The author addresses this.

He notes Tesla drivers are expected to be able to intervene at any time. Both companies rely on human intervention. But his argument is Tesla doesn’t have the infrastructure to learn from all its data with the accuracy necessary to account for edge cases, which are mortally important for safety.

Tesla, per the author, will need to go through exactly the staging Waymo is doing to move to driverless, but is years behind. That’s the argument.