Autonomous Autos – Cruise control

Cruise is mostly punished for bad behaviour.

• Although there was most certainly a technology failure, Cruise is off the roads mostly because the California DMV thinks that it failed to fully disclose data around an incident involving a pedestrian in a move that will delay the journey to autonomous driving yet again.
• Cruise has had its licence to operate in San Francisco suspended by the California DMV for an indefinite period as the DMV has lost confidence in Cruise’s ability to provide a safe service.
• Cruise was already under regulatory pressure after a series of incidents where Cruise vehicles have blocked the road or in one case collided with an emergency vehicle.
• Cruise’s fleet was already operating at 50% capacity while the incidents were being investigated but this appears to have been the final straw.
• It comes just as Waymo is beginning operations in the same area where Cruise has been operating which will come as a bitter blow to Cruise.
• I suspect that Waymo will suffer many of the same problems as the differences between the leading offerings are pretty slim these days and there is no reason to think that Waymo is better than Cruise or vice versa.
• The incident in question involved a pedestrian who crossed on a red signal and was hit by a human-driven vehicle adjacent to the Cruise vehicle and propelled into the path of the Cruise vehicle.
• The Cruise vehicle attempted to avoid the pedestrian but still made contact before aggressively coming to a stop.
• The Cruise vehicle then pulled over to the side of the road to avoid further issues, but it had not detected that the pedestrian was underneath the vehicle who was dragged 20 feet by the Cruise vehicle.
• This is another classic example of an edge case that no training or simulation is ever going to be able to prepare the vehicle to deal with once again highlighting the weakness of all systems that rely on deep learning.
• This weakness is that they have no causal understanding of what it is that they are doing meaning that they can not deal with a situation that they have not been explicitly trained for.
• This allows deep learning to work brilliantly for tasks where the dataset is both finite and stable but the road is neither of these things which is why there have been and still are so many problems.
• However, it looks like this is not the main reason for the suspension as the DMV believes that Cruise withheld video evidence of the pedestrian being dragged under the vehicle which Cruise emphatically denies.
• The DMV claims that it only found out about the dragging after speaking with another government agency.
• There is no way that Cruise could have expected that the DMV would not find out about the dragging and so I suspect that it did disclose the video in its original disclosure but for some reason, the DMV employees failed to find or watch it.
• Consequently, I think that the DMV has been overly harsh on Cruise but that does not mean that the incident should not trigger a safety investigation.
• I suspect that a Waymo vehicle would have performed in a similar manner to the Cruise vehicle so the safest thing to do would be suspend everyone until the investigations can be completed.
• Either way, this once again highlights the problems with autonomous driving as a human driver seeing a human go under their vehicle would have understood not to move the vehicle without any specific training.
• This is the difference between the humans and the machines and it is also why large language models hallucinate and make things up and progress to fixing this is glacially slow.
• Hence, I don’t think that autonomous driving is going anywhere beyond motorway driving with an alert driver for some time and I am sticking to my long-held target of 2028 before this becomes a reality.
• Given the lack of progress, even this may start to look optimistic.
• Consequently, falling valuations, shutdowns and dilutive consolidation look like they will continue for some time to come.
• I have no real desire to get involved here beyond the lidar companies, many of which have plenty of use cases outside of autonomous driving.