Hacker News

Waymo exec admits remote operators in Philippines help guide US robotaxis

102 points by anigbrowl ago | 47 comments

svat |next [-]

Waymo's article from 2024 about these operations (“much like phone-a-friend”): https://waymo.com/blog/2024/05/fleet-response/

svat |root |parent [-]

The blog post includes short videos that illustrate what the fleet response does:

• (29 seconds long) https://www.youtube.com/watch?v=T0WtBFEfAyo

• (59 seconds long) https://www.youtube.com/watch?v=elpQPbJXpfY

neuronexmachina |next |previous [-]

This was from congressional testimony this past week by executives from Waymo and Tesla, video and automated transcript here: https://www.c-span.org/program/senate-committee/tesla-and-wa...

i7l |next |previous [-]

With roughly 400 ms round-trip latency does that even make sense? A car travelling 50 km/h manages to advance by 5 m in that time.

SilverElfin |next |previous [-]

I think everyone knew this and is comforted by it. I’d be concerned if there weren’t humans ready to guide or take over. The company we should all be concerned about is Tesla, and their irresponsible way of falsely advertising full self driving capabilities. Who knows what those robotaxis are capable of.

flutas |root |parent [-]

Waymo's remote drivers have literally caused accidents and we only know about it because journalists did digging. Waymo simply removed all details of the remote ops roles in it in the NHTSA reporting.

jfjfjfjffhfjfj |root |parent [-]

What journalists should I look for to see that

flutas |root |parent [-]

https://www.forbes.com/sites/bradtempleton/2024/03/26/waymo-...

The description there is

    In January, an incident took place where a Waymo robotaxi incorrectly went through a red light due to an incorrect command from a remote operator, as reported by Waymo. A moped started coming through the intersection. The moped driver, presumably reacting to the Waymo, lost control, fell and slid, but did not hit the Waymo and there are no reports of injuries. There may have been minor damage to the moped.
While the description in the official report to the NHTSA is (ID: 30270-6981)

    On January [XXX], 2024 at 10:52AM PT a rider of a moped lost control of the moped they were operating and fell and slid in front of a Waymo Autonomous Vehicle (Waymo AV) operating in San Francisco, California on [XXX] at [XXX] neither the moped nor its driver made contact with the Waymo AV.

    The Waymo AV was stopped on northbound [XXX] at the intersection with [XXX] when it started to proceed forward while facing a red traffic light. As the Waymo AV entered the intersection, it detected a moped traveling on eastbound [XXX] and braked. As the Waymo AV braked to a stop, the rider of the moped braked then fell on the wet roadway before sliding to a stop in front of the stationary Waymo AV. There was no contact between the moped or its rider and the Waymo AV. The Waymo AVs Level 4 ADS was engaged in autonomous mode.

    Waymo is reporting this crash under Request No. 1 of Standing General Order 2021-01 because a passenger of the Waymo AV reported that the moped may have been damaged. Waymo may supplement or correct its reporting with additional information as it may become available.

bryan_w |next |previous [-]

> “They provide guidance. They do not remotely drive the vehicles,” Peña told the Senate committee. “The Waymo vehicle is always in charge of the dynamic driving tasks, so that is just one additional input.”

therealpygon |root |parent [-]

This indicates a very misleading title. It sounds like they aren’t remote “operators” any more than I “operate” the tv station by deciding what channel to watch.

unsupp0rted |root |parent [-]

It's more like you operate the code by vibe-coding alone. You don't write it, but you do tell it "what", and if you desire "where" and even "how".

ravenstine |next |previous [-]

Having physical brains in the loop seems like a good thing.

unsupp0rted |root |parent |next [-]

For now that's true, because it's early days and very much a hybrid system. In a few years having human brains in the loop will be like adding more and more orangutans around the Operating Room table.

jjeaff |root |parent |previous [-]

I think some are wondering if these overseas employees are driving cars in the US without a US driver's license.

rcxdude |root |parent [-]

They aren't driving the cars. If someone is giving you advice from the backseat, do they need a driver's license?

rcxdude |next |previous [-]

This headline implies this assistance is a new revelation when the only part that might be new is that it's now also being outsourced to the Philippines.

tengbretson |next |previous [-]

I was once in a waymo stopped at a red light. Prior to the light turning green I felt a split second where the car's brake had been released, anticipating the change and then accelerating immediately when the light changed.

Since this experience I've just assumed all waymos have some warehoused human drone pilot actually controlling it.

akanet |root |parent |next [-]

Waymo remote operators cannot drive the car

slim |root |parent [-]

how do you know ?

galkk |root |parent |next [-]

Physics. Have you ever played a competitive/reaction based video game with high ping? It is very, very hard. And it’s a game, where there are many tricks to hide latency from you.

Cloud console shows pings between Google data centers in us-west and ones that are in proximity of Philippines around 160-200ms. Then you also have inherent lag of wireless connection itself. Then you have also connectivity from google’s data center to Philippines.

If you want remote driving in uncontrolled environment, you reasonably can expect only the same city/county operators.

I’m obviously uninformed, but I’d expect the remote operators job (from another country) to be like “car is safe to proceed, based on that picture that I see” or, in the worst case scenario, put some waypoints in the ui and let car drive on its own.

hackable_sand |root |parent [-]

Thankfully technology hasn't advanced since 2010!

keeda |root |parent [-]

Sure, technology has advanced by leaps and bounds, but we haven't quite figured out how to exeed the speed of light just yet ;-)

rcxdude |root |parent |previous [-]

It would be a certifiably insane model with the latency and failure modes involved, for a start.

Marsymars |root |parent |next |previous [-]

> I was once in a waymo stopped at a red light. Prior to the light turning green I felt a split second where the car's brake had been released, anticipating the change and then accelerating immediately when the light changed.

This seems like it would be fairly straightforward to program, if not for all lights, at least for a lot (e.g. say half) of lights.

thebruce87m |root |parent |next |previous [-]

There are many other possibilities such as the system having learned the timings or another vehicle in the fleet observing the lights turn red at the other part of the junction.

The least likely possibility is a person controlling the vehicle directly over a variable latency connection that may fail completely at any time.

thenthenthen |root |parent [-]

Behind all the new smart city tech I encountered here in Shenzhen and Shanghai are actually human operators (drones, cars, vending machines). You can find the job ads online.

thebruce87m |root |parent [-]

I’m sure there are, but direct remote control of throttle, brakes etc in vehicles is _hopefully_ not part of that.

I could see certain situations where it could be authorised when a vehicle is stranded and unable to operate autonomously at all due to an error, but it would have to be extremely slow speed with a full-stop failsale on connection drop or high-latency detection.

That said I bet there are some who do not consider the safety implications and “move fast and break people”

rcxdude |root |parent |previous [-]

Why would you assume this from that experience?

etyhhgfff |next |previous [-]

Its turtles all the way down.

OsrsNeedsf2P |previous [-]

Ok, and?

MBCook |root |parent |next [-]

Then they’re not really self driving are they?

bryan_w |root |parent |next [-]

The assistants don't have access to the gas pedal or steering wheel input. The car is the only thing actually piloting the car i.e. self driving

* Unless it gets super stuck, then a human drives out and gets into the physical driver seat and takes over

MBCook |root |parent [-]

Which is certainly better than just having a human remote drive.

But it’s still not the impression they’ve been giving. It’s been an impression of full automation (ignoring getting stuck) and if it’s not navigating on its own that’s disingenuous.

SR2Z |root |parent [-]

The car will always get stuck every now and then.

This approach has two benefits: it can be unstuck without sending out a physical driver and while collecting training data, and it efficiently lets m humans control n cars with a wide range of acceptable m and n values.

It's intended for the ratio of m:n to smoothly shrink as the software gets better, but m will always be greater than zero.

lysp |root |parent |previous [-]

My assumption is they can provide input such as "you should make a left/right lane change" to get out of a "stuck" location.

So when the car's systems prevent it from taking a specific action, they can override it for a single instance.

anigbrowl |root |parent |next |previous [-]

People are somewhat surprised about this work being farmed out to the Philippines as opposed to being done by Americans. I'm pretty sure you don't need me to explain this, though.

jjeaff |root |parent |next |previous [-]

Do they have driver's licenses from a US state?

TheDong |root |parent |next [-]

Waymo has a blog post on what these remote assistance people do: https://waymo.com/blog/2024/05/fleet-response/

They effectively are answering questions like "is this road closed", or "is the object in front of me a solid object or a weird shadow".

These are not the sort of questions that US driver's license is really related to, it's not things like "can I legally turn right on red at this intersection".

Do we require a driver's license to solve Google reCapture questions like "what squares have a bike in them"? Because the waymo stuff is closer to image classification than driving.

naveen99 |root |parent [-]

I think if you took a buddy with you to the drivers license test in America and asked your buddy these questions during the test. You and your buddy are both failing. Unless test was in India over tea and not in a car.

TheDong |root |parent [-]

I guess you're saying that because a waymo car can't walk into the DMV and get a license, it shouldn't be on the road? (which of course it can't, you have to have a legal human identity to get a normal driver's license, and we don't let cars have humanity currently)

Driver's licenses are legal constructs. The DMV certifies self-driving cars as able to drive on the road differently, and sure, those two different processes are different.

I really don't get the point you're trying to make here.

throw4432334 |root |parent |next |previous [-]

If they have a driver’s license from the Philippines, then it should be enough. Just like foreign tourists can rent and drive cars in the US without needing a US state driver’s license.

victorbjorklund |root |parent |next |previous [-]

Do you think American tourists have to take a Filipino driver’s license when they rent a car on vacation?

whatever1 |root |parent |previous [-]

Is Waymo software licensed by a dmv ?

Grimblewald |root |parent [-]

Yes. Something you should intuit, and is eaisly confirmed with a quick search. It is licensed to drive and the conditions underwhich it may do so are clearly stipulated. If it didnt require a license elon would have his deathtraps littering roadsides with mangled flesh and steel everywhere. Perhaps ask yourself why you asked such a misguided question and consider what you can do different in your cognitive patternd to avoid it in the future.

whatever1 |root |parent [-]

So if it is licensed, the entire stack is licensed even if there are dogs pressing buttons on the other side of the planet.

Elons cars are already driving by themselves against the conditions they are licensed for, causing accidents and the liability falls to the drivers.

Aka, licensing means absolutely nothing. Specially today.

Grimblewald |root |parent |previous [-]

[flagged]

TheDong |root |parent [-]

You've imagined a scenario around remote drivers having access to the internal microphones.

Waymo tells you explicitly that all the microphones inside the car are off unless you press the button to call rider support yourself.

If you'd ever ridden in waymo, perhaps you'd recall them telling you that the first time you rode one.

> if you can't think of more perhaps you should keep your comments out of the discussion, because at present you've contributed nothing but ignorance.

You really shouldn't end your comment with that if you're not going to read up on whether a hypothetical scenario you've imagined up is ignorant or not.

Grimblewald |root |parent [-]

have you read terms and conditions? They can access video under near any circumstance, like wanting to check the general cleanliness of the car etc. audio is a bit different, or so they say, but when it comes to companies like this can we really trust what they say? They have an awful habit of lying an awful lot when it comes to data and privacy. Tesla for example recently got in trouble for not really doing as they should regarding sensors, if you recall. Waymo is several leagues above tesla in terms of general professionalism, however, I don't know if they professional enough to not do things they shouldn't, or under-employ folks in charge of implementing barriers to abuse etc.