That felt so wrong AND someone is cheating here. This felt really suspicious...
I got to the graffiti world and there were some stairs right next to me. So I started going up them. It felt like I was walking forward and the stairs were pushing under me until I just got stuck. So I turned to go back down and half way around everything morphed and I ended up back down at the ground level where I originally was. I was teleported. That's why I feel like something is cheating here. If we had mode collapse I'm not sure how we should be able to completely recover our entire environment. Not unless the model is building mini worlds with boundaries. It was like the out of bond teleportation you get in some games but way more fever dream like. That's not what we want from these systems, we don't want to just build a giant poorly compressed videogame, we want continuous generation. If you have mode collapse and recover, it should recover to somewhere new, now where you've been. At least this is what makes me highly suspicious.
Yes the thing that got me was i went through the channels multiple times (multiple browser sessions). The channels are the same everytime (the numbers don't align to any navigation though - flip back and forth between two numbers and you'll just hit a random channel everytime - don't be fooled by that). Every object is in the same position and the layout is the same.
What makes this AI generated over just rendering a generated 3D scene?
Like it may seem impressive to have no glitches (often in AI generated works you can turn around a full rotation and you're what's in front of you isn't what was there originally) but here it just acts as a fully modelled 3D scene rendering at low resolution? I can't even walk outside of certain bounds which doesn't make sense if this really is generated on the fly.
This needs a lot of skepticism and i'm surprised you're the first commenting on the lack of actual generation here. It's a series of static scenes rendered at low fidelity with limited bounds.
Ok playing with this more there's very subtle differences between sessions. As in there is some hallucination here with certain small differences.
I think what's happening is this is AI generated but it is very very overfitted to real world 3D scenes. The AI is almost rendering exactly a real world scene and not much more. They can't travel out of bounds or the model stops working since it's so overfitted to these scenes. The overfitting solves hallucinations but it also makes it almost indistinguishable from pre modelled 3D scenes.
Is it possible that this behavior is a result from training on Google Maps or something similar? I tried to walk off a bridge and you get completely stuck, which is the only reason I can think of that, other than not having first person video views of people walking off bridges.
Hi! CEO of Odyssey here. Thanks for giving this a shot.
To clarify: this is a diffusion model trained on lots of video, that's learning realistic pixels and actions. This model takes in the prior video frame and a user action (e.g. move forward), with the model then generating a new video frame that resembles the intended action. This loop happens every ~40ms, so real-time.
The reason you're seeing similar worlds with this production model is that one of the greatest challenges of world models is maintaining coherence of video over long time periods, especially with diverse pixels (i.e. not a single game). So, to increase reliability for this research preview—meaning multiple minutes of coherent video—we post-trained this model on video from a smaller set of places with dense coverage. With this, we lose generality, but increase coherence.
> One of the biggest challenges is that world models require autoregressive modeling, predicting future state based on previous state. This means the generated outputs are fed back into the context of the model. In language, this is less of an issue due to its more bounded state space. But in world models—with a far higher-dimensional state—it can lead to instability, as the model drifts outside the support of its training distribution. This is particularly true of real-time models, which have less capacity to model complex latent dynamics.
> To improve autoregressive stability for this research preview, what we’re sharing today can be considered a narrow distribution model: it's pre-trained on video of the world, and post-trained on video from a smaller set of places with dense coverage. The tradeoff of this post-training is that we lose some generality, but gain more stable, long-running autoregressive generation.
> To broaden generalization, we’re already making fast progress on our next-generation world model. That model—shown in raw outputs below—is already demonstrating a richer range of pixels, dynamics, and actions, with noticeably stronger generalization.
Why are you going all in on world models instead of basing everything on top of a 3D engine that could be manipulated / rendered with separate models? If a world model was truly managing to model a manifold of a 3D scene, it should be pretty easy to extract a mesh or SDF from it and drop that into an engine where you could then impose more concrete rules or sanity check the output of the model. Then you could actually model player movement inside of the 3D engine instead of trying to train the world model to accept any kind of player input you might want to do now or in the future.
Additionally, curious about what exactly the difference between the new mode of storytelling you’re describing and something like a crpg or visual novel is - is your hope that you can just bake absolutely everything into the world model instead of having to implement systems for dialogue/camera controls/rendering/everything else that’s difficult about working with a 3D engine?
This is super cool. I love how you delivered this as an experience, with very cool UI, background music, etc. It was a real trip. Different music and also an ambient mode where it’s atmosphere designed for the particular world would level this up artistically. But I gotta say, I haven’t been this intrigued by anything like this online in a while.
Does it stores a model of the world (like, some memory of the 3d structure) that goes beyond the pixels that are shown?
Or is the next frame a function of just the previous frame and the user input? Like (previous frame, input) -> next frame
I'm asking because, if some world has two distinct locations that look exactly the same, will the AI distinguish them, or will they get coalesced into one location?
> one of the greatest challenges of world models is maintaining coherence of video over long time periods
To be honest most of the appeal to me of this type of thing is the fact that it gets incoherent and morph-y and rotating 360 degrees can completely change the scenery. It's a trippy dreamlike experience whereas this kind of felt like a worse version of existing stuff.
Thanks for the reply and adding some additional context. I'm also a vision researcher, fwiw (I'll be at CVPR if you all are).
(Some of this will be for benefit of other HN non-researcher readers)
I'm hoping you can provide some more. Are these training on single video moving through these environments, where the camera is not turning? What I am trying to understand is what is being generated vs what is being recalled.
It may be a more contentious view, but I do not think we're remotely ready to call these systems "world models" if they are primarily performing recall. Maybe this is the bias from an education in physics (I have a degree), but world modeling is not just about creating consistent imagery, but actually being capable of recovering the underlying physics of the videospace (as opposed to reality which the videos come from). I've yet to see a demonstration of a model that comes anywhere near this or convinces me we're on the path towards this.
The key difference here is are we building Doom which has a system requirements of 100MB disk and 8MB RAM with minimal computation or are we building a extremely decompressed version that requires 4GB of disk and a powerful GPU to run only the first level and can't even get critical game dynamics right like shooting the right enemy (GameNGen).
The problem is not the ability to predict future states based on previous ones, the problem is the ability to recover /causal structures/ from observation.
Critically, a p̶h̶y̶s̶i̶c̶s̶ world model is able to process a counterfactual.
Our video game is able to make predictions, even counterfactual predictions, with its engine. Of course, this isn't generated by observation and environment interaction, it is generated through directed programming and testing (where the testing includes observing and probing the environment). If the goal was just that, then our diffusion models would comparatively be a poor contender. It's the wrong metric. The coherence is a consequence of the world modeling (i.e. game engine) but it coherence can also be developed from recall. Recall alone will be unable to make a counterfactual.
Certainly we're in research phase and need to make tons of improvements, but we can't make these improvements if we're blindly letting our models cheat the physics and are only capable of picking up "user clicks fire" correlating with "monster usually dies when user shoots". LLMs have tons of similar problems with making such shortcuts and the physics will tell you that you are not going to be able to pick up such causal associations without some very specific signals to observe. Unfortunately, causality can not be determined from observation alone (a well known physics result![0]). You end up with many models that generate accurate predictions, and these become non-differentiable without careful factorization, probing, and often requiring careful integration of various other such models. It is this much harder and more nuanced task that is required of a world model rather than memory.
Essentially, do we have "world models" or "cargo cult world models" (recall or something else).
That's the context of my data question. To help us differentiate the two. Certainly the work is impressive and tbh I do believe there is quite a bit of utility in the cargo cult setting, but we should also be clear about what is being claimed and what isn't.
I'm also interested in how you're trying to address the causal modeling problem.
[0] There is much discussion on the Duhem-Quine thesis, which is a much stronger claim than I stated. There's the famous Michelson-Morley experiment, which actually did not rule out an aether, but rather only showed that it had no directionality. Or we could even use the classic Heisenberg Uncertainty Principle which revolutionized quantum mechanics showing that there are things that are unobservable, leading to Schrodinger's Cat (some weird hypotheses of multiverses). And we even have String Theory, which the main gripe remains that it is non-differentiable from other TOES due to differences in predictions being non-observable.
It’s essentially this paper but applied to a bunch of video recordings of a bunch of different real world locations instead of counter strike maps. Each channel is just changing the location.
> Not unless the model is building mini worlds with boundaries.
Right. I was never able to get very far from the starting point, and kept getting thrown back to the start. It looks like they generated a little spherical image, and they're able to extrapolate a bit from that. Try to go through a door or reach a distant building, and you don't get there.
I mean... https://news.ycombinator.com/item?id=44121671
informed you of exactly why this happens a whole hour before you posted this comment and the creator is chatting with people in the comments. I get that you feel personally cheated, but I really don't think anyone was deliberately trying to cheat you. In light of that, your comment (and i only say this because it's the top comment on this post) is effectively a stereotypical "who needs dropbox" levels of shallow dismissal.
> I mean... https://news.ycombinator.com/item?id=44121671 informed you of exactly why this happens a whole hour before you posted this comment and the creator is chatting with people in the comments.
I apologize for replying with my experience and not reading every comment before posting. This was not the top comment when I wrote mine.
> and the creator is chatting with people in the comments.
This had yet to occur when I left my comment. I looked at their comments and it appears I'm the first person they responded to in this thread. Took me a bit to respond, but hey, I got stuff to do too.
> is effectively a stereotypical "who needs dropbox" levels of shallow dismissal.
I'll loop you into my response to the creator, which adds context to my question. This not a "who needs dropbox" so much as "why are you calling dropbox 'storing data without taking any disk space'". Sure, it doesn't take your disk space, but that's not no disk space... Things are a bit clearer now, but I gotta work with the context of what's given to me.
https://news.ycombinator.com/item?id=44147777
> and i only say this because it's the top comment on this post
I appreciate you holding people to high standards. No issues there. It's why I made my comment in the first place! Hopefully my other comment clarifies what was definitely lacking in the original.
It feels like an interpolated Street View imagery. There is one scene with two people between cars in a parking lot. It is the only one I have found that has objects you would expect to change over time. When exploring the scene, those people sometimes disappear altogether and sometimes teleport around, as they would when exploring Street View panoramas. You can clearly tell when you are switching between photos taken a few seconds apart.
Note that it isn't being created from whole cloth, it is trained on videos of the places and then it is generating the frames:
"To improve autoregressive stability for this research preview, what we’re sharing today can be considered a narrow distribution model: it's pre-trained on video of the world, and post-trained on video from a smaller set of places with dense coverage. The tradeoff of this post-training is that we lose some generality, but gain more stable, long-running autoregressive generation."
Well, that felt like entering a dream on my phone. Fuzzy virtual environments generated by "a mind" based on its memory of real environments...
I wonder if it'd break our brains more if the environment changes as the viewpoint changes, but doesn't change back (e.g. if there's a horse, you pan left, pan back right, and the horse is now a tiger).
I kept expecting that to happen, but it apparently has some mechanism to persist context outside the user’s FOV.
In a way, that almost makes it more dreamlike, in that you have what feels like high local coherence (just enough not to immediately tip you off that it’s a dream) that de-coheres over time as you move through it.
This is pretty much the same thing as those models that baked dust2 into a diffusion model then used the last few frames as context to continue generating - same failure modes and everything.
This is similar to the Minecraft version of this from a few months back [0], but it does seem to have a better time keeping a memory of what you've already seen, at least for a bit. Spinning in circles doesn't lose your position quite as easily, but I did find that exiting a room and then turning back around and re-entering leaves you with a totally different room than you exited.
Only at first glance. It can easily render things that would be very hard to implement in an FPS engine.
What AI can dream up in milliseconds could take hundreds of human hours to encode using traditional tech (meshes, shaders, ray tracing, animation, logic scripts, etc.), and it still wouldn't look as natural and smooth as AI renderings — I refer to the latest developments in video synthesis like Google's Veo 3. Imagine it as a game engine running in real time.
Why do you think this is so hard, even for technical people here, to make the inductive leap on this one? Is it that close to magic? The AI is rendering pillars and also determining collision detection on it. As in, no one went in there and selected a bunch of pillars and marked it as a barrier. That means in the long run, I'll be able to take some video or pictures of the real world and have it be game level.
I think an actual 3D engine with AI that can make new high quality 3D models and environments on the fly would be the pinnacle. And maybe even add new game and control mechanics on the fly.
it’s super cool. I keep thinking it kind of feels like dream logic. It looks amazing at first but I’m not sure I’d want to stay in a world like that for too long. I actually like when things have limits. When the world pushes back a bit and gives you rules to work with.
Doesn't it have rules? I couldn't move past a certain point and hitting a wall made you teleport. Maybe I was just rationalizing random events, though.
Readit News
I got to the graffiti world and there were some stairs right next to me. So I started going up them. It felt like I was walking forward and the stairs were pushing under me until I just got stuck. So I turned to go back down and half way around everything morphed and I ended up back down at the ground level where I originally was. I was teleported. That's why I feel like something is cheating here. If we had mode collapse I'm not sure how we should be able to completely recover our entire environment. Not unless the model is building mini worlds with boundaries. It was like the out of bond teleportation you get in some games but way more fever dream like. That's not what we want from these systems, we don't want to just build a giant poorly compressed videogame, we want continuous generation. If you have mode collapse and recover, it should recover to somewhere new, now where you've been. At least this is what makes me highly suspicious.
What makes this AI generated over just rendering a generated 3D scene?
Like it may seem impressive to have no glitches (often in AI generated works you can turn around a full rotation and you're what's in front of you isn't what was there originally) but here it just acts as a fully modelled 3D scene rendering at low resolution? I can't even walk outside of certain bounds which doesn't make sense if this really is generated on the fly.
This needs a lot of skepticism and i'm surprised you're the first commenting on the lack of actual generation here. It's a series of static scenes rendered at low fidelity with limited bounds.
I think what's happening is this is AI generated but it is very very overfitted to real world 3D scenes. The AI is almost rendering exactly a real world scene and not much more. They can't travel out of bounds or the model stops working since it's so overfitted to these scenes. The overfitting solves hallucinations but it also makes it almost indistinguishable from pre modelled 3D scenes.
To clarify: this is a diffusion model trained on lots of video, that's learning realistic pixels and actions. This model takes in the prior video frame and a user action (e.g. move forward), with the model then generating a new video frame that resembles the intended action. This loop happens every ~40ms, so real-time.
The reason you're seeing similar worlds with this production model is that one of the greatest challenges of world models is maintaining coherence of video over long time periods, especially with diverse pixels (i.e. not a single game). So, to increase reliability for this research preview—meaning multiple minutes of coherent video—we post-trained this model on video from a smaller set of places with dense coverage. With this, we lose generality, but increase coherence.
We share a lot more about this in our blog post here (https://odyssey.world/introducing-interactive-video), and share outputs from a more generalized model.
> One of the biggest challenges is that world models require autoregressive modeling, predicting future state based on previous state. This means the generated outputs are fed back into the context of the model. In language, this is less of an issue due to its more bounded state space. But in world models—with a far higher-dimensional state—it can lead to instability, as the model drifts outside the support of its training distribution. This is particularly true of real-time models, which have less capacity to model complex latent dynamics.
> To improve autoregressive stability for this research preview, what we’re sharing today can be considered a narrow distribution model: it's pre-trained on video of the world, and post-trained on video from a smaller set of places with dense coverage. The tradeoff of this post-training is that we lose some generality, but gain more stable, long-running autoregressive generation.
> To broaden generalization, we’re already making fast progress on our next-generation world model. That model—shown in raw outputs below—is already demonstrating a richer range of pixels, dynamics, and actions, with noticeably stronger generalization.
Let me know any questions. Happy to go deeper!
Additionally, curious about what exactly the difference between the new mode of storytelling you’re describing and something like a crpg or visual novel is - is your hope that you can just bake absolutely everything into the world model instead of having to implement systems for dialogue/camera controls/rendering/everything else that’s difficult about working with a 3D engine?
Or is the next frame a function of just the previous frame and the user input? Like (previous frame, input) -> next frame
I'm asking because, if some world has two distinct locations that look exactly the same, will the AI distinguish them, or will they get coalesced into one location?
To be honest most of the appeal to me of this type of thing is the fact that it gets incoherent and morph-y and rotating 360 degrees can completely change the scenery. It's a trippy dreamlike experience whereas this kind of felt like a worse version of existing stuff.
(Some of this will be for benefit of other HN non-researcher readers)
I'm hoping you can provide some more. Are these training on single video moving through these environments, where the camera is not turning? What I am trying to understand is what is being generated vs what is being recalled.
It may be a more contentious view, but I do not think we're remotely ready to call these systems "world models" if they are primarily performing recall. Maybe this is the bias from an education in physics (I have a degree), but world modeling is not just about creating consistent imagery, but actually being capable of recovering the underlying physics of the videospace (as opposed to reality which the videos come from). I've yet to see a demonstration of a model that comes anywhere near this or convinces me we're on the path towards this.
The key difference here is are we building Doom which has a system requirements of 100MB disk and 8MB RAM with minimal computation or are we building a extremely decompressed version that requires 4GB of disk and a powerful GPU to run only the first level and can't even get critical game dynamics right like shooting the right enemy (GameNGen).
The problem is not the ability to predict future states based on previous ones, the problem is the ability to recover /causal structures/ from observation.
Critically, a p̶h̶y̶s̶i̶c̶s̶ world model is able to process a counterfactual.
Our video game is able to make predictions, even counterfactual predictions, with its engine. Of course, this isn't generated by observation and environment interaction, it is generated through directed programming and testing (where the testing includes observing and probing the environment). If the goal was just that, then our diffusion models would comparatively be a poor contender. It's the wrong metric. The coherence is a consequence of the world modeling (i.e. game engine) but it coherence can also be developed from recall. Recall alone will be unable to make a counterfactual.
Certainly we're in research phase and need to make tons of improvements, but we can't make these improvements if we're blindly letting our models cheat the physics and are only capable of picking up "user clicks fire" correlating with "monster usually dies when user shoots". LLMs have tons of similar problems with making such shortcuts and the physics will tell you that you are not going to be able to pick up such causal associations without some very specific signals to observe. Unfortunately, causality can not be determined from observation alone (a well known physics result![0]). You end up with many models that generate accurate predictions, and these become non-differentiable without careful factorization, probing, and often requiring careful integration of various other such models. It is this much harder and more nuanced task that is required of a world model rather than memory.
Essentially, do we have "world models" or "cargo cult world models" (recall or something else).
That's the context of my data question. To help us differentiate the two. Certainly the work is impressive and tbh I do believe there is quite a bit of utility in the cargo cult setting, but we should also be clear about what is being claimed and what isn't.
I'm also interested in how you're trying to address the causal modeling problem.
[0] There is much discussion on the Duhem-Quine thesis, which is a much stronger claim than I stated. There's the famous Michelson-Morley experiment, which actually did not rule out an aether, but rather only showed that it had no directionality. Or we could even use the classic Heisenberg Uncertainty Principle which revolutionized quantum mechanics showing that there are things that are unobservable, leading to Schrodinger's Cat (some weird hypotheses of multiverses). And we even have String Theory, which the main gripe remains that it is non-differentiable from other TOES due to differences in predictions being non-observable.
https://diamond-wm.github.io/
Right. I was never able to get very far from the starting point, and kept getting thrown back to the start. It looks like they generated a little spherical image, and they're able to extrapolate a bit from that. Try to go through a door or reach a distant building, and you don't get there.
Deleted Comment
I call BS.
Deleted Comment
"To improve autoregressive stability for this research preview, what we’re sharing today can be considered a narrow distribution model: it's pre-trained on video of the world, and post-trained on video from a smaller set of places with dense coverage. The tradeoff of this post-training is that we lose some generality, but gain more stable, long-running autoregressive generation."
https://odyssey.world/introducing-interactive-video
https://diamond-wm.github.io/
I wonder if it'd break our brains more if the environment changes as the viewpoint changes, but doesn't change back (e.g. if there's a horse, you pan left, pan back right, and the horse is now a tiger).
In a way, that almost makes it more dreamlike, in that you have what feels like high local coherence (just enough not to immediately tip you off that it’s a dream) that de-coheres over time as you move through it.
Fascinatingly strange demo.
Deleted Comment
https://diamond-wm.github.io/
[0] Minecraft with object impermanence (229 points, 146 comments) https://news.ycombinator.com/item?id=42762426
What AI can dream up in milliseconds could take hundreds of human hours to encode using traditional tech (meshes, shaders, ray tracing, animation, logic scripts, etc.), and it still wouldn't look as natural and smooth as AI renderings — I refer to the latest developments in video synthesis like Google's Veo 3. Imagine it as a game engine running in real time.
Deleted Comment
https://m.youtube.com/watch?v=YXPIv7pS59o
Dead Comment
Eventually managed to leave the first room, but then got teleported somewhere else.