Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

A look under our trunk: what's in our compute

waymo.com

Post reply on HN
Print view View original

111–116 of 116 posts

Re: A look under our trunk: what's in our compute

#111

Post by tjwebbnorfolk »

fwip wrote: post #40
tjwebbnorfolk wrote: post #33

Earlier quoted context omitted.

Maybe it's different where you live, but 95% of the buses I see driving around here are completely empty.

Are you near the city center, or out near the suburbs? It's common for a bus to be crowded on one part of the route and nearly empty at the end.

I'm a half-mile from downtown. Yes I'm sure it's crowded downtown, but most of the mileage is not burned downtown, it's burned by driving way out of town and bringing in 0 or 1 or 2 people until it fills up right near the center.

Buses appear way more efficient than SOVs until you actually spend a couple minutes of critical thinking to realize they're not

Re: A look under our trunk: what's in our compute

#112

Post by tjwebbnorfolk »

fragmede wrote: post #102
cyberax wrote: post #44

Earlier quoted context omitted.

And this is a fundamental issue with transit. It _has_ to travel as mostly empty on avereage to be useful.

...no? In the heart of downtown, with people riding in both directions, I don't see how it follows that it has to mostly empty to be useful. If it's full you can't get on so that's not useful, but anything less than totally full should still be useful.

Because most of the mileage is not burned downtown, it's burned by having a bunch of empty buses drive way out of town and pick up in 0 or 1 or 2 people until it fills up right near the center.

Re: A look under our trunk: what's in our compute

#113

Post by fwip »

tjwebbnorfolk wrote: post #112
fragmede wrote: post #102

Earlier quoted context omitted.

...no? In the heart of downtown, with people riding in both directions, I don't see how it follows that it has to mostly empty to be useful. If it's full you can't get on so that's not useful, but anything less than totally full should still be useful.

Because most of the mileage is not burned downtown, it's burned by having a bunch of empty buses drive way out of town and pick up in 0 or 1 or 2 people until it fills up right near the center.

If you look at average ridership over time, you'll see that the times when it has 40-60 people on it give an average passenger density much higher than cars, even after averaging in the longer empty miles. (40 passengers is a pretty comfortable, not crowded bus.) Simplifying, but a mile of 40 people, averaged with 10 miles of 1 person, is still an average of 5 people. And those are deliberately bad numbers. As I understand it, average bus occupancy in Seattle is around 10 or 15 people per bus, during non-commuting hours. (Commute is often quite crowded).

Re: A look under our trunk: what's in our compute

#114

Post by NicuCalcea »

cyberax wrote: post #91
NicuCalcea wrote: post #88

Earlier quoted context omitted.

Surely a person driving 18 people will emit less than each person driving themselves? I can see how dollars can be a proxy for macroeconomic stuff, for example by comparing countries, but it doesn't make much sense for comparing individual units of transport. I don't understand why we have to do napkin maths to guess what's more polluting. There is countless research showing buses emit less than individual cars per p…

> Surely a person driving 18 people will emit less than each person driving themselves? No. A person driving themselves does not create any _incremental_ CO2 emissions. They don't need a dedicated full-time position to do the driving. > There is countless research showing buses emit less than individual cars per passenger per km, self-driving or not. Care to share it? I actually read through most of highly cited arti…

The very source you are providing is pretty clear to me. For cars, only hybrids and EVs are more efficient, and that is because it's comparing electric passenger cars with all buses (most of which still run on fossil fuels). There is no data for electric buses, but it's pretty intuitive to me that they would emit less per passenger kilometre than an electric car.

Re: A look under our trunk: what's in our compute

#115

Post by novia »

ButlerianJihad wrote: post #86
novia wrote: post #71

> Redundant: Because there is no human to take over, proactive safety and redundancy are natively built in. Our compute is designed like two independent engines. While they normally operate as one unit running full parallel workloads, if one experiences a fault, the other seamlessly takes over. uhhh i've been in a situation where a remote human took over before it's rare but it can happen (or at least it could 1.5 ye…

Waymo has zero "remote human drivers". It is literally impossible to "drive" a Waymo by remote commands. The Waymo agents are constantly monitoring the fleet for exceptions. When an exception is raised, a human reviews the footage and the situation. They can send commands to the car and suggestions about which way to branch or a course of action to choose. So ultimately, it is the Waymo Driver doing the actual drivin…

> Waymo has zero "remote human drivers". It is literally impossible to "drive" a Waymo by remote commands

And yet?

A guy phoned into our waymo with a thick indian accent and told us to hold on just one moment and steered us around the obstacles and then let the waymo get back to it.

I know I'm just some person on the internet but this is what happened.

Edit: I think I was wrong and the parent comment is right, although this is indeed what it seemed like happened to me, that human on the other end of the line was most likely a member of Fleet Response. More details here: https://waymo.com/blog/2024/05/fleet-response/

Also I was confused because now-a-days the word "agent" in my circles more often than not no longer refers to a human.

Re: A look under our trunk: what's in our compute

#116

Post by astrange »

nxc18 wrote: post #105
MostlyStable wrote: post #89

Earlier quoted context omitted.

Who is this audience that thinks unasked-for and un-controllable felonies is a good message, or who wants to fund that?

Everyone who is buying into RSI/AGI hype. That hype is the only thing sustaining this investment cycle. The bubble will pop as soon as people realize “AI” is a normal technology.

I don't think there's anyone who wants to fund "committing felonies", for obvious reasons.
Post reply on HN