why, as a language teacher and learner, I think it’s valuable or download HelloTalk today!


There are few who would challenge that one of Apple’s significant competitive advantages is its ability to design and manufacture physical products. Even fewer, perhaps, would challenge that Google’s significant competitive advantages are it’s access to data and algorithms and artificial intelligence (AI) which process that data. As a upcoming frontier in technology appears to be AI, many have assumed that Google’s advantages are so superior to Apple’s that the AI frontier is Google’s battle to lose, and Apple’s to struggle to win. It is my contention, though, that while Google has significant advantages in AI, Apple has advantages critical for competing in the AI frontier.

To argue for my position, I will identify the critical competencies necessary for competing in AI, then argue that Apple’s advantages in some competencies give them, at minimum, an equal fighting chance with Google’s advantages in different competencies.

What competencies do you need to compete in AI?

While there are dozens of competencies that a company needs to compete in AI, there are, at minimum, four:

  1. AI
  2. Data
  3. Viable business model
  4. Positive user experience

AI

This probably goes without saying, but you have to have the algorithms, machine learning, etc. that make AI possible. This is what most people think of when they think of the AI frontier, as it is this competency that is the newest technology frontier. Each year, as we hear of AI arriving, this is what people refer to.

Data

Data means two things. First, to develop an AI, you need access to massive quantities of data upon which the machines can learn and the algorithms can work. Second, AI needs user data on which the algorithms can run. An AI developed and trained with data but without access to user data simply isn’t the AI that we dream of.

Business model

When speaking about AI as the next technology frontier, most people ignore the need of a company to profit from the AI. The reality is, however, that people need incentives to do the immensely hard, research-intensive, groundbreaking work of developing AI. An AI has immense societal benefits, but it’s hard to see foundations, non-profits, and philanthropy producing an AI that’s useful to the masses. If history is any indicator, for AI to benefit hundreds of millions, even billions, there must be a business model that sustains it. A company must have an incentive to develop and maintain AI: either a profit-making venture or a profit-preserving moat against competition. More than that, the business model must be viable or else the most incredible AI will be adopted as widely as the Amazon Fire phone.

Positive user experience

For AI to achieve its promise, it must be used and, to be used, it must be useful or delightful or otherwise beneficial to users. This may sound pedantic, but the tech sector produces huge numbers of products that are technologically advanced and practically asinine (I’m looking at you , $70 Egg Minder or at you $99 HapiFork). If AI does not provide a positive user experience, users simply will not utilize it, and developing that AI becomes even harder as users are less likely to give something a second chance than a first.

This positive user experience is not simply in the quality of AI (the classic Siri v. Cortana v. Google Assistant v. Alexa contests) but also in the quality of whatever physical device contains the AI (phone, watch, ring, earpiece, home speaker, car, etc.), the physical attractiveness or fashionability of that item (if visible, worn, displayed, driven), the security and privacy risk of the AI, and the “human element” of AI (it won’t embarrass you, creep you out, creep your mother-in-law out, try to kill you, break the law, act immorally, etc.).

All Competencies Are Required

Without competency in all four areas, any product that is developed is doomed.

  • Data + Business model + Data + User experience – AI = unintelligent product
  • AI + Business Model + User experience – Data = theoretical product
  • AI + Data + User experience – Business model = short-lived product
  • AI + Data + Business model – User experience = disliked product
  • AI + Data + Business model + User experience = history-changing product

Apple’s AI Strengths and Weaknesses

I’ll address each competency and assess Apple’s strengths and weaknesses:

AI

Apple has repeatedly reference AI, machine learning, or constituent parts (eg. LSTM) in their keynotes, earnings calls, and frequent media appearances. Still, for all their talk, no “AI product” has been released, we have only a marginally-improved Siri, and the general consensus in the tech sector seems to be that Apple is relatively weak in its development of AI while Google seems years ahead.

I think this assessment is premature. While Google Photos seems to be far superior to Apple Photos, it is unclear how substantial the AI delta is compared to Apple’s other advantages: users may prefer Apple’s privacy stance for photos over Google’s, prefer its native integration with iCloud, etc. Even if Google’s AI-infused products are as superior to Apple’s as their Photos are, it is unclear if that advantage alone is enough to win users over to their product.

Yet it is not at all clear that Google’s AI products would be that much better than Apple’s. Google’s first foray major into AI–Google Assistant–is better than Siri at some tasks and worse at others (for example, see reviews by Marques Brownlee or Business Insider). Even if it is, overall, better, it is not a slam-dunk case and, again, it is not clear that Google’s advantage alone would be enough to win the AI battle.

However, Google has the clear advantage from a cultural point of view: Apple’s obsession to perfecting every detail before shipping is contrary to the ship-and-constantly-improve approach that an AI seems to require. This is the cultural reason why Siri has improved at such a slow pace relative to other companies. How significant an advantage that gives Google remains to be seen, but it does seem clear that the advantage in this area is Google’s.

Data

In this category, the advantage seems to be more clearly Google’s, as they have amassed massive amounts of proprietary data through their search, ad (tracking & identity), and Android businesses. That data enables them to train an AI in a way that it seems difficult Apple could match without a search, ad, or data-collecting phone business. Furthermore, Apple’s hard stance on privacy (“We don’t want your data“) seems to limit their ability to access user data.

Against this, it should be said that it is possible to train AI on publicly available data, then run the AI locally on your device; it takes immense amounts of computing power to do the former, but the latter is well within the capabilities of an iPhone processor. As Federighi shared on the WWDC Talkshow, Apple did this with their Photos app. This doesn’t eliminate Google’s advantage here, as they clearly have more data, especially user-specific data, but it does reduce their advantage to the point that it must be compared against Apple’s other advantages.

We also have yet to see whether Apple’s application of differential privacy enables them to get the data that they need to train and run an AI. It could be that the general wisdom of the tech industry is wrong in this regard, just as it has been before:

Still, Apple’s lack of cloud-centric identity that allows for cross-device data learning from data is a clear weakness. Google certainly has more access to data and does not have the qualms to share, store, and use that data to improve its services. That does give it an advantage here.

So far, Google has an advantage over Apple in AI and Data, an advantage widely recognized in the tech industry, but these are not the only competencies needed to succeed in AI.

Business Model

Both companies have clear incentives to generate or preserve profit from AI, as it could be worth hundreds of billions, so extensive are its potential applications. However, in terms of ability to profit from AI, Apple significant business model advantages over Google.

Apple’s AI business model is clear: consumers buy Apple products that come with AI and Apple profits off of lucrative margins on those products.

Google’s AI business model is less clear, as Ben Thompson argued on Exponent: Google’s native search ad unit has been “one of many options” from which a user makes a choice, but part of the premise of AI is that it would make choices for the user. Thus, AI is not monetizable through Google’s current ad business model (with Android as its moat). Likewise, though Android Wear would be a natural place for AI, Google has no path to monetization here, either through ads or licensing the OS to OEMs.

Google has suggested that chatbots could be a place for an embedded AI, but the present UX is inept and has an unclear method to monetization, even if Google’s messaging apps had sufficient users to be a mechanism for monetization.

What is Google’s business model for AI, then? The same as Apple’s: embedding AI like Google Assistant exclusively in Google devices like the Pixel phone and, I predict, Google-designed wearables. This is a clear business model for a Google AI, but there’s just one problem: designing and selling physical products is one of–if not the–strongest competency Apple has. It’s not just the ability to design a product with an attention to detail that Apple does, but the manufacturing process, supply chain, and distribution that Apple has masterfully developed over decades of selling electronics. Google can learn these skills, of course, or partner with those who have them, but Apple has decades of experience in these fields. This advantages is strengthened by the fact that you consider Apple has partnerships with IBM, SAP, and Cisco, and others that get their products into business sectors–as well as partnerships with every major carrier in the world and thousands of retail stores to distribute products. Google’s manufacturing, supply chain, and distribution network are minuscule in comparison. Google’s online store is great for techies, but you are hard pressed to find a store where you can buy one. In case Apple’s advantages here were not abundantly clear, consider Google’s history in producing and selling physical goods. The Nexus devices never had adoption widespread enough to monetize an AI; the Nexus Q died before it was born, and Google TV never went anywhere before it was killed. Chromecast is, perhaps, Google’s most successful physical product in terms of marketshare, but Google can’t even ship it in a box with corners that align:

Google has not developed the culture and skills needed to excel in designing, manufacturing, distributing, and selling physical goods. Apple’s advantage in this category is clear. It is unclear how great that advantage compares to Google’s advantages in other competencies, but it is overwhelming.

User Experience

As I noted earlier, a positive user experience comes not just from the AI, but the quality of the physical device, its security/privacy, and the human factor. In each of these, Apple has significant advantages.

Design. While AI exists abstractly, humans, as physical beings, can only access it physically. So long as the device is physical, then it must be designed as attractive and, if worn, fashionable. While Google’s design skills for products are decent and seem to be improving (Pixel seems better than Nexus, though OnHub & Home are a bit quirky IMO), their skill is less clear if AI is embedded in a wearable device. After all, Google’s first wearable was Glass and Apple’s was Apple Watch. Perhaps more important, whatever physical device has to have a build quality that creates a positive user experience; Google’s product history (above) gives little confidence in this regard. At best, Apple and Google’s competencies in this category are even, but I give the edge to Apple.

Security. In terms of security and privacy, the clear winner here is Apple. There is a reason that Android has been called a toxic hellstew for security as Android devices have been compromised on so many occasions its sickening. Meanwhile, iPhones are so secure that police have to steal them unlocked to access them and there was the entire FBI debacle. While Google has managed to still be successful with Android despite these security flaws without most consumers minding, the danger of a compromise AI is greater than merely a phone because of the amount of access to data it has, the trust a user would place in an AI, and the hiddenness of a hacked AI. Security becomes even more critical when it comes to AI that is embedded into home systems, wearables, or transportation. Still, while the advantage here is clearly Apple’s, it is unclear how great an advantage it is, as many consumers are either uninformed about security or do not care.

The human element. The last part of the user experience is what I call the human element, perhaps the most significant component of the user experience of AI. There’s a reason that nearly every movie featuring AI ends up with an AI that acts in inhumane and immoral ways. We want AI to have intelligence like human beings, but to only act in ways that are morally right–unlike many human beings. If AI merely trains off of existing human data, without a moral filter, it will become racist and sexist. It is not enough that AI learns to be like a human; AI must learn to be like a certain kind of human.

Put differently, we want AI to be like humanity in positive qualities, but not like humans in our negative qualities. We want AI to provide us the benefits of humanity without the errors of humanity; we want AI to think like a human, not like a computer. In I, Robot, just one of many AI-gone-wrong movies, the main character recalls how a robot saved his life instead of a child. He says:

I was the *logical* choice. [The robot] calculated I had a forty-five percent chance of survival, Sarah only had an eleven percent chance. That was somebody’s baby. Eleven was more than enough. A human would have known that. But robots, no, they’re just lights and clockwork.

Notice that what we are talking about here is not what algorithms can do, but what they should do: emulate moral behavior and disregard immoral behavior, even when it is human behavior. For AI to achieve the benefits we dream of, it must be a moral AI; it must be a certain kind of AI; it must take the good of humanity and discard our evil. Yet algorithms, logic trees, and machine learning cannot teach an AI to be good, kind, generous, truthful, or moral. Algorithms lead to fake news, racism and sexism.

technologyliberalartsAn AI that delights users with its humanity–not inhumanity–cannot be developed without intentionality to address the philosophical, moral, spiritual, and artful elements of humanity. Put differently, an AI that creates a positive user experience cannot be built except at the intersection of technology and liberal arts. There’s only one company that explicitly seeks to live at this intersection, and that’s Apple. The advantage for creating an AI that creates a positive and not immoral user experience is Apple’s.

Conclusion

I’ve argued that a company must have four competencies to compete in the AI space: AI, data, a viable business model, and positive user experience. In each competency, Google and Apple have different advantages and disadvantages. Google may have greater advantages in AI and data, but Apple has greater advantages in its business model and user experience. In totality, who has the greater advantage? I don’t think anyone knows the answer to that question. But it is far from a foregone conclusion that Google’s advantages in AI and data assure them of victory over Apple. The battle is just beginning.

You can subscribe to markdmill.com via RSSTumblr or by following MarkDMill on Twitter.


Sponsored App: HelloTalk

hellotalk_vertical_1000x761While writing my thesis for my Master’s degree in ESL, I was amazed to see the strong connection in research between mastering a language and receiving large amounts of quality input in that foreign language. In short, the more you interact in a foreign language, the faster you’ll learn it. The trouble is, most of us don’t live with easy access to native speakers of the language we want to learn. That’s where the free app HelloTalk comes in.

HelloTalk connects you to native speakers of the language you want to learn who are learning your native language. They have support for over 100 languages, including Spanish, Mandarin, French, Japanese, Italian, Arabic, Russian, and German. Since I studied Mandarin Chinese for 4 years while I lived in China, I searched for native Mandarin Chinese-speaking men who are my age and was immediately connected with dozens. A nice touch is that HelloTalk lets you search by age, gender, language, and even location, so that you can find exactly the person you’d like to talk with.

I have legitimately had fun using Mandarin to communicate again and helping others with their English at the same time. More importantly, I’m using the language and getting linguistic input, which is the most critical factor in acquiring a language.

If you want to learn, practice, or refresh a language, why not give HelloTalk a try? It’s free to download on the App Store today.

link-badge-appstore_2x

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s