🧠 How AI Makes Face ID and Social Media Work

Subscribe | AI Toolkit | Meet The Team

In today’s Sunday Special:

Read Time: 7 minutes

📜THE PRELUDE

AI is so seamlessly integrated into our smartphones that we rarely notice it. We tap, swipe, and type through the digital world without a second thought, unaware of the AI optimizing every interaction.

Perhaps the most overlooked and underappreciated feature is your smartphone’s Facial Recognition. Despite being used billions of times per day, it’s rarely discussed.

In contrast, social media platforms like TikTok, YouTube, and Instagram are constantly in the news. We marvel at their Almighty Algorithms, sometimes even feel violated by their eerie accuracy, yet continue to devour the content they supply.

Today, we’ll explore how Facial Recognition and the Almighty Algorithms work. More specifically, Apple’s Face ID and Instagram’s Explore Recommendations System.

🍎HOW APPLE’S FACE ID RECOGNIZES YOU

At the heart of Apple’s Face ID is the TrueDepth Camera, which has three hardware components and employs ML techniques to collect, refine, and adjust your facial data. What’s even crazier is all this happens within a fraction of a second. Here’s how it works:

When you swipe up to unlock your iPhone, the TrueDepth Camera springs to life. First, the Flood Illuminator shines a controlled beam of IR Light onto your face. Then, a Diffuser evenly spreads this IR Light across your face to prevent bright areas or dark spots from interfering with the process. While the IR Light illuminates your face, the Dot Projector projects more than 30,000 IR Dots that capture your unique facial features.

Once the Flood Illuminator lights up your face with IR Light, the IR Camera within the TrueDepth Camera converts the facial features captured by the IR Dots into an image that’s sent to the Apple Neural Engine (ANE): an NPU that analyzes the image to extract the distance between your eyes, the shape of your nose, and the contour of your jaw. Then, these facial features are converted into a unique numerical representation of your face, known as a Facial Template. Apple’s Face ID compares this Facial Template to the Stored Facial Template you created when you first set up your iPhone. If they match, your iPhone’s unlocked.

The Stored Facial Template isn’t static; it evolves as your face changes, such as growing a beard, wearing sunglasses, or aging. So, how is this possible? Apple’s Face ID leverages the following ML techniques:

This blend of the TrueDepth Camera and ML techniques reduces the probability of someone else unlocking your iPhone to 1 in 1,000,000.

Your Stored Facial Template is kept inside the Secure Enclave, a special hardware component built into Apple’s iPhone, iPad, and MacBook processors that acts as a mini-fortress inside the central processor. This mini-fortress keeps your data ultra-secure.

📱HOW INSTAGRAM CURATES YOUR FEED

After unlocking our iPhones, we often navigate to social media platforms, where the Almighty Algorithms guide us. Instagram provides curated content to over 500 million people every day through Recommendation Systems like Explore. Here’s how it works:

Instagram’s Explore Recommendations System pulls thousands of potentially engaging Instagram posts from a massive pool of content during Retrieval. To do this efficiently, Instagram uses a Two-Tower NN:

By comparing the numerical similarities between a User Embedding and a Media Embedding, the Explore Recommendations System selects thousands of Instagram posts that likely align with your interests.

After Retrieval, the Explore Recommendations System uses multiple ML techniques to Rank the thousands of selected Instagram posts based on Engagement Predictions: how likely you’ll like it, how long you’ll view it, and how likely you’ll share it. Instagram posts with the highest Engagement Predictions move to the top of your Explore Page.

After Ranking Instagram Posts, the Explore Recommendations System deploys Caching, which involves storing pre-ranked Instagram posts so Instagram doesn’t have to create new Rankings every time you navigate to the Explore Page. There are three types of Caches:

A/B Testing allows Instagram to refine your Explore Page by testing different variations of Instagram posts through two Groups:

By comparing the Key Metrics between these Groups, like Dwell Time (i.e., how long you spend on the Explore Page) or Tap Rate (i.e., how often you tap on an Instagram post), Instagram can measure the effects of different variations of the Explore Page to optimize these Key Metrics.

🔑KEY TAKEAWAY

From unlocking our iPhones to scrolling through Instagram, AI quietly shapes our digital world. While OpenAI’s Reasoning Engines and Anthropic’s AI Agents consume the spotlight, the seamless integration of AI through Apple’s Face ID or Instagram’s Explore Recommendations System helps contextualize the latest AI developments and, ultimately, separates AI Hype from AI Reality.

📒FINAL NOTE

{{rp_personalized_text}}

Copy and paste this link to friends: {{rp_refer_url}}