4 ms·

Waymo works in specific areas only though. As soon as you get out of San Francisco, Phoenix, Los Angeles, or Austin, the car wouldn't have self-driving capabili

Waymo works in specific areas only though. As soon as you get out of San Francisco, Phoenix, Los Angeles, or Austin, the car wouldn't have self-driving capabilities. I suspect it'll be some kind of AI-powered assistant that will take self-driving vehicles to people, like [1]. You retrofit a device onto your car, which'll hook to the car's sensors (as well as bring it's own sensors, like camera) and get a better experience in self-driving. Still nowhere near reliable to just stop paying attention though. Hopefully, they'll make reliable self-driving vehicles within the next 10 years or so... [1] https://comma.ai/ https://comma.ai/

hiddencost 2y ago
Atlanta coming soon. They scaled from 10k to 100k per week in a year. They've been operating successfully on highways for months now. Once they can do highways, a huge fraction of the US will open up. And generally speaking, driving outside of cities is quite a bit more easy. They seem like they're moving very quicky to massive deployment. Comma is still just driver assist. Custom vehicles that can strip out all the complexity of a steering wheel will reduce costs pretty aggressively.
- AStonesThrow 2y ago
  Any taxi service will rely on several elements which limit their service area: 1. Availability. How many vehicles are in an area, that can respond to dispatch? 2. Demand. Can't have too many idle vehicles scooting around and taking up space. 3. Service. Must establish depots where the vehicles can return for charging, cleaning and maintenance. 4. Support. Waymo Support is a tap away, and if a vehicle gets particularly jammed up, they will dispatch a technician in their own vehicle. Support can also monitor telemetry and push commands to give nudges (but cannot remotely drive a car). So with the current model, I do not envision Waymo service areas expanding beyond the reach of their depots, or offering long-haul trips between urban centers.
tracker1 2y ago
Even then... I happen to live in Phoenix, and locally wouldn't mind self-driving, but I like to drive on road trips as well. So electric first and limited self-driving as an option for my home area would be really nice, at least for me specifically. Of course, I don't have enough trust in self-driving to even use the things. I know a lot of people love them, and have a couple acquaintances working on self-driving stuff. Just me being a bit hesitant with new tech in general. I have similar feelings on a lot of the smart home stuff, well justified after companies dropping, deprecating hardware or outright dying. Maybe in another decade we might see better options all around.
refulgentis 2y ago
Comma? Geohit's scarce attention already flitted away, quite some time ago.
everforward 2y ago
Self-driving being affordable enough for people to own a self-driving car is going to rely on either massive price reductions on LiDAR sensors (an order of magnitude or 2) or a Tesla-style approach without LiDAR working out. I worked in the space 5ish years ago, and the LiDAR sensors were monstrously expensive. They were the most expensive component we put in the car; if I'm remembering right, a single LiDAR sensor cost more than the car itself and there were half a dozen or so of them on the car. I want to say it was ballpark $500k to build out one of the cars, with the majority of that being LiDAR sensors followed by the servers in the car. Comma doesn't appear terribly close to cracking that based on their features. It's certainly a neat project, but I wouldn't put them much closer to cracking self-driving than someone just starting out today.
- ra7 2y ago
  Your numbers are way off. Waymo reported 5 years ago their Jaguars cost between $120k-$150k. That was after reducing their LiDAR cost by 90% from $75,000 to ~$7500. Their newly announced 6th gen sensors are even cheaper. LiDAR sensors are no longer expensive. We are seeing automakers including sub $1000 LiDARs. Granted they only offer L2 ADAS, but the rapid price cratering means it's only a matter of time for consumer-owned self driving cars.
  - everforward 2y ago
    That's likely still too expensive; sub-$1,000 would probably work. The currently deployed Waymos have 5 $7,500 (at cost) LiDAR sensors in them. That's $37.5k, almost exactly $10k short of the average new car sale price. To be clear, I'm talking about the point where normal people get new cars that have self-driving. It's affordable for wealthy people now, and was available to the super wealthy before that. I think the whole package is going to have to get somewhere in the $5k-$10k range before it's "normal" to have a self-driving car and be able to afford the maintenance. Even those super cheap LiDARs may be over $10k installed by the time you add the other sensors, compute, labor, etc. I have no idea what the lifetime of a LiDAR sensor is, so maintenance is a concern too. I think we'll get there, but it'll be a while after robotaxis become a thing. The news seems to think LiDAR sensors will settle around $200, which would definitely be affordable.

hiddencost 2y ago

Atlanta coming soon. They scaled from 10k to 100k per week in a year. They've been operating successfully on highways for months now. Once they can do highways, a huge fraction of the US will open up. And generally speaking, driving outside of cities is quite a bit more easy. They seem like they're moving very quicky to massive deployment. Comma is still just driver assist. Custom vehicles that can strip out all the complexity of a steering wheel will reduce costs pretty aggressively.

AStonesThrow 2y ago
Any taxi service will rely on several elements which limit their service area: 1. Availability. How many vehicles are in an area, that can respond to dispatch? 2. Demand. Can't have too many idle vehicles scooting around and taking up space. 3. Service. Must establish depots where the vehicles can return for charging, cleaning and maintenance. 4. Support. Waymo Support is a tap away, and if a vehicle gets particularly jammed up, they will dispatch a technician in their own vehicle. Support can also monitor telemetry and push commands to give nudges (but cannot remotely drive a car). So with the current model, I do not envision Waymo service areas expanding beyond the reach of their depots, or offering long-haul trips between urban centers.

tracker1 2y ago

Even then... I happen to live in Phoenix, and locally wouldn't mind self-driving, but I like to drive on road trips as well. So electric first and limited self-driving as an option for my home area would be really nice, at least for me specifically. Of course, I don't have enough trust in self-driving to even use the things. I know a lot of people love them, and have a couple acquaintances working on self-driving stuff. Just me being a bit hesitant with new tech in general. I have similar feelings on a lot of the smart home stuff, well justified after companies dropping, deprecating hardware or outright dying. Maybe in another decade we might see better options all around.

refulgentis 2y ago

Comma? Geohit's scarce attention already flitted away, quite some time ago.

everforward 2y ago

Self-driving being affordable enough for people to own a self-driving car is going to rely on either massive price reductions on LiDAR sensors (an order of magnitude or 2) or a Tesla-style approach without LiDAR working out. I worked in the space 5ish years ago, and the LiDAR sensors were monstrously expensive. They were the most expensive component we put in the car; if I'm remembering right, a single LiDAR sensor cost more than the car itself and there were half a dozen or so of them on the car. I want to say it was ballpark $500k to build out one of the cars, with the majority of that being LiDAR sensors followed by the servers in the car. Comma doesn't appear terribly close to cracking that based on their features. It's certainly a neat project, but I wouldn't put them much closer to cracking self-driving than someone just starting out today.

ra7 2y ago
Your numbers are way off. Waymo reported 5 years ago their Jaguars cost between $120k-$150k. That was after reducing their LiDAR cost by 90% from $75,000 to ~$7500. Their newly announced 6th gen sensors are even cheaper. LiDAR sensors are no longer expensive. We are seeing automakers including sub $1000 LiDARs. Granted they only offer L2 ADAS, but the rapid price cratering means it's only a matter of time for consumer-owned self driving cars.
- everforward 2y ago
  That's likely still too expensive; sub-$1,000 would probably work. The currently deployed Waymos have 5 $7,500 (at cost) LiDAR sensors in them. That's $37.5k, almost exactly $10k short of the average new car sale price. To be clear, I'm talking about the point where normal people get new cars that have self-driving. It's affordable for wealthy people now, and was available to the super wealthy before that. I think the whole package is going to have to get somewhere in the $5k-$10k range before it's "normal" to have a self-driving car and be able to afford the maintenance. Even those super cheap LiDARs may be over $10k installed by the time you add the other sensors, compute, labor, etc. I have no idea what the lifetime of a LiDAR sensor is, so maintenance is a concern too. I think we'll get there, but it'll be a while after robotaxis become a thing. The news seems to think LiDAR sensors will settle around $200, which would definitely be affordable.