Hacker News

points by numlock86 2 days ago | 0 comments

Reading the comments almost everyone seems not impressed at all and points at obvious errors in the samples. Meanwhile it's been just roughly over 3 years since we've seen an AI generated abomination of Will Smith "eating" (or whatever you want to call that act) spaghetti. Yes, AI gets things wrong. You could say the same about LLMs still making obvious errors here and there. To me, personally, this seems super impressive and it's clearly another leap forward. Not perfect, yes. But an improvement.

[–] edstarch 2 days ago

I'm by no means anti-anything but I think what the comments are getting at is that these "world models" do not have a world model. It's a fundamental qualitative gap that no amount of compute has been able to close, just mask by everything else getting better.

This doesn't show up as much where the verifiable reward is "code compiles and tests pass" but you can still absolutely see it in things like movies, prose, or slide decks. Even frontier models are pretty bad at keeping track of narrative threads even as the technical quality drastically improves.

This begs the question of whether the gap can be closed just with more compute (presumably yes, or at least errors can be minimized beyond human capacity to see them) and is that going to be worth it with current tech (possibly no).

[–] exit 1 day ago

does your model of models admit the possibility of partial models?

the average human cannot vividly recollect a walk through a familiar neighborhood, much less invoke a coherent fantasy of going on such a walk. do they contain a world model?

[–] tyg13 1 day ago

Must we, every time the limitations of AI are discussed, always imagine some simpleton human to defend it? It's trite and rarely adds anything to the conversation. The point is to judge the model on its merits. If this is a good world model, then it's a good world model. It doesn't matter whether or not a human can recollect every detail about a familiar neighborhood. It's irrelevant

[–] handoflixue 1 day ago

It matters whether humans can pass the test or not.

Humans also get details wrong when they do stuff like this. So right away, we learn this isn't a limitation of AI. If you want to say that getting that wrong is "proof they don't have a world model", that cuts both ways - it means humans do just fine despite not having one of those either.

But that also lends us insight into fixing the issue on merit! If humans also obviously lack a world model, how do we cope with this? Sometimes we use "live action" and reality forces consistency. Sometimes we use editors and QA. Sometimes it ends up on a website listing every little shot-to-shot inconsistency in every single TV show and movie ever made.

And now you have a few clear ideas on how to train the next generation of image models to do better! And you can probably dramatically improve your own prompting by providing reference images, using consistent descriptions, etc..

So, hardly irrelevant.

[–] cman1444 1 day ago

Exactly. Also, for practical purposes, an AI's capability to meet or exceed human performance is a very important benchmark. It makes the decision between using AI vs paying a human creator very easy.

[–] shoelessone 1 day ago

I think what we're talking about though here in reality is our perception of the model's output.

Our perception in realtime needs to be better than our ability to remember things from the past.

[–] handoflixue 1 day ago

I strongly suspect that if you fed this video through an LLM, it would spot all of the mistakes - the problem is that parsing and generating video is still pretty expensive, so that sort of automatic multi-pass process isn't showing up yet. Better to let the human decide whether they mind enough to pay for edits.

[–] yreg 1 day ago

The point is that humans obviously have world models, just imperfect ones.

[–] ainch 1 day ago

My problem with it is the shift in focus. The original Genie was pitched as a simulator of the world that an agent can interact with to learn so that it could be deployed into the real world. This new version looks like a good, controllable video generator. The fact that you're comparing it to the will smith video, rather than a simulator, is indicative of the fact they've strayed from that original goal. I think that's a shame

[–] roryirvine 1 day ago

I mostly agree with you, though I think it's more of a sharpening of focus onto a narrow subset of the original grand vision that's more likely to be immediately useful as a product.

Whether it's a product that anyone actually wants is another matter, of course - as is whether they'll ever be able to zoom out and bring the entirety of the original vision into focus if this more limited version does become a success.

[–] onion2k 1 day ago

It's a huge leap from 3 years ago, but not a huge leap from other current AI models. People who are deep in the AI rabbit hole have seen impressive generative video that has fewer problems (or possibly just better chosen examples).

[–] Valakas_ 1 day ago

People are emotionally driven. When emotions drive us to do something we will (unconsciously) find plausible justifications for it without realizing we're doing it.

Anti-AI crowd is anti AI. Maybe they were (mostly) happy with status quo and don't like change. Maybe scared for the future. Or any other reason. Went from "AI can't even tell what 2+2 is", to "AI can't even tell whether these simply connected smooth 4-manifolds are merely homeomorphic or actually diffeomorphic after stabilization" without batting an eye. The goalposts will keep being moved and the dismissal will continue.

[–] metalrain 1 day ago

Visually it looks better than last generation, some materials and lighting, very good. But also there is lot of fuzziness, it's not good enough for TV ad. It doesn't beat the best video generation models, but being interactively steerable it could have more utility or mean you need less "takes" to get the right angle.

[–] kleiba2 2 days ago

I agree. This is nothing short of magic -- or perhaps 95% of magic. Imagine yourself 5 years ago, and someone from today going back in time and telling you "hey, check this out, in just 5 years you'll be able to write this short text and a computer will generate this movie from it"...

[–] freecodeio 1 day ago

"and nobody is going to watch it because for some reason it feels soulless no matter the output"

[–] cman1444 1 day ago

Until it gets good enough that you can't tell the difference. It will climb out of the uncanny valley eventually.

[–] freecodeio 1 day ago

it's already good enough, the flat emotional response we have towards ai is that nobody in the screen put any sweat to it

for the time being we've all grown up to value hard work and that's a depth vector we subconsciously consider when watching something, when we watch an AI video it's just pixels, maybe it will change with future generations as what's real and what's not makes no difference

[–] kleiba2 1 day ago

I'm excited that I might actually be able to realize the animated movie I've had in my head for a long time.

[–] sebzim4500 1 day ago

Plenty of people will watch it, souls don't exist.

[–] freecodeio 1 day ago

I have never seen a person impressed with the contents of an AI video other than the fact that AI can make the said video.

also > souls don't exist

yes i know, but if you think you're no better than an LLM then why don't you put your personality into a system prompt and spare us the carbon footprint

[–] kleiba2 1 day ago

> I have never seen a person impressed with the contents of an AI video other than the fact that AI can make the said video.

That's only because you cannot see me.

[–] sebzim4500 1 day ago

>yes i know, but if you think you're no better than an LLM then why don't you put your personality into a system prompt and spare us the carbon footprint

There are still things I can do much better than AI, making videos just happens not to be one of them.

[–] freecodeio 20 hours ago

and your videos would still suck even with AI, what's your point

[–] sebzim4500 15 hours ago

Yes my videos would suck whether I use AI or not I really don't understand what you are trying to say

[–] v4dm 2 days ago

Impressive... at populating our visual world with more slop? Sure.

Would this be impressive in terms of convincing you to buy something or teach you something or entertain you?

Genuinely asking.

[–] thunky 1 day ago

> Would this be impressive in terms of convincing you to buy something or teach you something or entertain you?

Obviously yes. I mean have you seen what entertains us? And have you seen what a TV ad looks like? It's not a high bar and this is way above it.

Btw your comment is anti-ai slop. Slinging mud with no real criticism or even a personal observation to be found.