Uber Says Waymo Autonomous Vehicles Outperforming 99% of Human Drivers in Austin
- Reference: 0177366367
- News link: https://tech.slashdot.org/story/25/05/07/1445228/uber-says-waymo-autonomous-vehicles-outperforming-99-of-human-drivers-in-austin
- Source link:
He cited the performance to "Waymo's safety record and rider experience coupled with Uber's scale and reliability." Uber has rapidly expanded its autonomous vehicle operations, reaching an annual run-rate of 1.5 million mobility and delivery AV trips across its network. The company plans to scale to hundreds of vehicles in Austin in the coming months, while preparing for a launch in Atlanta by early summer. Khosrowshahi said that autonomous vehicle technology represents "the single greatest opportunity ahead for Uber."
[1] https://s23.q4cdn.com/407969754/files/doc_earnings/2025/q1/transcript/Uber-Q1-25-Prepared-Remarks.pdf
And after wide straight roads with 90deg turnings. (Score:4, Interesting)
Why not see how well it does in London, Paris, Rome or Delhi in the rush hour with narrow streets, drivers often ignoring the rules, having to push in from a turning as no one will let you in and frequent use of headlights to signal others (can the AI understand that yet?). It'll be lucky to make it to the destination at all in any reasonable time.
Re: (Score:2)
Same reason it was only SF and Phoenix to start, those were the only cities to create openings in the laws that would allow them to stop operating. As far as I know places like those cities are not allowing Waymo type systems to pilot out much less run service in.
I think VW's Moia AV system is testing out in Germany though, it's just a wholly different regulatory environment.
You are right though in that transiting EU cities with their tight roads is a new challenge but I have no doubt Waymo will get there
Re: (Score:2)
> Same reason it was only SF and Phoenix to start
Also those cities don't have weather, to speak of. Would love to see one of those things operate in the wake of a nor'easter when lanes markings, crosswalks, road signs, and curbs are all invisible and you need to navigate around plows and people walking in the street because sidewalks are obstructed.
Re: (Score:2)
Yeah when I have taken them in SF they worked well but I was thinking how it will feel when these are barreling down the highway at 75mph and you look at nobody in the driver seat. No reason it shouldn't work though.
The weather point ties into that ethos though, gotta be able to handle easy weather before you think about introducing the system to inclement weather. I think the drip by drip method here is obviously paying off than trying to tackle everything at once like Tesla is trying.
You can see that in
Who are they comparing to? (Score:2)
Are they comparing to humans in modern vehicles with proximity sensors and all the safety equipment or are they comparing to people in old vehicles with huge blind spots?
Re: Who are they comparing to? (Score:1)
They're comparing them to Austin drivers. Which is a low bar. Just by using the turn signal you're outperforming most of Texas.
Re: (Score:2)
I can’t find the article but I was reading about how two Waymo cars caused a huge traffic jam in Austin. Two Waymo cars met at a four way stop intersection that had a lot of pedestrian traffic. With that scenario, both cars tried to yield to the other Waymo car while traffic backed up for miles. Complicating that is any pedestrian traffic would cause the car to reset whatever decision it decided to make.
Re: (Score:2)
They are “in it to win it”, so they would utilize whichever category of human-driver data gives them the best results in comparison. This is marketing and business, not scientific control group in a well-designed experiment.
Tone down the bragging (Score:2)
This wreaks of hubris, when you brag like this and there's even one accident .. the whole thing will be in jeopardy. Don't ruin the party for everyone.
The 1% (Score:2)
Worry not, this is slashdot.org. You will soon see where all of the 1% of drivers that are better are hanging out all day.
So? (Score:2)
This really isn't the concern. There are far greater concerns about the cost of operating AVs and the need for these companies to eventually recoup their investments. Once these things have reached a certain critical mass, these companies will start lobbying governments to ban human drivers in their markets of operation (mainly cities for now), under the guise of public safety, in order to establish their monopoly on transportation.
Doubt (Score:4, Insightful)
I had a waymo use a left turn lane to pass stopped traffic on the left and jump two lanes to the right, so no lol.
Re: (Score:2, Flamebait)
You ever saw a human driver do something stupid too?
Remember that the goal isn't perfection - its performance better than human drivers.
Re: (Score:2)
> You ever saw a human driver do something stupid too?
> Remember that the goal isn't perfection - its performance better than human drivers.
... better than American drivers.
Re: (Score:2)
Have you ever driven in Italy?
Re: Doubt (Score:2)
Being better than human drivers means not making any mistakes a human driver would never make.
Re: (Score:2)
Considering some of the crap I've seen people do that's a very low bar to clear.
Remember; the automatic car doesn't have to be better than the BEST person. It has to be better than the AVERAGE person.
Re: (Score:2)
The AVERAGE person does not utilize the left-turn pocket as a passing lane!
Re: (Score:2)
Anecdote: I have observed, at least in this area, that many drivers will quite easily use *any* available lane to pass if they think it will save them five seconds of their drive. This includes turn lanes, highway merge lanes, slip roads, etc. On a regular basis (especially on the Interstate highways) I also see people changing two or more lanes, frequently without signaling. The probability of this happening increases with certain makes and colors of car. A red Tesla, for example, has a much higher than
Re: Doubt (Score:2)
Then it's pretty sad they haven't cleared that bar yet.
Re: (Score:2)
> Being better than human drivers means not making any mistakes a human driver would never make.
Not necessarily. Because these systems approach problems in a different way, their types of mistakes may be very different in nature than mistakes humans would make. The types of mistakes isn't really important. Its the rate of the mistakes, the severity, and the safety record over time.
If an AI can drive with a record that results in less traffic injuries and/or fatalities, and either fewer accidents or an accident total that is lower in monetary cost, then they're doing better than human drivers, even
Re: Doubt (Score:2)
Tell that to the pedestrian who got dragged because the self driving's cars sensors didn't detect them bouncing off the hood.
Re: (Score:2)
Yes, but they're saying that if you retire drivers for doing things like that, you'll be mistakenly retiring 99 humans for every replicant.