This is a huge sign of confidence that they think they can do this safely and at scale... Freeways might appear "easy" on the surface, but there are all sorts of long tail edge-cases that make them insanely tricky to do confidently without a driver. This will unlock a lot for them with all of the smaller US cities (where highways are essential) they've announced plans for over the next year or so.
I don’t live in a served market yet so I haven’t yet tried Waymo. However I have used SuperCruise and BlueCruise from GM and Ford.
What I’ve noticed from those other systems is that a human in the loop makes the system so much more comfortable. I’ve had times where I can see the red lights ahead and the system is not yet slowing because the car immediately in front of me isn’t slowing yet. It’s unsettling when the automated system brakes at the last moment.
Because of this experience the highway has been the line in the sand for me personally. Surface streets where you’re rarely traveling more than 45 mph are far less likely to lead to catastrophic injury vs a mistake at 70 mph.
I don’t think Waymo is necessarily playing fast and loose with their tech but it will be interesting how this plays out. A few fatal accidents could be a fatal PR blow to their roll out. I’m also very curious to see how the system will handle human takeover. Stopping in the middle of a freeway is extremely dangerous. Other drivers can have a lapse in attention and getting smoked by a semi traveling 65 mph is not going to be a good day.
How will Waymos handle speed limits on highways? In the city, they seem to stick to the rules. A large percentage of drivers in the bay area, including non-emergency police, drive well above the legal limit regularly. Unless Waymo sticks to the slow lane, it's going to be a weird issue.
All of the serious problems I have seen with Waymo navigation so far have had to do with busy urban streets. Trying to make use of blocked non through way alleys, turning around in driveways when other vehicles are exiting, coming to a complete dead stop on busy one way streets, failing to brake predictably for pedestrians walking into lanes, suddenly backing up a half block from stopped at a red light in order to change lanes, and so on. Freeways are a simplified driving environment that should suit current technologies well.
I have seen a Waymo do a very stupid thing where it darted across a busy street, and it left very little margin of error for the oncoming traffic, which happened to be a loaded dump truck that could not have stopped. The dump truck driver was clearly surprised. It was a move that I never would have made as a driver. Did they dial the aggression up? I'm sure they're safer than humans in aggregate as there are some dumb humans out there but it's not infallible.
I was in SF a few weekend ago and rode both Waymo and normal Lyft style taxi cars. the Waymo was a better experience in every single way. One of the Lyfts i was in drove on the shoulder for a while like it was a lane. The Waymos were just smooth consistent driving. No aggressive driving to get you dumped off so they can get to the next fair.
Is that human vs robo or is it just that one had the enshittification dial turned up earlier?
If you were running a private equity robotaxi firm and your bonus relied on 1% more rides wouldn't you be dialing up the aggressive driving? Repeat for a few quarters and the robot will be cutting the same corners that the human is forced to.
Some future Fight Club reboot will reference your ChatGPT logs that show you asked how much the corporation would need to pay to the people killed in crashes vs increased profit to find the profit maximising level of dangerous driving.
An interesting prospect is that a bunch of autonomous cars on the freeways might have a meaningful impact in preventing traffic jams (specifically those "phantom jams") [0] simply by driving in a calm and pondered way always at a constant distance.
Road in a Waymo last weekend in Austin. Amazing experience. I was surprised at how mundane it felt. I had to keep looking at the empty driver's seat to remind myself that I was experiencing science fiction becoming reality.
I will say, I was surprised that the interior of the car was kind of dirty. I would imagine this is going to be a growing issue these FSD taxi fleets are going to have deal with. Lots of people will behave poorly in them.
I am really excited for this. Once going home with my family via Uber in SFO we realized on the freeway that our driver was high and driving at 80-85 mph.
It was a really scary experience and I couldn’t do much about it in the moment.
Don't bother installing unless you know Waymo is available.
Otherwise the App frustratingly runs you through onboarding and then tells you it is unavailable in your area. I had tried because they were supposed to be coming to New Orleans.
A lot of people rely on Uber and Lyft for supplemental or primary income, so this could be very disruptive if it continues to scale. Are we not worried about this in the medium-term?
Also I appreciate many of the random human interactions I've had with Uber/Lyft drivers. Of course not every ride was great, but many drivers had stories and experiences that no one I usually meet would have. For me, the safe but bland experience of a self-drivng car isn't worth losing the human touch, not to mention taking away income for human drivers.
Driving all happens locally in the onboard computer on the Waymo. The car does maintain internet connectivity - but generally it's used for non-driving scenarios. (traffic info, next pickup drop off, entertainment systems, etc)
There are cases where the onboard computer can't make a decision or needs "help" - in which case a support specialist is presented with options the onboard computer needs help deciding between. To be clear - the human is not driving it's more the car asks "Hey - there's something ahead and I am unsure if it's safe to proceed. Here's a video clip of the thing I'm seeing. Help?" Common cases might be an out of distribution thing like steam or an unidentifiable object in the road.
In a "worst case" mode - a human can remotely give the onboard computer a directed path to follow - eg "draw points and follow this path" to get back to where it needs to be. Even then - the onboard computer is following the path but still maintaining it's constraints "eg don't hit pedestrians."
This makes Dallas much more viable. I'm not sure if I'd have a usecase for waymo other than going to the airport. The other use cases such as going to work when my car is in the shop require the freeway too. Only non freeway thing I'd use it for otherwise is coming home from a neighborhood party or local restaurant/bar when I've been drinking and I don't really drink.
39 comments
[ 1.8 ms ] story [ 64.7 ms ] threadWhat I’ve noticed from those other systems is that a human in the loop makes the system so much more comfortable. I’ve had times where I can see the red lights ahead and the system is not yet slowing because the car immediately in front of me isn’t slowing yet. It’s unsettling when the automated system brakes at the last moment.
Because of this experience the highway has been the line in the sand for me personally. Surface streets where you’re rarely traveling more than 45 mph are far less likely to lead to catastrophic injury vs a mistake at 70 mph.
I don’t think Waymo is necessarily playing fast and loose with their tech but it will be interesting how this plays out. A few fatal accidents could be a fatal PR blow to their roll out. I’m also very curious to see how the system will handle human takeover. Stopping in the middle of a freeway is extremely dangerous. Other drivers can have a lapse in attention and getting smoked by a semi traveling 65 mph is not going to be a good day.
If you were running a private equity robotaxi firm and your bonus relied on 1% more rides wouldn't you be dialing up the aggressive driving? Repeat for a few quarters and the robot will be cutting the same corners that the human is forced to.
Some future Fight Club reboot will reference your ChatGPT logs that show you asked how much the corporation would need to pay to the people killed in crashes vs increased profit to find the profit maximising level of dangerous driving.
These vehicles very regularly block traffic because they can’t maneuver in congested areas with the finesse of a human driver.
Aggressive driving isn’t always bad. Sometimes it’s to unblock others waiting behind you so they can get somewhere they need to be.
[0] https://www.youtube.com/watch?v=m74zazYPwkY&t=1860s
it seems like these robotaxis have been around long enough to have conclusions now
I will say, I was surprised that the interior of the car was kind of dirty. I would imagine this is going to be a growing issue these FSD taxi fleets are going to have deal with. Lots of people will behave poorly in them.
It was a really scary experience and I couldn’t do much about it in the moment.
Otherwise the App frustratingly runs you through onboarding and then tells you it is unavailable in your area. I had tried because they were supposed to be coming to New Orleans.
Also I appreciate many of the random human interactions I've had with Uber/Lyft drivers. Of course not every ride was great, but many drivers had stories and experiences that no one I usually meet would have. For me, the safe but bland experience of a self-drivng car isn't worth losing the human touch, not to mention taking away income for human drivers.
I keep seeing them around my home in Menlo Park (Redwood City), but they're still in testing phase and not available for booking yet.
There are cases where the onboard computer can't make a decision or needs "help" - in which case a support specialist is presented with options the onboard computer needs help deciding between. To be clear - the human is not driving it's more the car asks "Hey - there's something ahead and I am unsure if it's safe to proceed. Here's a video clip of the thing I'm seeing. Help?" Common cases might be an out of distribution thing like steam or an unidentifiable object in the road.
In a "worst case" mode - a human can remotely give the onboard computer a directed path to follow - eg "draw points and follow this path" to get back to where it needs to be. Even then - the onboard computer is following the path but still maintaining it's constraints "eg don't hit pedestrians."
You can read more about this here: https://waymo.com/blog/2024/05/fleet-response