HN Explorer

89 comments

[ 2.6 ms ] story [ 71.1 ms ] thread
xnx
> The Waymo World Model can convert those kinds of videos, or any taken with a regular camera, into a multimodal simulation—showing how the Waymo Driver would see that exact scene.

Subtle brag that Waymo could drive in camera-only mode if they chose to. They've stated as much previously, but that doesn't seem widely known.

OGEnthusiast
What's going to happen to all the millions of drivers who will lose their job overnight? In a country with 100 million guns, are we really sure we've thought this through?
8note
this sounds like a major benefit.

i dont want my uber driver bragging anout how theyre going to shoot me before i get out of the car

bdangubic
what did all the farmers do when the first tractors rolled into the fields?
mgaunard
Still needs to be trained on the final boss: dense cities with narrow streets.
NullHypothesist
I wonder if they can simulate the Beatles crossing the street at Abbey Road in the late '60s
devmor
Wow, interesting timing for this PR blast considering the admission in the Senate Commerce Committee hearing. Not transparent at all!
ra7
The novel aspect here seems to be 3D LiDAR output from 2D video using post-training. As far as I'm aware, no other video world models can do this.

IMO, access to DeepMind and Google infra is a hugely understated advantage Waymo has that no other competitor can replicate.

anupamchugh
This is the real story buried under the simulation angle. If you can generate reliable 3D LiDAR from 2D video, every dashcam on earth becomes training data. Every YouTube driving video, every GoPro clip, every security camera feed.

Waymo's fleet is ~700 cars. The internet has millions of hours of driving footage. This technique turns the entire internet into a sensor suite. That's a bigger deal than the simulation itself.

PeterStuer
Imagine driving in a Waymo 'out of a raging fire'.

Talk about edge cases.

But, what would you do? Trust the Waymo, or get out (or never get in) at the first sign of trouble?

m0llusk
Seems interesting, but why is it broken. Waymo repeatedly directed multiple automated vehicles into the private alley off of 5th near Brannan in SF even after being told none of them have any business there ever, period. If they can sense the weather and stuff then maybe they could put out a virtual sign or fence that notes what appears to be a road is neither a through way nor open to the public? I'm really bullish on automated driving long term, but now that vehicles are present for real we need to start to think about potentially getting serious about finding some way to get them to comply with the same laws that limit what people can do.
turtlesdown11
How many Filipinos, who do not have US drivers licenses, does it take to drive this new model?
andrewmcwatters
It feels intellectually negligent to me that models like those used by Openpilot have no auditable concept of what a car, or a sign, or a person is.

Just where lines are and when a car should accelerate or break. The rest of the latent state is "based on pixels."

mempko
One interesting thing from this paper is how big of a LiDaR shadow there is around the waymo car which suggests they rely on cameras for anything close (maybe they have radar too?). Seems LiDaR is only useful for distant objects.
999900000999
It doesn't look like they're going to open sources or anything, but I could imagine this would be great for city planning.

Or the most realistic game of SimCity you could imagine.

mattlondon
Suddenly all this focus on world models by Deep mind starts to make sense. I've never really thought of Waymo as a robot in the same way as e.g. a Boston Dynamics humanoid, but of course it is a robot of sorts.

Google/Alphabet are so vertically integrated for AI when you think about it. Compare what they're doing - their own power generation , their own silicon, their own data centers, search Gmail YouTube Gemini workspace wallet, billions and billions of Android and Chromebook users, their ads everywhere, their browser everywhere, waymo, probably buy back Boston dynamics soon enough (they're recently partnered together), fusion research, drugs discovery.... and then look at ChatGPT's chatbot or grok's porn. Pales in comparison.

rswail
Maybe they were focusing on a real world use that basically requires AI, but not LLMs.

Tesla claimed that all their "real world" recording would give them a moat on FSD.

Waymo is showing that a) you need to be able to incorporate stuff that isn't "real" when training, and b) you get a lot more information from alternate sensors to visible spectrum only.

ericzundel
I just listened to a fantastic multi-hour Acquired (https://www.acquired.fm/) podcast episode on Google and AI that talks about the history of Google and AI and all the ways they have been using it since 2012. It's really fascinating. You can forgive them for not focusing on Reader or any of their other properties when you realize they were pulling in hundreds of billions of dollars of value by making big bets in AI and incorporating it into their core business.
theptip
So is this a model baked into the VLLM layer? Or a scaffold that the agent sits in for testing?

If the former then it’s relevant to the broader discourse on LLM generality. If the latter, then it seems less relevant to chatbots and business agents.

Edit to add: this is not part of the model, it’s in a separate pillar (Simulator vs Driver). More at https://waymo.com/blog/2025/12/demonstrably-safe-ai-for-auto....

YeGoblynQueenne
>> Suddenly all this focus on world models by Deep mind starts to make sense.

The apparent applicability to Waymo is incidental, more likely because a few millions+ were spent on Genie and they have to do something with it. DeepMind started to train "world models" because that's the current overhyped buzzword in the industry. First it was "natural language understanding" and "question answering" back in the days of old BERT, then it was "agentic", then "reasoning", now it's "world models", next years it's going to be "emotions" or "social intelligence" or some other anthropomorphic, over-drawn neologism. If you follow a few AI accounts on social media you really can't miss when those things suddenly start trending, then pretty much die out and only a few stragglers still try to publish papers on them because they failed to get the memo that we're now all running behind the Next Big Thing™.

Vosporos
The new frontier is manifestly the Phillipines.
hazrmard
cue the bell curve meme for learning autonomy:

                 ____.----.____
          ______/              \______
    _____/                            \_____
    ________________________________________

    (simulations)  (real world data)  (simulations)
Seems like it, no?

We started with physics-based simulators for training policies. Then put them in the real world using modular perception/prediction/planning systems. Once enough data was collected, we went back to making simulators. This time, they're physics "informed" deep learning models.

tgrowazay
This page crashes my browser.

Vivaldi 7.8.3931.63 on iOS 26.2.1 iPhone 16 pro

AndrewKemendo
For whatever it’s worth World models is going to be the dominant computing structure of the future

I started working heavily on realizing them in 2016 and it is unquestionably (finally) the future of AI

ge96
What is the 5/3 tiles? Cameras?
01100011
Nvidia has had this for years. What am I missing?