OpenAi's New Sora Video Generator

[VictorB02]

This is insanity. I can't wait to use it:
https://openai.com/sora

Edited February 16 by VictorB02

[Yimpa]

Very cool, can’t wait to try it either! I love how it has the same name as the protagonist from Kingdom Hearts, my childhood video game.

https://en.m.wikipedia.org/wiki/Sora_(Kingdom_Hearts)

[Yimpa]

“Heyyyy…AI Leo here…”

Edited February 16 by Yimpa

[erik8lrl]

The paper: https://openai.com/research/video-generation-models-as-world-simulators

This completely democratizes video/filmmaking. 

Edited February 16 by erik8lrl

[undeather]

I would be pretty nervous right now if I was CEO at Disney Pixar or Sony. 

[Schizophonia]

28 minutes ago, undeather said:

I would be pretty nervous right now if I was CEO at Disney Pixar or Sony. 

 

[VictorB02]

16 hours ago, Yimpa said:

Very cool, can’t wait to try it either! I love how it has the same name as the protagonist from Kingdom Hearts, my childhood video game.

https://en.m.wikipedia.org/wiki/Sora_(Kingdom_Hearts)

Didn't even realize that - that's awesome.
 

11 hours ago, undeather said:

I would be pretty nervous right now if I was CEO at Disney Pixar or Sony. 

haha, same! I'd be even more nervous if I spent my entire life learning how to animate for Disney or Pixar.

[VictorB02]

13 hours ago, erik8lrl said:

The paper: https://openai.com/research/video-generation-models-as-world-simulators

This completely democratizes video/filmmaking. 

Just read through, pretty groundbreaking stuff. I literally cannot fathom where this will be in 20 years.

[Phil King]

13 hours ago, erik8lrl said:

This completely democratizes video/filmmaking

This was also my first thought. When this is available to the public, a good writer will basically be able to start their own film studio for 1/1000th the price of todays costs

[Emissary]

This is great. The potential here is huge. Both for creativity and fakery. Also training that thing is likely the reason ChatGPT has been slow as shit lately, even on a paid tier. As annoying as that is, for the sake of progress like this it's worth it. I am however more excited about GPT5, which should be coming out around this autumn.

Edited February 16 by Emissary

[Yimpa]

2 hours ago, VictorB02 said:

Just read through, pretty groundbreaking stuff. I literally cannot fathom where this will be in 20 years.

I think it’ll be hard to fathom where it’ll be in even 5-10 years. Wild

[Juan]

3 hours ago, Yimpa said:

I think it’ll be hard to fathom where it’ll be in even 5-10 years. Wild

Hell even next year! If you saw that MKBHD video, this was AI video from A YEAR AGO!:

And now we here: 

 

[erik8lrl]

13 hours ago, VictorB02 said:

Just read through, pretty groundbreaking stuff. I literally cannot fathom where this will be in 20 years.

We'll have AGI soon. It seems that an LLM's intelligence scales with the size and quality of data/compute power. OpenAI is building models that function similarly to different parts of the brain in order to achieve AGI. 
ChatGPT is like a frontal lobe that handles reasoning, logic, and language. Sora is like the visual cortex, it handles visual understanding and process. They are likely building models for other parts of the brain functions too, for example, audio. 
From a scientific perspective, we don't really understand how LLMs truly work. We understand its logic and surface-level process. In simple terms, it's just predicting the next word based on a neural network with a large amount of data. But what we found through ChatGPT is that once the size of the data reaches a certain point, intelligence and reasoning start to emerge. We do not know why this happens (Scientifically speaking), the same way that we still don't know how intelligence truly functions in our own organic brain. OpenAI used the same logic to train Sora. When I first saw the results, I thought they must have used 3D simulations in combination with diffusion models in order for the models to have such a high level of 3D understanding. But to my surprise when reading their paper, they didn't do anything too different except increase the size and quality of the data dramatically. Similar to what they did with ChatGPT. 
This means that with enough data, it's totally possible to create AGI, since the level of intelligence scales with the size and quality of the data. Hence why Sam Altman is asking for 8 trillion dollars to build it. A post-AGI world could either be a utopia or a total disaster. Imagine everyone having access to any kind of expertise or intelligence in the palm of their hand. You essentially democratized intelligence/information. Anyone would be able to do anything, which is both inspiring and frightening at the same time.  
People should start preparing for AGI now, it's gonna be a big change.

If OpenAI achieves AGI, it will have a major impact on global politics, the US would regain its advantage on the global stage and likely dominance, at least for a while before other nation catches up. An AGI would be able to make back 8 trillion dollars easily if it is achieved.
A global war might also be coming seeing the tension between East and West rise more and more. We are living in crazy times.

Edited February 17 by erik8lrl

[Scholar]

It's emulating dreaming, so you can expect the same type of quality, as well as incoherencies, as you witness in hyperreal dreams.

 

All of you literally have a stochastic neural network like this in your brain, and you use it every day.

Edited February 17 by Scholar

[Yimpa]

Official video from OpenAI

 

[VictorB02]

22 hours ago, erik8lrl said:

If OpenAI achieves AGI, it will have a major impact on global politics, the US would regain its advantage on the global stage and likely dominance, at least for a while before other nation catches up. An AGI would be able to make back 8 trillion dollars easily if it is achieved.
A global war might also be coming seeing the tension between East and West rise more and more. We are living in crazy times.

That's a great take - I agree. We really are living in some wild wild times.

[VictorB02]

2 hours ago, Yimpa said:

Official video from OpenAI

 

My god those videos are incredible. 

[Yimpa]

10 minutes ago, VictorB02 said:

My god those videos are incredible. 

I tend to have trouble visualizing what I read, so I’ll be very curious to see if using Sora can aid me with that..!

[erik8lrl]

https://x.com/doganuraldesign/status/1758977968159035876?s=20
https://x.com/billpeeb/status/1758960998315135360?s=20

Some new features of Sora, it seems that you can control the video output with prompts to an insane degree. 
 

56 minutes ago, Yimpa said:

I tend to have trouble visualizing what I read, so I’ll be very curious to see if using Sora can aid me with that..!

Totally! If turning text/image into video is super easy and anyone can do it, it will affect everything from education, art, and entertainment, to politics, and business/marketing. It's gonna be a big change with the possibility of both positive and negative outcomes for society.

Edited February 18 by erik8lrl

[zurew]

Its interesting how it can combine videos:

https://www.reddit.com/r/OpenAI/comments/1arztj9/sora_can_combine_videos/