Welcome back, AI prodigies!

In today’s Sunday Special:

  • 📜 The Prelude

  • 🛰️ The Seismometer-Based Approach

  • 📱 The Smartphone-Based Approach

  • 🏚️ The Real-World Impact?

  • 🔑 Key Takeaway

Read Time: 7 minutes

🎓Key Terms

  • Machine Learning (ML): Leverages data to recognize patterns and make predictions without being explicitly programmed to do so.

  • Convolutional Neural Networks (CNNs): A network of specialized filters that detect complex visual patterns like edges and corners across images and videos.

🩺 PULSE CHECK

Do you take cover when your smartphone sends you an earthquake alert?

Vote Below To View Live Results

Login or Subscribe to participate

📜 THE PRELUDE

The night cradles the world in stillness, while your bedroom encloses you within a cocoon of calm. You’re fast asleep, dreaming of hidden realms where time drifts softly, slowly, and seamlessly. The steady hiss of the humidifier lulls you deeper into your wandering mind.

Suddenly, a high-pitched alarm shatters the stillness, jolting you awake. As your hands scramble to silence it, a piercing robotic voice warns: “Earthquake! Earthquake! Prepare to drop, cover, and brace.” You spring out of bed and dive beneath the bedroom desk, pressing yourself flat against the floorboards.

Within a heartbeat, a distant rumble rolls through the bedroom: frames swaying on walls, lamps rattling on tables, and drawers sliding on tracks. The ceiling fan sways violently, slicing the air with a frantic whoosh before smashing into the bed. You begin to wonder: Will I survive? Before you can think of the answer, the tremors cease.

For decades, EEW Systems have primarily been implemented across wealthy countries like the U.S. and Japan due to the high cost of constructing and maintaining scientific-grade seismometer networks. They’re extremely fast and highly accurate, but leave billions of people without early warnings.

Today, AI has transformed smartphones into seismic sensors, giving billions worldwide a chance to prepare for intense earthquake tremors. So, how exactly does AI predict when the ground beneath your feet will shake?

🛰️ THE SEISMOMETER-BASED APPROACH

⦿ 1️⃣ The Science of Earthquakes?

The Earth consists of four distinct layers: crust, mantle, outer core, and inner core. The crust is made up of many pieces like a puzzle, constantly sliding past one another and bumping into each other. These puzzle pieces are called tectonic plates, and the edges of these tectonic plates are called plate boundaries.

The plate boundaries are made up of many rigid, rough, and rocky “faults.” Sometimes, the plate boundaries get stuck along these “faults.” When they finally slip, energy is released in a sudden burst, and that’s what causes earthquakes.

Based on current geological monitoring and seismic data, the most active plate boundaries considered most prone to major shifts are concentrated along the Pacific Ring of Fire. For context, approximately 90% of the world’s earthquakes occur there.

⦿ 2️⃣ The Concept of Early Warnings?

In 1985, Caltech seismologist Thomas H. Heaton published the first paper on the modern concept of developing early warning systems for earthquakes. He proposed SCAN,” which relied on continuous monitoring of seismic data to provide highly reliable advanced alerts of tremors, enabling proactive actions such as seeking cover, stopping trains, or shutting down industrial machinery.

In the early 2000s, the U.S. and Japan built EEW Systems by carefully placing thousands of seismometers: sensitive sensors anchored deep within the crust to detect and record ground vibrations. Despite this, during severe earthquakes above magnitude 6.0, these sensitive sensors can underreport the earthquake’s strength because the shaking exceeds their optimal measurement range.

To maintain accuracy, seismic scientists integrate data from the GNSS, which measures how far the Earth’s surface physically shifts relative to satellites. For example, during the 2011 Tōhoku earthquake and tsunami, which was a magnitude 9.0, Japan’s main island shifted over 2.4 m east.

📱 THE SMARTPHONE-BASED APPROACH

⦿ 3️⃣ Android Devices = Earthquake Detectors?

Google Research recently developed Android Earthquake Alerts (AEA),” which leverages 2.5 billion Android devices worldwide to detect the initial tremors of an earthquake and send life-saving alerts before destructive shaking begins.

So, how did Google Research turn Android devices into powerful, pocket-sized earthquake detectors? Every Android device has an accelerometer: a tiny silicon sensor that detects tilts, shifts, and forces across three axes (i.e., X, Y, Z). It enables “smart” features like screen rotation and fitness tracking.

They discovered that when Android devices are placed on a stationary surface, ranging from a nightstand to an office desk, it eliminates everyday “macro” noise. For example, if you’re walking with your Android device in your pocket, the accelerometer is quickly overwhelmed by constant leaning, swinging, and vibrating. Without this everyday “macro noise,” the tiny silicon sensor becomes acutely aware of specific vibrational frequencies.

An earthquake produces a very specific waveform called a “seismic signature.” Google Research trained an on-device ML Model to recognize the vibrational differences between the constant hum of a washing machine, the slamming of an apartment door, and the high-frequency jolt of P-waves”: the initial, typically weak, shaking that often goes undetected before the destructive S-waves arrive.

⦿ 4️⃣ Crowdsourced Confirmation?

When Android devices suddenly detect these high-frequency jolts, they quickly capture up to 100 recordings, collectively called a Vibration Window (VW). These recordings are sent to Google’s earthquake detection server, which deploys CNNs to identify “coherent motion,” where thousands of Android devices in the same area move in unison along the three axes (i.e., X, Y, Z).

Once CNNs confirm the earthquake, AEA simultaneously calculates the estimated epicenter and forecasted magnitude to predict which local communities will experience the strongest shakes. These predictions are expressed via the MMI Scale (i.e., 1-10), where a 10 would override all system settings and emit loud siren sounds with a push notification stating: “DROP, COVER, HOLD ON!”

🏚️ REAL-WORLD IMPACT?

⦿ 5️⃣ Do the Facts and Figures Add Up?

More than 50% of earthquake injuries can be prevented if people receive an early warning and respond appropriately. Between 2021 and 2024, Android devices detected more than 11,000 earthquakes across 98 countries, successfully delivering over 790 million early warnings via push notifications.

This peer-reviewed study surveyed over 1.5 million Android users and found that roughly 79% of those who received early warnings but didn’t feel any shaking still rated them “extremely helpful,” suggesting that simply being readily informed reduces anxiety during a potential natural disaster.

🔑 KEY TAKEAWAY

Earthquakes are a constant threat to populations around the globe. AEA provides billions with a few precious seconds of warning before the shaking starts. Those precious seconds could be enough time to get off a ladder or unplug the stove.

📒 FINAL NOTE

FEEDBACK

How would you rate today’s email?

It helps us improve the content for you!

Login or Subscribe to participate

❤️ TAIP Review of The Week

“Nice job, guys! Love all the in-depth explanations on cool AI use cases.”

-Luca (1️⃣ 👍Nailed it!)
REFER & EARN

🎉Your Friends Learn, You Earn!

{{rp_personalized_text}}

Share your unique referral link: {{rp_refer_url}}