When it comes to the companies that are leading the charge into artificial intelligence (AI), Apple (AAPL -0.57%) has always been something of a black sheep. Leaders in the field like Alphabet (GOOGL 0.32%) (GOOG 0.37%), Amazon.com (AMZN -1.16%), and Microsoft (MSFT -1.82%) have aggregated their data in the cloud, which resulted in AI systems that could learn more quickly thanks to the constant stream of new data.
Apple has been more focused on user privacy and has taken a different path. The company has used differential privacy, a technique that is capable of identifying trends without transmitting personal data. This is accomplished by adding digital "noise" to the information contained on individual devices, thereby protecting the user's data and aggregating and transmitting only the trends. There has been an unintended consequence of this approach, and Apple stands to reap the rewards.
The iPhone X uses AI offline to protect user privacy. Image source: Apple.
The standard approach
Deep learning is the technology at the heart of the AI revolution. It is a technique that uses a neural network, a computer model based on the structure of the human brain, that tries to recreate our capacity to learn. It is taught to recognize patterns that might otherwise be missed. By processing all their data in the cloud, many tech companies have accelerated the training of their AI systems, giving them what many believed was a decided advantage.
Aside from the obvious privacy concerns, one of the downsides to this process is that the devices could be useless or inefficient in cases where they were disconnected from their "brain" in the data center, which could be located thousands of miles away. This presented limitations in cases where connectivity was limited or non-existent.
The Apple way
By focusing on the privacy of its users, Apple was forced to innovate in a particular area of the technology. The company has focused on at-the-edge or on-device computing, which moves the AI capabilities from the cloud to the device itself, allowing it to use AI offline.
With the release of the iPhone 8 and iPhone X, Apple gained an advantage by developing an A11 Bionic Chip with its own built-in neural engine, which Apple described as "the most powerful and smartest chip ever in a smartphone." This AI chip has "a six-core CPU design with two performance cores that are 25 percent faster and four efficiency cores that are 70 percent faster." Its leading edge "performance controller can harness all six cores simultaneously." Apple says the chip was designed specifically for machine learning algorithms.
Apple also released the Core ML framework with its iOS 11 software update. This gave developers the ability to access the A11 chip for certain machine-learning functions, and integrate AI capabilities into their apps, without the data ever leaving the phone. During Apple's earnings conference call in August, CEO Tim Cook said:
With iOS 11, we're also bringing the power of machine learning to all Apple developers with Core ML, enabling capabilities like space detection, object tracking and natural language interpretation. Core ML lets developers incorporate machine learning technologies into their apps with all the processing done right on device so it respects our customers' data and privacy.
Ahead of the pack
The ability to process data locally while protecting personally identifiable information provides significant advantages. Privacy rights advocates and regulators in the U.S. and abroad are calling for stricter controls and greater insight into the data that is being sent to the cloud.
Companies like Google and Amazon could be at the center of a potential firestorm should any of the collected data be leaked, fall into the wrong hands, or be made available to third parties.
With its longtime emphasis on protecting user privacy and data, Apple has gained the lead in the realm of offline AI. Maybe it wasn't unintended after all.
