Waymo Blog

Feel free to check my math, but Waymo's 2 accidents in 1,000,000 driverless miles (https://blog.waymo.com/2023/02/first-million-rider-only-miles-how.html… is higher than the rate for human drivers in any of the years from my quick NHTSA stats search. They've had no fatalities, which is great (and certainly better than other "self-driving" car options)! But it's hard to make really good comparisons as long as Waymo is off by a factor of millions of miles.

Waymo is currently the only SAE Level 4[0] self-driving car in use, right? The wheel must not be used. Even though it is limited to two cities. Their technology seems an order of magnitude above the rest, although factoring in its cost could make it look worse. (Can it turn a profit, when including the development costs?) Tesla FSD is at best SAE level 3 (can need human fallback), and at worst level 2 (needs constant human monitoring). [0]: https://blog.waymo.com/2020/10/revealing-our-approach-to-safety.html

Disclosure: I work at Waymo. This is big news, and I encourage you to apply if you're in San Francisco. For the HN crowd, I'd also recommend last week's blog post [1] which includes some more technical material on "how" we're driving. [1] https://blog.waymo.com/2021/08/MostExperiencedUrbanDriver.html

Other founder here! For a high level overview of this framing of the problem, I recommend reading this Waymo blog post [1]. One nice feature is that by using embeddings produced by a user's model, which has been trained in the context of their domain, we can do this sort of smart sampling in domains we've never seen before. Embeddings are also naturally anonymized, so we can do this without access to a user's potentially private raw data streams. [1] https://blog.waymo.com/2020/02/content-search.html

There seems to be a lot of misconception and misinformation about mapping. People (mostly influencers associated with Tesla) constantly cite it as a major bottleneck for scaling. And there are barely any technical details about the mapping process. Are you able expand on the following: - Is the initial mapping effort very high and expensive? How long does it take, on an average, to fully map a city the size and complexity of San Francisco? - How much of the mapping process is automated? - How are the maps kept up-to-date? We know from Waymo's blog post a while ago [1] that the cars map continuously. But is human intervention required to verify an update (like a new construction zone that was encountered)? - How much of the country have you already mapped? Waymo says they are testing in 25 cities, so I'm assuming the number is at least 25 cities. - What does a vehicle do if it detects a mismatch of road conditions vs the onboard map? (I already know the answer, but this is one of the most frequent points of misinformation where people say it stops working) Anything you can share without revealing your secret sauce is appreciated! [1] https://blog.waymo.com/2020/09/the-waymo-driver-handbook-mapping.html

"I can tell you with confidence they have not really put effort into weather" Ah, the Internet, where people who know absolutely nothing about a topic make bold assertions about it with 100% confidence. https://blog.waymo.com/2022/11/using-cutting-edge-weather-research-to-…

Waymo has even developed custom localized climate models for San Francisco: https://blog.waymo.com/2022/11/using-cutting-edge-weather-research-to-…

Waymo buys renewable energy for charging https://blog.waymo.com/2023/03/paving-way-toward-fully-electric-ride.h…

This person just has no idea how AVs work and thinks an openly shared approach to AVs is a conspiracy https://medium.com/cruise/hd-maps-self-driving-cars-b6444720021c https://blog.waymo.com/2019/09/building-maps-for-self-driving-car.html https://zoox.com/journal/putting-our-robots-on-the-map/ When you're ignorant on a topic, it's not surprising that every fact you come across might end up being misconstrued as discovering some well-hidden secret.

Waymo does pose estimation to help determine pedestrian intent and react accordingly [1]. [1] https://blog.waymo.com/2022/02/utilizing-key-point-and-pose-estimation…

Waymo was collecting data before Tesla, and switched to simulated training a long time ago because it's more effective. The self driving AI can gain 100 years experience in just 1 day using simulation: https://blog.waymo.com/2020/04/off-road-but-not-offline--simulation27.…

I think getting a license for full commercial services in California can be an inflection point. Waymo is currently serving outside SF downtown (edited: outside, not in) without too much hiccups, so I don't expect significant technical blockers for future expansion into the entire bay area and then the LA and its metro. Looks like this is their plan as well. https://blog.waymo.com/2022/10/next-stop-for-waymo-one-los-angeles.htm…

I think this blog goes into more detail: https://blog.waymo.com/2021/06/beyond-the-bike-lane.html

Why is this considered a big stride when Waymo is doing it for 6 months already?https://blog.waymo.com/2021/08/welcoming-our-first-riders-in-san.html

Waymo built a bespoke, 113 acre facility for this very purpose back in 2017[0][1]. Full disclosure: Google employee, working on nothing to do with Waymo. [0]: https://blog.waymo.com/2020/09/the-waymo-drivers-training-regime.html [1]: https://www.wired.com/story/google-waymo-self-driving-car-castle-testi…

They also train in simulation, where they modify real scenarios to run through variations and edge cases. [1] As of last year, they'd driven 15 billion miles in simulation - which is more than a typical person will drive in a million years. Still a ways to go to get to a billion though. [1] https://blog.waymo.com/2020/04/off-road-but-not-offline--simulation27.…

> Once Waymo can replace some human drivers some of the time, things will start changing very quickly. But that happened 1y+ ago [1][2] without much changing since? [1] https://www.theverge.com/2019/12/9/21000085/waymo-fully-driverless-car… [2] https://blog.waymo.com/2020/10/waymo-is-opening-its-fully-driverless.h…

What is your definition of "replace"? Waymo operates a driverless taxi service in Phoenix. Sign ups are open to the general public. IMO this counts as replacing some drivers as there is less demand for taxi service in the operating area. https://blog.waymo.com/2020/10/waymo-is-opening-its-fully-driverless.h…

There's of course legal ways to do similarly. I thought this was a good example of John Krafcik's resignation which to a certain extent meets the same aim: Official resignation letter: https://blog.waymo.com/2021/04/capstone-of-my-career-email-from-john.h… "Interpreted" letter: https://gist.github.com/ChuckM/ff5fc8c800c7fe9160483b68ec45aac9

You can, today, go to Pheonix Arizona, download the Waymo One app, and summon a full self driving vehicle. This happened silently at the end of 2020, and was overshadowed by the ongoing pandemic doom and gloom. https://blog.waymo.com/2020/10/waymo-is-opening-its-fully-driverless.h… On the subject of danger. Literally everything we do is dangerous to some degree. Everything. Self-driving does not need to be perfect, but only better than human drivers. We're likely already at that point with Waymo vehicles. On the subject of automation. Humanity is on track to have the computational power to perform whole brain emulation decades before 2100. Even if we don't solve the general AI problem through other pathways, we will solve it through this path. Ethical problems aside, once this is achieved, everything will be automatable at a human level of competence.

Not anymore. As of a couple months ago its been opened up. anyone (in the service area) can use it. Relevant blog post: https://blog.waymo.com/2020/10/waymo-is-opening-its-fully-driverless.h…

"I think the really AI winter will come once everyone will realize that self driving cars are not coming anytime soon." And yet : https://blog.waymo.com/2020/10/waymo-is-opening-its-fully-driverless.h… I guess "the really (sic) AI winter" may have to wait! I commend the power of reading (and grammar) to people who write blogs...

Waymo did 6.1M miles in Phoenix, Arizona alone. Waymo did a lot of simulated driving, they claim about 200M miles a day: https://blog.waymo.com/2020/04/off-road-but-not-offline--simulation27.…

Google engineer not working on Waymo, opinions are my own. I think "driverless" here is different from "self-driving", i.e. 'Waymo uses “driverless” here to refer to operations in which the ADS controls the vehicle for the entire trip without a human driver (whether in the vehicle or at a remote location) expected to assume any part of the driving task.' at page 5 of that report. I think this driverless mode is only available to public on Oct 8th [0]. [0]: https://blog.waymo.com/2020/10/waymo-is-opening-its-fully-driverless.h…

Waymo uses ML/AI extensively. Here's a blog post from earlier this year about how they're forgoing CNNs for a "hierarchical graph neural network." https://blog.waymo.com/2020/05/vectornet.html

https://blog.waymo.com/2020/10/waymo-is-opening-its-fully-driverless.h… > We’ll start with those who are already a part of Waymo One and, over the next several weeks, welcome more people directly into the service through our app So...this is just an extension of what they currently do where it's not really open to the public? Waymo One is something you have to apply for and almost nobody gets in. It's not clear to me what "welcome more people directly into the service through our app" means. Waymo is terrible at PR. I can't even tell what they're trying to say here. But there is no date. There is no definitive statement. It's just another vague press release about how they're going driverless "soon"

Direct link to the announcement: https://blog.waymo.com/2020/10/waymo-is-opening-its-fully-driverless.h…

It's not that simple, for example Waymo has external investors alongside Alphabet, https://blog.waymo.com/2020/03/waymo-raises-first-external-investment.…

It's in the official waymo statement and was pretty easy to find: https://blog.waymo.com/2019/08/a-note-on-our-lawsuit-against-otto-and.… "Recently, we received an unexpected email. One of our suppliers specializing in LiDAR components sent us an attachment (apparently inadvertently) of machine drawings of what was purported to be Uber’s LiDAR circuit board — except its design bore a striking resemblance to Waymo’s unique LiDAR design." It was also in many articles. The story is that they were reluctant to pursue it until that email, you can find a ton of sources in the wikipedia entry: https://en.wikipedia.org/wiki/Anthony_Levandowski

Also on the blog (WSJ article has more details): https://blog.waymo.com/2020/03/waymo-raises-first-external-investment.…