4 ms·
California publishes robocar “intervention” reports – Google/Waymo is far ahead
- Fricken 10y agoIt's very early in the game for most of Waymo's competitors, and next year's disengagement reports will be more telling, as it will give us an indication as to the rate of progress being made by the various players in this race. GM's Cruise appears to be on a solid trajectory. Tesla appears to be in a state of clusterfuck. They made the committed decision to build expensive sensor suites and a PX drive 2 computer worth 'a few thousand dollars' (according to Jen-Hsun Huang) into every new Tesla. This marries Tesla to an incremental development approach that certainly good for generating excitement and engaging users near term, but the holy grail is L4, and Tesla is putting a whole bunch of unnecessary obstacles between themselves and that longer term objective. So long as it isn't safety critical, it isn't hard to demonstrate impressive proof of concept, but what is clear by Waymo's gigantic numbers is that mopping up those last remaining edge cases and validating their system against them is actually a monstrous undertaking. Not impossible, just really hard. The closer you get to the finish line, the harder it gets. So my guess is that whatever progress Tesla makes with HW2, they will have to set up a separate dedicated division for testing and validating a proper L4 vehicle if they wish to keep pace with the rest of the peloton. Every other big player has abandoned any notion of taking an incremental approach towards full robotaxi-grade autonomy for good reasons.