An inside look at how Waymo collaborates with researchers from Google Brain in using AI, deep learning, and neural nets for its driverless vehicles
Andrew J. Hawkins / The Verge :
Context & Ripple Effects
This 2018 look inside the Waymo–Google Brain collaboration sits early in a decade-long pattern: Alphabet pointing its research labs at autonomy. A year earlier, The Atlantic had profiled Waymo's Castle test base and Carcraft simulations running 25,000 virtual cars over 8 million miles a day; weeks before this piece, The Verge compared how Waymo and Tesla each turn billions of driven miles into training data.
What makes the Google Brain story worth tracking is where it leads: by 2019 Waymo formalizes a similar tie-up with DeepMind on evolutionary-algorithm-inspired driving AI, and by 2026 it is using DeepMind's Genie 3 to generate synthetic worlds for edge-case training. The Brain collaboration is the first documented instance of Waymo treating Alphabet's labs as core infrastructure rather than optional research.
First-order effects
- Waymo gains direct access to Google Brain's deep learning and neural network research for its driver stack, an input rivals like Tesla must replicate with their own in-house AI teams since no equivalent lab is for hire.
Second-order effects
- Alphabet's lab-to-vehicle pipeline pushes competitors toward their own research bets — startups like Wayve, Waabi, and Autobrains later pitch end-to-end AI learning explicitly as a way to leapfrog Waymo's approach rather than out-collect it.
Third-order effects
- If the pattern holds, autonomy consolidates around companies that own frontier AI research: Waymo's arc from Brain neural nets through DeepMind algorithms to Genie 3 world models suggests the durable moat is the lab relationship, not the vehicle fleet.
The trend: Autonomous driving is becoming a contest between corporate AI labs, with Alphabet's internal research pipeline — Brain, then DeepMind — functioning as Waymo's compounding technical moat.