Hacker News Reader

Comment by labrador | Hacker News Reader

Based on anecdotal evidence, I strongly suspect these are remotely driven cars, not completely self-driving, which means there is a remote saftey operator ready to take over if the ride starts to go bad. "Self-driving" sounds sexier and is the ultimate goal, but if it were my company I'd probably lead with "remotely monitored by safety driver" to calm any fears people might have. I'm guessing the percentage of people who are willing to put their lives in the hands of this tech is rather small and the percentage of people who don't trust tech is rather large.

View on HN

Waymo has been pretty clear about this: they do have remote navigators/coaches who tell the cars what to do if it gets stuck/confused, but they're not directly driving the car, they're telling it where to go, like a navigator in the front passenger seat.

Remote operation is considered dangerous due to possible issues with network connection latency or stability. If it was actually happening, Waymo is big enough now (IIRC they said 50,000 paid rides each week) to where someone would've leaked such a secret.

> they do have remote navigators/coaches who tell the cars what to do if it gets stuck/confused, but they're not directly driving the car, they're telling it where to go

That sounds a lot like remote operation.

Waymo is notoriously tight-lipped about this. Look at the number of journalists and reporters who have, over the years, asked very basic questions, like:

1. How often do remote operators intervene?

2. How many miles are driven per intervention? How does this compare to FSD?

3. How much of a typical ride is remotely operated and not actually driven by the car itself?

Waymo never provides answers, and one can only imagine it is because they are not proud of the answers.

I really don't understand why they aren't transparent about this, but judging by the downvotes I'm getting it seems people really, really want to believe they are intelligent and independent self-driving machines. It's so much more futuristic with a "wow" factor that attracts press attention and future investement.

The downvotes are because your suspicions are entirely wrong. There’s no remote “driving” or “take over”. They wrote a blog post on how it works just a few weeks ago: https://waymo.com/blog/2024/05/fleet-response/

So I've learned but it's not reassuring because the safer option would be to give a remote operator the ability to take over

That would be the least safe option due to network latencies involved.

Drone operators in Ukraine seem to do fine and besides, I'm talking about having the option, not using it 100% of the time

Surely, it's obvious that Waymo operates in a different safety critical setting? It's a 5000 pound vehicle with pedestrians and other vehicles around it. You can't depend on a network connection to prevent injuries and deaths.

Alphabet has way more money to spend on it than Ukrainians. I was a professional driver in a couple of jobs and as far as I can tell Google Waymo doesn't care about the millions of Americans they want to put out of work, possibly by having cheap foreign labor drive their cars, so you see I can't let it go that easily

> Waymo never provides answers

Found this by searching "waymo stats"

https://waymo.com/blog/2023/12/waymo-significantly-outperfor...

That article does not answer any of the 3 questions I proposed.

It's an example of how easy it is to find some numbers they posted, countering that they "never" do

I'm not trying to answer your question because I don't want to put in the time with someone who seems to only want to be combative in conversation

Case in point. Plugging your ears and shouting does not invalidate the question. These are very basic questions that, if Waymo were at all confident about their technology, would not hesitate to answer.

And most of these are just simple numbers! Like stating, the number of miles per intervention is 10. Yet that number is nowhere to be found from Waymo's press relations department.

Having seen how the sausage is made, a lack of confidence in what the numbers show is not why the information is withheld. The numbers are highly relevant to competitors and anything that could potentially be used to build a negative media narrative is very carefully considered before release.

Cruise had their numbers leaked during the incident last year and they came up in several negative media pieces despite being fairly good overall.

Don't the HN guidelines have something about "assuming good faith"? I can't be arsed to check now because obviously the comment above is not in good faith.

It's assume the best possible interpretation and the other commenter has provided evidence to the contrary through multiple posts

The guidelines explicitly say "Assume good faith":

Please respond to the strongest plausible interpretation of what someone says, not a weaker one that's easier to criticize. Assume good faith.

The guidelines don't say anything about being free to be a dick when the other person is being a jerk. They even say not to:

Comments should get more thoughtful and substantive, not less, as a topic gets more divisive.

Re 2 question - Tesla doesn't provide any reliable stat on miles per intervention neither, so you would not be able to compare anyway.

> That sounds a lot like remote operation.

If you have a friend in the front passenger seat directing you where to go, are they operating the car?

I had someone remotely intervene on my last ride. A car was trying to parallel park on a narrow street and the waymo stayed put, despite there being plenty of room to go around. I heard a chime and a pop up on the screen said “we’re getting your waymo back on track” and it went around the car, like any sane driver would have done from the get go

That's reassuring to me. I don't know why Waymo doesn't reassure safety conscious people like me by being explicit about this. As it stands, as a computer programmer and professional driver at different times in my career, whose seen a lot of unforeseen bugs and weird driving situations, I'm not putting my life in the hands of this software.

https://waymo.com/blog/2024/05/fleet-response

Interesting. This is about a month and half old, so they are getting more transparent. I am not reassured though because based on my reading, it appears the fleet response can't actually take over and drive the car in an emergency. I double checked with Claude AI

Me: "does it appear the remote operator can take over and drive the car if needed."

Claude: "Based on my reading of the document, it does not appear that the remote operator (referred to as "fleet response agent") can take over and drive the car directly"

Double checking with an LLM is not the way to find the truth. They are more like creative assistants than fact checkers

> I'm not putting my life in the hands of this software.

If you were a professional driver, you know you put your life in the hands of people right? The average person is a pretty terrible driver. I’ll take a safely trained machine all day every day.

I'm a better and safer driver than Waymo. I trust myself over software. I also trust human professional drivers over software.

The data disagrees with you, significantly. Level 5 technology is only improving, and you’ll regress over time. A waymo never loses focus or gets distracted or tired in the way humans always will.

100,000 miles as a paid professsional driver and a couple 100k more for pleasure and I never once drove into a telephone pole. But keep talking about the mirage Level 5. I can think of many snap decisions I've had to make to prevent death or serious injury, including once when someone tried to commit suicide by jumping in front of me. I seriously doubt AI ability to outperform me in those situations. You're right, humans do regress, that's why I don't drive anymore. But no matter, Waymo's goal is to put millions of American drivers out of a job and they'll do it no matter what I say. Consider me shouting from the void as I disappear along with my working driver brothers and sisters, who Alphabet couldn't care less about.

Waymo recalls driverless cars to make them less likely to drive into poles https://www.cnn.com/2024/06/13/business/waymo-recalls-driver...

This conversation is invisible since everyone downvoted it so I don't know why you bothered to comment

They are only taken over in certain edge cases, and even then the remote operator is only placing the next destination, he is not actually remotely "driving" the vehicle. More on this here: https://waymo.com/blog/2024/05/fleet-response/

>> Fleet response can influence the Waymo Driver's path, whether indirectly through indicating lane closures, explicitly requesting the AV use a particular lane, or, in the most complex scenarios, explicitly proposing a path for the vehicle to consider. The Waymo Driver evaluates the input from fleet response and independently remains in control of driving. This collaboration enhances the rider experience by efficiently guiding them to their destinations

In my reading this is trying very hard to put a self-driving spin on an interface that allows a human operator to set waypoints for a robot to follow. This should at best be called "human remote-assisted self-driving", rather than "self-driving".

Btw, NASA uses a similar setup for its Mars rovers with autonomous navigation capabilities, like Perseverance. There was a report that described how humans on Earth can guide Perseverance to plan paths avoiding areas with potentially deep sand where the rover can get seriously stuck. The human operators draw the limits of what seems to them to be the safe area and the rover's onboard planner than plans a path through the safe area. That's still the farthest that autonomous navigation can go; pun not intended.

Ref: Autonomous robotics is driving Perseverance rover’s progress on Mars

https://www.science.org/doi/10.1126/scirobotics.adi3099

> In my reading this is trying very hard to put a self-driving spin on an interface that allows a human operator to set waypoints for a robot to follow.

I can 100% confirm you're reaching way too far to find excuses to indicate Waymo isn't self-driving.

OK what the bloody fuck. What is it with the tone of these discussions? You can 100% confirm I'm reaching... the hell is the matter with you? Can't you disagree like an intelligent adult?

Fucking internets.

(lectures about tone while getting personal and swearing and not asking any legible question)

Come with curiosity! Ex. "How would one be able to 100% confirm Waymo isn't just being live monitored and incremental way points set?"

(I'll leave it to you ask if you want, you seem upset and I worry you'll think the answer is obvious and that's not what you meant and I'm creating strawman. I want to deescalate and make you un-angry!)

Well done for de-escalating but I can't see what "curiosity" you expect when you "confirm with 100% certainty" anything you want. Curious discussion doesn't mean you pretend the other person is not a fool just for the sake of proving they are.

Oh okay!

Wouldn't the fear just become that not enough safety drivers will always be available or that the internet connection they use to "drive" the car remotely would fail?

It's a taxi that you don't have to risk having a conversation with a stranger in. Outside of that it seems that all other benefits are marginal or well crafted illusions.

Anyways, I'm sure I'm just being a pessimist, it's not as if large monopolistic companies with no human customer service have failed us significantly before. I mean, I trust them with my data completely, so why not my life, too?

If it loses remote monitoring I'd rather it would pull over and park rather than barelling down the steep hills of San Francisco without a safety monitor.

Reply on news.ycombinator.com