Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The dawn of a world simulator

odyssey.ml

Post reply on HN
Print view View original

41–50 of 60 posts

Re: The dawn of a world simulator

#41

Post by KaiserPro »

godelski wrote: post #16

As a machine learning researcher, I don't get why these are called world models. Visually, they are stunning. But it's nowhere near physical. I mean look at that video with the girl and lion. The tail teleports between legs and then becomes attached to the girl instead of the tiger. Just because the visuals are high quality doesn't mean it's a world model or has learned physics. I feel like we're conflating these thi…

I think the issue is that "world models" are poorly defined.

With this kind of image gen, you can sorta plan robot interactions, but its super slow. I need to find the paper that deepmind produced, but basically they took the current camera input, used a text prompt like "robot arm picks up the ball", the video generated the arm motion, then the robot arm moved as it did in the video.

The problem is that its not really a world model, its just image gen. Its not like the model outputs a simulation that you can interact with (without generating more video) Its not like it creates a bunch of rough geo that you can then run physics on (ie you imagine a setup, draw it out and then run calcs on it.)

There is lots of work on making splats editable and semantically labeled, but again thats not like you can run physics on them so simulation is still very expensive. Also the properties are dependent on running the "world model" rather than querying the output at a point in time

Re: The dawn of a world simulator

#42

Post by godelski »

KaiserPro wrote: post #41
godelski wrote: post #16

As a machine learning researcher, I don't get why these are called world models. Visually, they are stunning. But it's nowhere near physical. I mean look at that video with the girl and lion. The tail teleports between legs and then becomes attached to the girl instead of the tiger. Just because the visuals are high quality doesn't mean it's a world model or has learned physics. I feel like we're conflating these thi…

I think the issue is that "world models" are poorly defined. With this kind of image gen, you can sorta plan robot interactions, but its super slow. I need to find the paper that deepmind produced, but basically they took the current camera input, used a text prompt like "robot arm picks up the ball", the video generated the arm motion, then the robot arm moved as it did in the video. The problem is that its not real…

  > poorly defined.
Poorly defined is not the same as undefined. There are bounds and we have a decent understanding of what this means. Not having the details all worked out is not the same. Though that lack of precision is being used to get away with more slop.

  > I need to find the paper that deepmind produced
I've seen that paper and the results pretty close to the action. I've even personally talked with people that worked on that paper. It very frequently "forgets" what is outside its view and it very frequently performs non-physically consistent actions. When you evaluate those models don't just try standard things, do weird things. Like keep trying to extend the grabber arm and it shouldn't jump to other parts of the screen.

  > The problem is that its not really a world model, its just image gen.
Yes, that was my point. Since you agree I'm not sure why you're disagreeing.

Re: The dawn of a world simulator

#43

Post by KaiserPro »

godelski wrote: post #42
KaiserPro wrote: post #41

Earlier quoted context omitted.

I think the issue is that "world models" are poorly defined. With this kind of image gen, you can sorta plan robot interactions, but its super slow. I need to find the paper that deepmind produced, but basically they took the current camera input, used a text prompt like "robot arm picks up the ball", the video generated the arm motion, then the robot arm moved as it did in the video. The problem is that its not real…

> poorly defined. Poorly defined is not the same as undefined. There are bounds and we have a decent understanding of what this means. Not having the details all worked out is not the same. Though that lack of precision is being used to get away with more slop. > I need to find the paper that deepmind produced I've seen that paper and the results pretty close to the action. I've even personally talked with people tha…

I don't think I'm disagreeing, just adding more colour.

> It very frequently "forgets" what is outside its view

This was the observations that I saw when we were testing it. My former lab was late to pivoting to robotics, so we were looking at the current state of play to see what machine perception stuff is out there for robotics.

Re: The dawn of a world simulator

#44

Post by godelski »

KaiserPro wrote: post #43
godelski wrote: post #42

Earlier quoted context omitted.

> poorly defined. Poorly defined is not the same as undefined. There are bounds and we have a decent understanding of what this means. Not having the details all worked out is not the same. Though that lack of precision is being used to get away with more slop. > I need to find the paper that deepmind produced I've seen that paper and the results pretty close to the action. I've even personally talked with people tha…

I don't think I'm disagreeing, just adding more colour. > It very frequently "forgets" what is outside its view This was the observations that I saw when we were testing it. My former lab was late to pivoting to robotics, so we were looking at the current state of play to see what machine perception stuff is out there for robotics.

Ah, thanks for the clarification. It can be hard to interpret on these forums sometimes.

Re: The dawn of a world simulator

#45

Post by jyunth »

Interesting. I imagine quite a few issues would seem to stem out of the inherent nature of generative AI, we even see several in these demos themselves. One particularly stood out to me, the one where the man is submerged, and for a good while bubbles come out quite consistently out of his mask, and then suddenly one of the bubbles turn into a jellyfish. At a specific frame, the AI thought it looked more like a jellyfish than a bubble and now the world has a jellyfish to deal with now.

It'll surely take a looot of video data, even more than what humans can possibly produce to build a normalized, euclidean, physics adherent world model. Data could be synthetically generated, checked thoroughly and fed to the training process, but at the end of the day it seems.... Wasteful. As if we're looking at a local optima point.

Re: The dawn of a world simulator

#46

Post by andy12_ »

godelski wrote: post #16

As a machine learning researcher, I don't get why these are called world models. Visually, they are stunning. But it's nowhere near physical. I mean look at that video with the girl and lion. The tail teleports between legs and then becomes attached to the girl instead of the tiger. Just because the visuals are high quality doesn't mean it's a world model or has learned physics. I feel like we're conflating these thi…

You just have to extrapolate the improvements in consistency in image model from the last couple of years and apply it to these kinds of video models. When in a couple of years they can generate consistent videos of many physical phenomena such that they are nearly indistinguishably from reality, you'll se why they are called "world models".

Re: The dawn of a world simulator

#48

Post by IAmGraydon »

slashdave wrote: post #27
nl wrote: post #21

The reason they are called "world models" is because the internal representation of what they display represents a "world" instead of a video frame or image. The model needs to "understand" geometry and physics to output a video. Just because there are errors in this doesn't mean it isn't significant. If a machine learning model understands how physical objects interact with each other that is very useful.

> The model needs to "understand" geometry and physics to output a video. No it doesn't. It merely needs to mimic.

Correct. The fact that AI is a black box means we can easily imagine anything we want happening within that box. Or perhaps the more accurate way to say it - AI companies can convince investors of amazing magic happening within that box. With LLMs, we anthropomorphize and imagine it’s thinking. With video models, they’re now trying to convince us that it understands the world. None of these things are true. It’s all an illusion.

Re: The dawn of a world simulator

#49

Post by nurettin »

godelski wrote: post #25
nurettin wrote: post #18

Earlier quoted context omitted.

> Visually, they are stunning. The input images are stunning, model's result is another disappointing trip to uncanny valley. But we feel Ok as long as the sequence doesn't horribly contradict the original image or sound. That is the world model.

> But we feel Ok as long as the sequence doesn't horribly contradict the original image or sound. Is the error I pointed out not "horribly contradicting"? > That is the world model. I would say that if it is non-physical[0] then it's hard to call it a /world/ model. A world is consistent and has a set of rules that must be followed. I've yet to see a claimed world model that actually captures this behavior. Yet it's…

> Yet if one studied physics (beyond your basic college courses) you'd understand the naïveté of that.

I studied enough physics to get a mech. eng. diploma. And I still understand the naivete. Observational physics can be derived with ml, and I have derived them, but not with neural nets. Or if you do it with neural nets, you can't alpha zero it, you need to cheat.

Re: The dawn of a world simulator

#50

Post by slashdave »

IAmGraydon wrote: post #48
slashdave wrote: post #27

Earlier quoted context omitted.

> The model needs to "understand" geometry and physics to output a video. No it doesn't. It merely needs to mimic.

Correct. The fact that AI is a black box means we can easily imagine anything we want happening within that box. Or perhaps the more accurate way to say it - AI companies can convince investors of amazing magic happening within that box. With LLMs, we anthropomorphize and imagine it’s thinking. With video models, they’re now trying to convince us that it understands the world. None of these things are true. It’s all…

It's worse than that. It's not a black box. We know how the architecture is constructed. We can read the weights.
Post reply on HN