Front end collision avoidance/braking systems on all cars would make a significant dent in that number.
One thing I will also note, which sort of gets away from the topic of AV, is that roughly 45% of road fatalities every year are from unbelted drivers or passengers. There’s some stupid that apparently can’t be fixed.
Well, the CAV that won’t potentially kill people has yet to be invented.
Granted, but you seemed to suggest that self driving vehicles, as a concept, wouldn’t be a big improvement.
If they work, if they actually don’t kill people, it’s a big improvement, if they don’t work, it’s like any other technology that doesn’t work. Telephones that don’t work aren’t better than the telegraph.
As a concept, cars that successfully and safely drive themselves would change the world.
Assuming CAV stands for “Connected and Automated Vehicles”:
Thanks; as far as I can tell, @tullsterx dropped that in this thread without any sort of introduction. It would nice if people didn’t do that sort of thing.
Ah, that’s helpful. I thought it meant “Computationally Aware Vehicle”.
My Google search gave me “clean air vehicle” which didn’t seem to make any sense in the context of a discussion of self-driving cars.
Several reasons: People are living longer, and the United States is set up in ways that make it difficult to live without driving.
Parking spaces take up huge amounts of space in cities. Imagine being dripped off and then having the car go park itself someplace else We could redesign cities with wider sidewalks, more outdoor eating areas, etc.
Delivery of groceries, food, mail, and drugs would be much cheaper.
Traffic jams would be largely eliminated and highways and roads could be smaller
Drafting in convoys ar highway speeds would reduce drag and dramatically improve efficiency.
10s of thousands of deaths would be saved.
They weren’t loud about it at all (maybe some clickbaity media sites were, though). There were a few tweets that their FSD system no longer needed radar, and then it was quietly removed from the cars a couple of months later.
What Musk has always said, in any case, is that if you are going to have sensors other than cameras, they should at least work on a different wavelength. LIDAR fails at all the same times that cameras fail (fog, etc.), because it’s still just light. So there’s no advantage, particularly since the resolution is so low. And the depth information doesn’t get you much over what the cameras provide.
Radar is radio frequency, though, and can work in situations where cameras don’t. So it’s at least potentially useful. I think it’s unlikely to be critical (they aren’t retrofitting any cars with it), but it is at least interesting enough to experiment with. Note that the new radar is only going into S/X cars, which sell in fairly low quantities–I personally suspect it’s just an test to see if it’s useful. They can get a few tens of thousands on the road and collect data from it. A cheap experiment.
Even using the current FSD version, it’s clear that the cameras aren’t the limiting factor. It sees just about everything–better than I can in some cases (it often spots pedestrians in the dark before I do).
Where it fails is in the high-level decision-making, like figuring out what lane it should be in or what to do at a 4-way stop. That’s not going to be improved with LIDAR. It needs more general intelligence. The big question, as I see it, is if a “ChatGPT-level” intelligence is sufficient or if they really need something approaching AGI. FSD is a long way off if it’s the latter, but we’ll see.
Your first & third things that I quote above make no sense, and even less sense together.
As long as everyone is traveling in single occupancy vehicles, whether they are driving it themselves or an AI is driving and they’re snoozing in the back does nothing to alter traffic counts. Peak demand will still be peak demand and you’ll still need one vehicle per user to deliver the thing they need (door to door transport) at the time they need it (the same as most other working people).
To the degree your remote parking areas are more than a couple blocks away from the origin or destination, the cars will be ferrying themselves empty for umpteen miles once they’ve delivered their single passenger to the morning or evening commute stop. All you’ve done is increase the miles driven at the shoulders of the “rush hour”. Making it last longer and have more cars in it.
I mean, you’re splitting hairs…he tweeted at least half a dozen times saying radar wasn’t helpful and was being removed, and here is their actual announcement:
We are continuing the transition to Tesla Vision, our camera-based Autopilot system. Beginning with deliveries in May 2021, Model 3 and Model Y vehicles built for the North American market will no longer be equipped with radar. Instead, these will be the first Tesla vehicles to rely on camera vision and neural net processing to deliver Autopilot, Full-Self Driving, and certain active safety features. Customers who ordered before May 2021 and are matched to a car with Tesla Vision will be notified of the change through their Tesla Accounts prior to delivery.
It was understandably a big deal at the time.
I’m posting these questions separately to uncouple them from the previous post.
Again, what is the physical mechanism for training and then implementing generalized AI on the vehicles? This is not a question of doubt, I assume there is some sort of answer to it.
Has the Tesla FSD gotten better at driving in rain, snow, and fog? I’m comparing to Waymo, which as far as I can tell has made zero progress. (this is actually a question of doubt)
Was about to say… those points basically add up to “anything but redesign our society to not need so many individual cars”. I mean, they’re not “wrong”, in the sense that I don’t see anywhere in the horizon a true collective will to reverse 90 years of door-to-door-individual-conveyance as THE, first aspiration, and then expected norm.
Plus, true autonomous driving would open up enormous subscription services. I bet a HUGE number of urban and suburban folks would love to have a car show up in 5-10 minutes and whisk them to dinner without the hassles of owning it, parking it, or potentially driving it home after a few glasses of wine. If a autonomous car is in service +/- 16 hours a day that take a lot of cars off the street and out of parking spaces. I’m 10 miles out of town, but I’m injured and can’t drive for another 4 weeks. My wife is getting awfully sick of driving me to PT.
Uber, Lyft and taxicabs already make this sort of thing possible.
[quote=“Tride, post:1356, topic:798499”]can’t drive for another 4 weeks. My wife is getting awfully sick of driving me to …
[/quote]
The HUGE market for a true self driving car will be families with kids. Send your car to take the kid to her music lesson, then return to your office.so you can work a steady job, without asking the boss to leave work early…
True, but I think there would be a huge market for a commoditized product instead of looking at reviews of the driver. I don’t do Lyft too much and haven’t had a bad experience yet, but I’ve definitely had sketchy taxi rides. Plus, some people (myself included) would prefer to be alone in a car.
I’m really skeptical that a self-driving car is going to make the economics of this much better. Also, how old does your child need to be to get sent somewhere alone in care of a robot?
Tesla has a supercomputer, called Dojo, which they use to train the models. The training data comes from Tesla cars participating in the FSD beta program. Once the model has been trained, then each car has an onboard self driving computer which uses the output of the model to interpret what is in the cameras, and respond (somewhat) appropriately.
The vision system is surprisingly good in the rain. The cameras and vision system do an excellent job at picking out road lines in the rain with lots of glare. Times when it is difficult for me, the car is doing fine.
In the rain, the system will say it has detected bad weather, but continues to work. Similar when it is just falling snow. In my opinion, by the time the system refuses to run, it isn’t safe for a human to be driving either, because visibility is too low. Of course humans continue to drive in those conditions.
Snow on the road is a different matter. On completely white roads it can’t pick up lines. On roads with black pavement in the tire ruts, but white snow elsewhere, it won’t be able to figure out where the lines are, or can get confused.
I don’t have much experience with it in extremely heavy fog. In lighter fog, it seems to do fine, but that’s conditions where near visibility is fine.
ETA: One additional note. When the system used radar, snow would break things, because snow stuck to the front of the car would block the radar. The front cameras get swept by the wipers, and have a defroster, so they stay clean when there is precipitation.
I cannot imagine checking the reviews of an uber/lyft driver. They’re all good enough or they’d have been fired already.