Anybody familiar with the design Tesla uses for self driving? Do they simply input video clips of common scenarios and what the correct output should be? Is it

Anybody familiar with the design Tesla uses for self driving? Do they simply input video clips of common scenarios and what the correct output should be? Is it just a single neural net, or are there rules hard coded in? Multiple neural nets handling different things? I assume there are many approaches to automated driving that are effectively dead ends. But would be interesting if just having enough video clips/correct corresponding behavior was enough to build a fully functioning model.