Mind blowind OpenAI's text to video SORA.

[BrianHG]

The visuals speak for themselves...



Go here to see all 48 videos with their prompt text: https://openai.com/sora

[tszaboo]

I've used text to picture AI models on my PC out of curiosity. The output was total garbage 99% of the time. Dall-e seems to produce better results out of the box, unless you need photorealistic images, but that's paywalled now. So when we see these short videos, yes, they look amazing. And I can totally believe that they weren't made just by typing up a prompt in 10 seconds and pressing go. I would bet on that all these videos are highly curated, and they spent maybe a 100 hours on a 10 second clip, telling the AI which part to redraw, maybe even frame by frame.

[Eagcress88]

I also think that there were very detailed prompts given there, which probably took more time to compose than if a person had drawn it all themselves. Even text-to-image generators sometimes still draw people with six fingers and so on.

[nctnico]

I saw the clips on a different site. Without going into the good and bad, let's dwell on the moment a little bit. IMHO it is mind blowing that such technology exists nowadays. My grandfather used to travel by steam trains and model-T Fords. Go figure.

[Bud]

The old man in the video just hated his burger 
They need to work more on generating happy faces 

[BrianHG]

[tom66]

I do seriously wonder if being a 3D artist, animator or even a semi-professional videographer will be a job any more in 2030.  This technology is incredible.

[Bud]

Cinema has never replaced Theatre, though many at the time were saying it will.

[BrianHG]

I do seriously wonder if being a 3D artist, animator or even a semi-professional videographer will be a job any more in 2030.  This technology is incredible.

If not by 2030, you can bet by 2040 that most except the most talented will either have to find new jobs, or become an expert at text prompting the AI to get the results you want to see.

On the other hand, this will become a god-send for directors.  They will just be fiddling/editing lines of text of their scene script into their AI generator until they get every seen looking exactly the way they want.  Hundreds of trials until it looks and feels like the final production scene they want.

On the other hand, now novelists may release new books and they might create a real life looking film simultaneously.

We will also see old scanned in comic books turned into real life movies, with almost no effort except for style cues.

[BrianHG]

Cinema has never replaced Theatre, though many at the time were saying it will.

Right you are.  Real life performances will always be a thing, though the big-scale production stuff will become harder to find, it will never disappear.  The live large scale stuff will be reserved for those who can afford it.

But, how much do you want to bet that there will be multiple live performance adaptations of of AI generated movies in the future.  I guarantee this will happen on a regular basis.

[BrianHG]

Go here to see all 48 videos with their prompt text: https://openai.com/sora

[BrianHG]

[Bud]

Watched the first 2 minutes and every single clip immediately feels fake.I do not know who that thing "shocked", it certainly did not shock me. There may be applications to it but the guy saying "it is so realistic" is talking BS.

[ebastler]

On the other hand, now novelists may release new books and they might create a real life looking film simultaneously.

What makes you so optimistic regarding the future need for human novelists? 
At least as far as Hollywood-style film scripts are involved, AI should be able to produce those nicely.

[SiliconWizard]

On the other hand, now novelists may release new books and they might create a real life looking film simultaneously.

What makes you so optimistic regarding the future need for human novelists? 

FTFY.

[RAPo]

Is it really, or are we blinded by the beatifull images?

[BrianHG]

On the other hand, now novelists may release new books and they might create a real life looking film simultaneously.

What makes you so optimistic regarding the future need for human novelists? 
At least as far as Hollywood-style film scripts are involved, AI should be able to produce those nicely.

Someone still has to come up with the prompts, for now anyways.  However, soon everything will be almost all AI generated and the human reaction will become the want to have some story from an authentic human.  Having human actors, actual locations and sounds will be reserved for live theatrical performances.

[berke]

All of this is going too fast.

What is the likelihood that the big players have training algorithms and/or hardware that is much more efficient than what has been revealed publicly?  I have a hard time believing that gradient descent-like algorithms (ADAM etc.), which are theoretically inefficient (much slower convergence than second derivative Newton methods) but fit in memory, are a good use of resources to train these humongous models.  Do they really just throw obscene amounts of computing power at these kinds of algorithms with no or very little secret sauce?  Or did they figure out how to run something with a higher rate of convergence on these big models?  Are they using some kind of analog hardware (not even talking quantum here)?

Also I wouldn't be surprised if TLAs already had 10 to 15 years ago technology comparable to what we have today, specifically something at least like OpenWhisper and for images maybe something similar to the first versions of Stable Diffusion.  After all we know they had voice keyword spotting systems in the 90s.

How good was the accuracy of classical, publicly known "non-AI" voice recognition algorithms back then?  It has to have been be pretty damn good to be usable for mass surveillance without triggering a deluge of false alarms that have to be cleared by tens of thousands of human operators. 

In other words, I think the AI we're seeing now has been recently technology-transferred to the public, however some of the secret sauce (for fast training) might still be held as secrets.  Heck, maybe OpenAI subcontracts the NSA to train their models on their massive farms and/or non-classical computers.  Someone call the WTO, I sense unfair government subsidies!

[KE5FX]

In other words, I think the AI we're seeing now has been recently technology-transferred to the public, however some of the secret sauce (for fast training) might still be held as secrets.  Heck, maybe OpenAI subcontracts the NSA to train their models on their massive farms and/or non-classical computers.  Someone call the WTO, I sense unfair government subsidies!

That's not realistic, given that the silicon needed to train these models takes up entire warehouses and didn't exist only a few years ago.

It's been a long time since the three-letter agencies had access to better tech than the public.

[berke]

That's not realistic, given that the silicon needed to train these models takes up entire warehouses and didn't exist only a few years ago.

What I'm saying is that they might have better algorithms or custom hardware.

It's been a long time since the three-letter agencies had access to better tech than the public.

If you're in a position to know you shouldn't be revealing such secrets!

[voltsandjolts]

This looks so much more complex than many engineering tasks, why can't AI do my PCB designs yet?

[BrianHG]

That's not realistic, given that the silicon needed to train these models takes up entire warehouses and didn't exist only a few years ago.

What I'm saying is that they might have better algorithms or custom hardware.

Custom hardware is being developed.  Including training and inference directly on the memory as well as analog memory with analog comparators, which wont be as fast, but also use magnitudes less power for the learning phase and inference phases.

This looks so much more complex than many engineering tasks, why can't AI do my PCB designs yet?

Nvidia and Google are already doing this for the chip design and advanced reports show much better designs than humans working with pre-fab cell libraries.  However, I'm not sure a 4090 has enough juice to do this as I'm sure it was done on a server rack of their top tier processors.

[BrianHG]

     I can say one thing, in the future, if someone goes through the trouble to film a real authentic live stunt they've been planning for months and invested thousands of dollars to complete will be written off as AI generated.

     Such filmed stunts or events, even actual crime scene security camera footage will never have any credibility to them and such filmed stunts will be a thing of the past.  (Which also wont be believed since everyone will say that old filmed stunt was AI generated and its release date was faked...)

[SiliconWizard]

Is it really, or are we blinded by the beatifull images?

It's more or less the second option, but it may not make a lot of difference for many. We'll also probably grow into not seeing the difference, or even knowing what "real" art creation is, or even what reality is altogether.

One important aspect of this IMO is the fact that machine-generated content does not confront us to our own inability in the common way, that is by measuring ourselves to others (that's until we grow into measuring ourselves to machines, if it ever happens, which I don't know, but doubt for now.) In other words, we stop potentially feeling inferior to some other people who are able to do stuff we are not able to. While that can be a trigger for actually willing to improve ourselves, and has been, I'm afraid we'll more easily fall into the comfort of not doing anything to go out of our way, and not feel anything about it, because we'll have no one else to compare to, if we all become just passive consumers of everything.

[Alex Eisenhut]

and for the immature among us ... like me... (maybe nsfw?)
https://youtu.be/fY1EIArdsto
https://youtu.be/4LlV3_wvUE8
https://youtu.be/PiEow9uEPsU
https://youtu.be/6emCWRSOUuY
it's soooo stupid, I love it