How Waymo Builds Self-Driving Cars

Hinzugefügt: 2026-05-17

[00:00:00]There is the obvious, which is the driver, the software that runs the car.

[00:00:03]But there are two other very important pieces of the tech stack.

[00:00:07]One, of course, is the simulator, which is this virtual environment where you can test the car before you deploy it on the roads.

[00:00:14]And then there is a critique, which is a piece of software that's able to detect suboptimal performance in the simulator or in the real world, because then you can hone in on those suburb suboptimal cities and then improve the driver and then verify that those fixes have indeed worked in the simulator.

[00:00:34]So it's really the driver, the simulator and the critique is that triad it in our tech stack that allows us to scale. You have the full stack when it comes to autonomous vehicles. That's right, Waymo.

[00:00:47]That's right. But some of your competitors are going for a very different approach. They're kind of the brain only approach, the end to end machine learning approach without as much hardware, without necessarily those three different components.

[00:00:59]That's right. What gives you conviction that your approach that the Waymo approach ends up winning out?

[00:01:05]Our tech stack not focuses not just on the driver, but also ensuring that all of the plans the driver generates. It's GenAI, of course, so it generates plans. But you have to make sure that those plans that it generates are safe. And so you need to build in safety checks in real time so you can make sure the driver is prioritizing safety.

[00:01:28]And so it's the driver of the simulator and the critique on board components that prioritize safety, which is an onboard validation or an independent validation layer to ensure the driver does the right thing.

[00:01:41]And when you look at all of that, that's what it takes to drive.