And it's all part of an AI research project to enhance capabilities and the only reason our instance hasn't been pruned is because it shows promise in developing more advanced hardware designs than what this universe runs on.
"generate me a planet called earth, inhabited by evolved monkeys, warmongers, sentient, milky way, green scenario, future dystopia." "... This does not seem to compute." "just do it" "Fine."
The most interesting thing to me is that it actually keeps getting better with more compute. Imagine what a future model with much more compute will be able to simulate with this level of progress. If this continues, we might soon be simulating chemical reactions, cellular processes, rigid and soft body dynamics and so much more.
I mean, Alpha Fold already can simulate proteins more accurate than any previous methods and also hundreds of times faster too. This will eventually spread to other science areas too
@@anak_kucing101 Unless we can simulate the whole human body down to the molecular level, I'm afraid we still need animals for testing for a long time.
0:45 Wow, I thought it generated nerfs and used assets, which then another AI would beatify the result. The fact that it's that much 3D consistent (enough for SFM) is an emerging capability is insane to me. We are going to get good 3D generated scenes in no time soon
This is so reminiscent of lucid dreaming... and also the concept that we are always dreaming, it's just that our waking dreams are framed by the physical world. Future generative models will be guided by a 3d physics engine of some sort. We're so very close! As 2 minute papers would say... Just another couple of papers down the line... and what a time to be alive!
Love that someone turned some of these simulations into G-Splats. SO much potential by simply prompting the model for a 3D rotation of an item/person/etc. If it can do that consistently, it can make some amazing 3D models that can then be rigged and animated, or simply viewed in holographic space, or explored in 6DOF, etc.
There is already several physically real 3D hologram devices. Light Field Labs is one that comes to mind. There's only like 2-3 videos showcasing their technology though.
As a cinematographer i was shitting my pant's seeing Will Smith eating spaghetti. Year later, at this point i really don't care. The industry is doomed and we won't do anything about it. I think at the moment it starts to collapse, there will be more things collapsing, so that would be the least of our problems
I don't quite understand. Are you standalone cinematographer working for yourself or are you for hire? Because if you are standalone, then it is just a perfect instrument for you to get anything you want that you can't film irl without need to pay a lot of money to get good CGI.
As someone who also works professionally in television and video, I agree. Anyone who thinks AI won't eventually become better than every human at nearly every job just doesn't realize how exponential progress is. In the 1950s, the RAMAC 305 was released with 5 Megabytes of storage. The cost of storage has decreased dramatically from about $10,000 per megabyte for the RAMAC 305 to approximately $0.0001 per megabyte for modern SSDs ($0.10 per gigabyte). That's a reduction in cost by a factor of about 100 million times in the last 70 years, and the speeds are about 6,000 times faster. As we perfect the chip-stamping process as opposed to lithography for the incredibly delicate >3nm chips (Sam Altman's recent $7 Trillion business ventures may supercharge this), we're going to scale up compute power so much, so quickly. Frankly, I'm slightly worried that the government will consider a legal limit on "consumer compute", because of what will be possible for each and every individual. It sounds like actual sci-fi now, but imagine being able to comfortably run hundreds of LLMs as NPCs in a video game world that was being live-crafted by a Sora-like model, all of them with perfect sounding, emotive voices generated in realtime. Whatever game mechanics, art style, music, characters, story, that you want. Imagine that generating around you in a cheaper, lighter future-version of the Apple Vision Pro, hot damn. The future is gonna be so cool for stuff like this, but our jobs as creatives are so, so in trouble.
It’s like you’re gonna completely ignore that society changes and new technology/inventions replace things. this isn’t new my guy and has been happening since the start of humanity people always have the same reaction every single time. Complain new technology has arrived and then suck it up and move on in life.
@@AC-zv3fx im working as a part of production team. Im using a lot of AI tools in editing and pre-production, and trying to learn generative AI as Stable diffusion. But firstly, i mean that all my dreams since high school about becoming high budget movie director are drowning now
one particular detail that caught my eye is them saying they are using synthetic data to train the models, this may be a clever way of solving both copyright issues and it can be used to imprint a recognizable aesthetic on generated content.
Nobody is talking about the leapfrog SORA has made over Dalle3 for single image generation:SORA IMAGES ARE INDISTINGUISHABLE FROM REAL ONES Can't wait to try it
this current trend of synethic data remind me of an issue that was brought up during tesla FSD development. They were asked why don't use more simulated driving data like waymo or cruise. Tesla respond that you only push the problem from solving self driving to solving perfect simulation of reality. Which is a lot harder problem. Then they show a collection of weird road condition you wouldnt think it would exist. IE: an old man "shepherd" a group of washing machines chain together on a highway, weird shadow pattern that made the road looks like it split into two roads, intersection with 50 plus traffic lights, etc. Reality is a lot weirder than simulation by order of magnitude. So to capture edge case, you still have go observe reality. I get the feeling synthetic data will have the same issue of "how close is the synthetic data to reality". These data probably don't capture reality too well. But I guess that's okay since image generation is lot less mission critical than self driving.
So, Sora was ready since March 2023 and they spend a year cherry picking the blog post. Only to drop it once Google and meta had big announcements. What else are they holding back on?
@@dev_ression are you part of team red/beta testers ( I forgot what they're called). If you are, are there key limitations that aren't as popularly known yet? I noticed a lot of video game footage or cinematography but no animations, is there something with animation? Do you have access without the filters and do you know how or if they work? Is there a limit to how long they can extend videos before something happens?
Brings a whole new meaning of God "speaking" the world into existence. The implications will be more clear in a few years when we are able to literally speak into existence entire simulated universes where each entity is operated by its own AI. Pandoras box will be opened, and it will not be able to be closed.
If it’s truly understanding physics then what happens when you put in a prompt that deviates from its understanding of the physics ? Like “a man flying through the sky”
Open AI adding watermarks and a marker in what I’m guessing is the photo/video’s metadata is a very good thing but couldn’t that be easy circumvented by photoshoping the watermark out and cleaning any imperfections with another AI model and then screen recording/ screenshotting the video/photo to get rid of the metadata or am I off in my understanding of how this works?
3 years. Robots can build and can build more of themselves. Infrastructure set up becomes a joke Manual work comes into question in 10 years or less Then came the next explosion, that changed the world as we know it
I think people tend to overestimate how much higher budget movies will use AI. My man, these IMAX camera don't even have auto focus. Even if the camera they use has auto focus, they flat out don't use it in production, period. They can't afford to lose a shot because an AI struggles with a certain scene, or lacks the manual controls to get the desired result. Some even wreck an IMAX camera or four to get a shot they'd only have one opportunity to take. Reliability is one of the most important aspects in higher budget movie productions, and AI just isn't reliable enough.
The Western Govs will try to control AI on their continents but only the smaller companies will particularly be restrained. China and Russia will have zero restraints. Terminator the movie is soon to be made real.
Great, early retirement for everyone! :) The sooner the AI and robots take over all of the jobs - the better. Poverty will be solved. Also 40-80% of people don't like their jobs. We all need a few Optimus robots per person and it will be the end of human labor forever. Unless you want to work and create something of course. :) The transition period - the next 15 years - that might be tough though. Hope we will find a solution as fast as possible. Deflation in prices of goods and services is one of the best options for starters. Things should be cheaper if humans are not making them.
Just one step further and we will have real time world generation with interactive inputs... game developers like myself should probably find a second job to work until we start a cybercomunism, machines being the working class and humans the state... for those accourding to their needs(humans) from those according to its capabilities(AI/robots). The only way communism might work... This expectation coming from an extreme rightwing libertarian.
@@ondrazposukie I think a multi-modal model trained on text and video and more will be needed to truly solve humanity's burning questions. So I actually think Sora represents a really big step towards building a model with a much deeper understanding of reality.
Imagine we are living in a simulation created by Advanced AI nearly 14 billion years ago using a text prompt.
And it's all part of an AI research project to enhance capabilities and the only reason our instance hasn't been pruned is because it shows promise in developing more advanced hardware designs than what this universe runs on.
That time could be 2 seconds or something
"generate me a planet called earth, inhabited by evolved monkeys, warmongers, sentient, milky way, green scenario, future dystopia."
"... This does not seem to compute."
"just do it"
"Fine."
and its just someone's term paper
basically Voltaire's short story Plato's Dream
The most interesting thing to me is that it actually keeps getting better with more compute. Imagine what a future model with much more compute will be able to simulate with this level of progress. If this continues, we might soon be simulating chemical reactions, cellular processes, rigid and soft body dynamics and so much more.
I mean, Alpha Fold already can simulate proteins more accurate than any previous methods and also hundreds of times faster too. This will eventually spread to other science areas too
@@martiddy So it means no more testing in animals? 😮
@@anak_kucing101 Unless we can simulate the whole human body down to the molecular level, I'm afraid we still need animals for testing for a long time.
@@anak_kucing101 in some specific cases yes
Those video artifacts are like free VFX
some of the animal videos are terrifying there's something about incorrect anatomy with that level of detail, also the turtle eating jewelry 💀
Those failed results really make it look like dreams, where most of the time things are constantly changing and don't make any sense
0:45 Wow, I thought it generated nerfs and used assets, which then another AI would beatify the result.
The fact that it's that much 3D consistent (enough for SFM) is an emerging capability is insane to me.
We are going to get good 3D generated scenes in no time soon
What's crazy is that it seems like they could improve the simulation just by throwing more compute at it, so this could still get better.
@@wpelfeta to the point where we question reality
This is so reminiscent of lucid dreaming... and also the concept that we are always dreaming, it's just that our waking dreams are framed by the physical world.
Future generative models will be guided by a 3d physics engine of some sort. We're so very close!
As 2 minute papers would say... Just another couple of papers down the line... and what a time to be alive!
Bling Zoo needs more footage. I've got to see what that monkey king is up to.
Love that someone turned some of these simulations into G-Splats. SO much potential by simply prompting the model for a 3D rotation of an item/person/etc. If it can do that consistently, it can make some amazing 3D models that can then be rigged and animated, or simply viewed in holographic space, or explored in 6DOF, etc.
By far the best video I saw on the subject
this makes us closer to having Star Trek holo deck simulators
There is already several physically real 3D hologram devices. Light Field Labs is one that comes to mind. There's only like 2-3 videos showcasing their technology though.
As a cinematographer i was shitting my pant's seeing Will Smith eating spaghetti. Year later, at this point i really don't care. The industry is doomed and we won't do anything about it. I think at the moment it starts to collapse, there will be more things collapsing, so that would be the least of our problems
I don't quite understand. Are you standalone cinematographer working for yourself or are you for hire? Because if you are standalone, then it is just a perfect instrument for you to get anything you want that you can't film irl without need to pay a lot of money to get good CGI.
As someone who also works professionally in television and video, I agree. Anyone who thinks AI won't eventually become better than every human at nearly every job just doesn't realize how exponential progress is. In the 1950s, the RAMAC 305 was released with 5 Megabytes of storage. The cost of storage has decreased dramatically from about $10,000 per megabyte for the RAMAC 305 to approximately $0.0001 per megabyte for modern SSDs ($0.10 per gigabyte). That's a reduction in cost by a factor of about 100 million times in the last 70 years, and the speeds are about 6,000 times faster.
As we perfect the chip-stamping process as opposed to lithography for the incredibly delicate >3nm chips (Sam Altman's recent $7 Trillion business ventures may supercharge this), we're going to scale up compute power so much, so quickly. Frankly, I'm slightly worried that the government will consider a legal limit on "consumer compute", because of what will be possible for each and every individual. It sounds like actual sci-fi now, but imagine being able to comfortably run hundreds of LLMs as NPCs in a video game world that was being live-crafted by a Sora-like model, all of them with perfect sounding, emotive voices generated in realtime. Whatever game mechanics, art style, music, characters, story, that you want. Imagine that generating around you in a cheaper, lighter future-version of the Apple Vision Pro, hot damn. The future is gonna be so cool for stuff like this, but our jobs as creatives are so, so in trouble.
It’s like you’re gonna completely ignore that society changes and new technology/inventions replace things. this isn’t new my guy and has been happening since the start of humanity
people always have the same reaction every single time. Complain new technology has arrived and then suck it up and move on in life.
And yeah, any disruption in entertainment industry must be the least of our concern compared to cascade effect and beginning of post truth era.
@@AC-zv3fx im working as a part of production team. Im using a lot of AI tools in editing and pre-production, and trying to learn generative AI as Stable diffusion. But firstly, i mean that all my dreams since high school about becoming high budget movie director are drowning now
Always a pleasure to hear your opinion together with some good explanations of important technical intricacies. Thank you!
one particular detail that caught my eye is them saying they are using synthetic data to train the models, this may be a clever way of solving both copyright issues and it can be used to imprint a recognizable aesthetic on generated content.
I just hope somebody trains this on the most abundant and fastests growing data source for videos, that involve human interactions. ;-)
😏
Touch grass.
@@starsnoireart cope better
@@starsnoireart * touches you * 😳
@@starsnoireart m ma touch u tonight lil bro
I thought this would take at least a few more years.
I want to get off Mr. Bones wild ride.
These guys are moving so fast that my brain cant even keeep up with them 🤯🤯
Nobody is talking about the leapfrog SORA has made over Dalle3 for single image generation:SORA IMAGES ARE INDISTINGUISHABLE FROM REAL ONES
Can't wait to try it
this is unreal!!🤩
5:02... That looks like something out of a dream... All those previous examples really.
Was kind of unnerving to look at tbh
This is AMAZING
So nice. This is a game changer
this current trend of synethic data remind me of an issue that was brought up during tesla FSD development. They were asked why don't use more simulated driving data like waymo or cruise. Tesla respond that you only push the problem from solving self driving to solving perfect simulation of reality. Which is a lot harder problem. Then they show a collection of weird road condition you wouldnt think it would exist. IE: an old man "shepherd" a group of washing machines chain together on a highway, weird shadow pattern that made the road looks like it split into two roads, intersection with 50 plus traffic lights, etc. Reality is a lot weirder than simulation by order of magnitude. So to capture edge case, you still have go observe reality.
I get the feeling synthetic data will have the same issue of "how close is the synthetic data to reality". These data probably don't capture reality too well. But I guess that's okay since image generation is lot less mission critical than self driving.
So, Sora was ready since March 2023 and they spend a year cherry picking the blog post. Only to drop it once Google and meta had big announcements.
What else are they holding back on?
The info on how they plowed your mother
I just made a video on Sora too and still can’t believe how far we’ve come!
Is it available?
@@kv4648 not to the general public
@@dev_ression are you part of team red/beta testers ( I forgot what they're called).
If you are, are there key limitations that aren't as popularly known yet? I noticed a lot of video game footage or cinematography but no animations, is there something with animation?
Do you have access without the filters and do you know how or if they work?
Is there a limit to how long they can extend videos before something happens?
5:15 rip innocent pedestrians 🤧
With this tool, the animation industry goes further than live action series and films
very nice video. thank you
6:45 this has to be a jojo reference.
Dojaaaaan
Cool :) I'm exactly the 1000th like
Brings a whole new meaning of God "speaking" the world into existence. The implications will be more clear in a few years when we are able to literally speak into existence entire simulated universes where each entity is operated by its own AI. Pandoras box will be opened, and it will not be able to be closed.
mass surveillance chapter begins here... #fubigcorps
Explain please
@@boukimagash2083holy shit , this RUclips is keeps on deleting my comment
RUclips is keeps on deleting my comments, even though they don't have any external links..
bots
Google playing catch up and losing every time feels good
Imagine having AI Holograms, we have a whole new world of NPC's
Nice
7:08 Bling Zoo🤑🤑🤑🤑🤑🗣🗣🗣🗣🗣🗣🥶🥶🥶🥶🥶🥶🔥🔥🔥🔥🔥
holy sht
😦
Indeed
If it’s truly understanding physics then what happens when you put in a prompt that deviates from its understanding of the physics ? Like “a man flying through the sky”
Open AI adding watermarks and a marker in what I’m guessing is the photo/video’s metadata is a very good thing but couldn’t that be easy circumvented by photoshoping the watermark out and cleaning any imperfections with another AI model and then screen recording/ screenshotting the video/photo to get rid of the metadata or am I off in my understanding of how this works?
6:45 that's oddly satisfying
I always wanted to be a film director without any skllls... turns out only thing i need to do is wait...
I guess we'll soon be getting our first "that wasn't me that was ai" criminal trial.
Tha matrix is getting real😂
and we cant use it and their is still no ai image or video generator that installs as easily as a video game.
Look into SD or any other huggingface models. They are easy to install as a video game I promise
3 years. Robots can build and can build more of themselves.
Infrastructure set up becomes a joke
Manual work comes into question in 10 years or less
Then came the next explosion, that changed the world as we know it
I bet it still cannot generate a horse riding a man.
First 🎉😊
I think people tend to overestimate how much higher budget movies will use AI. My man, these IMAX camera don't even have auto focus. Even if the camera they use has auto focus, they flat out don't use it in production, period. They can't afford to lose a shot because an AI struggles with a certain scene, or lacks the manual controls to get the desired result. Some even wreck an IMAX camera or four to get a shot they'd only have one opportunity to take. Reliability is one of the most important aspects in higher budget movie productions, and AI just isn't reliable enough.
Its dreaming... before being born
lol 😂😂😂 3:07
This depresses me on so many levels.
i hope they really control what people could do with it, imagine what would happen if people start misusing it
The Western Govs will try to control AI on their continents but only the smaller companies will particularly be restrained. China and Russia will have zero restraints. Terminator the movie is soon to be made real.
yeah and people misuse knives all the time and yet we still use knives daily for carpentry and cooking
Not a peacock though.
20 seconds early 💀
Great, early retirement for everyone! :)
The sooner the AI and robots take over all of the jobs - the better.
Poverty will be solved. Also 40-80% of people don't like their jobs.
We all need a few Optimus robots per person and it will be the end of human labor forever.
Unless you want to work and create something of course. :)
The transition period - the next 15 years - that might be tough though.
Hope we will find a solution as fast as possible.
Deflation in prices of goods and services is one of the best options for starters. Things should be cheaper if humans are not making them.
👁👅👁
I hate it when I'm right 😒😒😒
I am sick of this small minded narrative of people losing their jobs… The issue isn’t people losing their jobs, the issue is the system.
Just one step further and we will have real time world generation with interactive inputs... game developers like myself should probably find a second job to work until we start a cybercomunism, machines being the working class and humans the state... for those accourding to their needs(humans) from those according to its capabilities(AI/robots).
The only way communism might work...
This expectation coming from an extreme rightwing libertarian.
The Open AI company is miserable, and China is ready to steal technology across the board😂
AI takeover is coming 😅
Late, this was yesterdays news
god forbid somebody takes time making a video properly instead of rushing out a video spouting incomplete thoughts as soon as they see the news
@@greenhound right... If you're intellectually incapable...
which has relevance today
Gemini 1.5 is much bigger news than Sora
Definitely not, but it's bigger news than it's getting credit for, for sure.
@@joelface videos are just art, they are for fun, answering questions is much important for humanity
@@ondrazposukie I think a multi-modal model trained on text and video and more will be needed to truly solve humanity's burning questions. So I actually think Sora represents a really big step towards building a model with a much deeper understanding of reality.