Apple’s Grounded A.I. Philosophy: Practicality in Pursuit of Excellence

Apple’s AI Progress

During Apple’s WWDC conference, the company highlighted its advancements in artificial intelligence (AI) and machine learning (ML).

Apple’s Previous Position

While other companies like Microsoft, Google, and OpenAI embraced cutting-edge ML technologies, Apple seemed less involved.

Apple’s Announcement

On Monday, Apple made significant AI announcements, including an improved iPhone autocorrect feature based on a machine learning program. It uses a transformer language model, similar to ChatGPT, and learns from user input to enhance accuracy.

Autocorrect Improvement

Apple’s chief of software, Craig Federighi, joked about the autocorrect’s tendency to replace certain words. The improved feature will learn user preferences, adapting to their typing habits.

Focus on AI and AR

While the new augmented reality headset, Vision Pro, gained attention, Apple showcased its commitment to state-of-the-art ML and AI. It aimed to enhance a feature used daily by one billion iPhone owners.

On-Device AI

Unlike competitors using server farms and large amounts of data, Apple prefers on-device AI models. The new autocorrect feature impresses because it runs directly on the iPhone, reducing the need for extensive data collection.

Apple’s Practical Approach

Apple refers to “machine learning” instead of “artificial intelligence” and focuses on the practical features enabled by the technology. It avoids discussing specific models, training data, or future improvements.

Example Feature:

AirPods Pro Apple introduced an enhancement to AirPods Pro that automatically disables noise cancelling during conversations. Although not explicitly mentioned as ML, it relies on AI models to solve a challenging problem.

Digital Persona and Neural Networks

Apple’s Digital Persona feature creates a virtual representation of the user’s face and body during videoconferencing with the Vision Pro headset. The company also mentioned other features utilizing neural networks, like identifying fields in PDFs.

Pet Recognition Feature

Apple received enthusiastic applause for a machine learning feature that identifies users’ pets and organizes pet photos into a dedicated folder.