&gt; All videos on this page were generated directly by Sora without modification.I hope there is at least some cherrypicking here. 
This also seems like some shots fired at some of the other gen video startups

If we go from DALL-E 3, it won&#x27;t be nowhere near competitive while they have the superior ground. Generating a high quality 1024x1024 image with costs around ~$0.002, but $0.08 on DALL-E 3 (20x more expensive per-image). For videos with very high computational needs (since each frame needs to be temporally consistent, you need huge GPUs to serve this) I&#x27;m expecting this to be so much more expensive than its competitors (Pika or SVD1.1)

samples look amazing , Looking forward for access, and hope they price it competitively

yeah I think a tech company showing how their tech can be used to cause damage to a humanities field as one of their leading product demos is bad

Yes, but the point is that in a few years, there won&#x27;t be a difference. Those clickbait accounts already exist for AI generated images. How many impressionable or young people have been fooled into believing history that never happened?More importantly, how can these accounts subtly direct the generations to instill modern ideology or politics into &quot;historical&quot; images, giving them historical credibility? Think of all the subtly white supremacist &quot;retvrn&quot; accounts, for example, falsely recontextualizing inventions and accomplishments to support their ideology.We all need to be thinking much more creatively and cynically about how these tools will be abused. The technology will get better. The people who want to abuse it will get smarter. And your capability to distinguish fake information is likely much worse than you believe - to say nothing of younger people who have less context and experience to form a mental &quot;immune system&quot;.

It&#x27;s a test prompt to demo the model, not a clickbait social media post.

Yeah, my heart sank when I saw that.Social media is really good at separating content from context, things like this will distort people&#x27;s understanding of history.

I think most people consider &quot;history&quot; to mean &quot;things that have actually happened&quot; and not &quot;the aesthetic of the past&quot; as you seem to be suggesting.

only if you consider &quot;historical footage&quot; to exclusively mean the &quot;[original] historical footage [stored in archiving]&quot; versus e.g. &quot;historical[ly accurate] footage&quot;if &quot;historical&quot; is going to be used subjectively with no further qualifying statements then the meaning of &quot;history&quot; will be subjucated to the context it&#x27;s being presented in, I don&#x27;t see it&#x27;s use here as contradictory

&gt; Prompt: Historical footage of California during the gold rush.this is the opposite of history

Just lots of videos from Youtube probably.

Amazing.One wonders how you might gain a representation of physics learned in the model. Perhaps multimodal inputs with rendered objects; physics simulations?

These samples look pretty amazing. I&#x27;m curious the compute required to train and even deploy something like this. How would it scale to making something like a CGI Pixar movie?

I love the downvote, could be an OpenAI employee in the know.

Pretty sure you plus tier not be using this free, too much processing power needed

Totally a coincidence that it&#x27;s announced immediately after the new Gemini reveal.

AGI at the quality of sora or dalle but for intelligence is gonna be quite the thing to witness

Eh. MSFT owns 49% of OpenAI. Doesn&#x27;t really seem like they need to do much except support them.

Microsoft doesn’t have access to OpenAI’s research, this was part of the deal. They only have access to the weights and inference code of production models and even then who has access to that inside MS is extremely gated and only a few employees have access to this based on absolute need to actually run the service.AI researcher at MSFT barely have more insights about OpenAI than you do reading HN.

Many people say the same about Google&#x2F;DeepMind.

Agreed.  It&#x27;s amazing how much of a head start OpenAI appears to have over everyone else.  Even Microsoft who has access to everything OpenAI is doing.  Only Microsoft could be given the keys to the kingdom and still not figure out how to open any doors with them.

The video of the gold rush town just makes me think of what games like Red Dead and GTA could look like.

The diffusion is almost certainly taking place over some sort of compressed latent, from the visual quirks of the output I suspect that the process of turning that latent into images goes latent -&gt; nerf &#x2F; splat -&gt; image, not latent -&gt; convolutional decoder -&gt; image

I&#x27;m almost speechless. I&#x27;ve been keeping an eye on the text-to-video models, and if these example videos are truly indicative of the model, this is an order of magnitude better than anything currently available.In particular, looking at the video titled &quot;Borneo wildlife on the Kinabatangan River&quot; (number 7 in the third group), the accurate parallax of the tree stood out to me. I&#x27;m so curious to learn how this is working.[Direct link to the video: <a href="https:&#x2F;&#x2F;player.vimeo.com&#x2F;video&#x2F;913130937?h=469b1c8a45" rel="nofollow">https:&#x2F;&#x2F;player.vimeo.com&#x2F;video&#x2F;913130937?h=469b1c8a45</a>]

Must be intimidating to be on the Pika team at the moment...

It&#x27;s movies the shots are way to deliberate to have random YouTube crap in the dataset.

Where is the training material for this coming from? The only resource I can think of that&#x27;s broad enough for a general purpose video model is YouTube, but I can&#x27;t imagine Google would allow a third party to scrape all of YT without putting up a fight.

In terms of following the prompt and generating visually interesting results, I think they&#x27;re comparable. But the resolution for Sora seems so far ahead.Worth noting that Google also has Phenaki [0] and VideoPoet [1] and Imagen Video [2][0] <a href="https:&#x2F;&#x2F;sites.research.google&#x2F;phenaki&#x2F;" rel="nofollow">https:&#x2F;&#x2F;sites.research.google&#x2F;phenaki&#x2F;</a>[1] <a href="https:&#x2F;&#x2F;sites.research.google&#x2F;videopoet&#x2F;" rel="nofollow">https:&#x2F;&#x2F;sites.research.google&#x2F;videopoet&#x2F;</a>[2] <a href="https:&#x2F;&#x2F;imagen.research.google&#x2F;video&#x2F;" rel="nofollow">https:&#x2F;&#x2F;imagen.research.google&#x2F;video&#x2F;</a>

The big stand out to me beyond almost any other text video solution is that the video duration is tremendously longer (minute+). Everything else that I&#x27;ve seen can&#x27;t get beyond 15 to 20 seconds at the absolute maximum.

It&#x27;s very good. Unclear how far ahead of Lumiere it is (<a href="https:&#x2F;&#x2F;lumiere-video.github.io&#x2F;" rel="nofollow">https:&#x2F;&#x2F;lumiere-video.github.io&#x2F;</a>) or if its more of a difference in prompting&#x2F;setttings.

The examples are most certainly cherry-picked. But the problem is there are 50 of them. And even if you gave me 24 hour full access to SVD1.1&#x2F;Pika&#x2F;Runway (anything out there that I can use), I won&#x27;t be able to get 5 examples that match these in quality (~temporal consistency&#x2F;motions&#x2F;prompt following) and more importantly in the length. Maybe I am overly optimistic, but this seems too good.

<a href="https:&#x2F;&#x2F;twitter.com&#x2F;sama&#x2F;status&#x2F;1758200420344955288" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;sama&#x2F;status&#x2F;1758200420344955288</a>They&#x27;re literally taking requests and doing them in 15 minutes.

It doesn&#x27;t matter if they&#x27;re cherrypicked when you can&#x27;t match this quality with SD or Pika regardless of how much time you had.and i still prefer Dalle-3 to SD.

In the past the examples tweeted by OpenAI have been fairly representative of the actual capabilities of the model. i.e. maybe they do two or three generations and pick the best, but they aren&#x27;t spending a huge amount of effort cherry-picking.

&gt; Stable Diffusion is still the go-to solution. I strongly suspect the same thing with Sora.Sure, for people who want detailed control with AI-generated video, workflows built around SD + AnimateDiff, Stable Video Diffusion, MotionDiff, etc., are still going to beat Sora for the immediate future, and OpenAI&#x27;s approach structurally isn&#x27;t as friendly to developing a broad ecosystem adding power on top of the base models.OTOH, the basic simple prompt-to-video capacity of Sora now is good enough for some uses, and where detailed control is not essential that space is going to keep expanding -- one question is how much their plans for safety checking (which they state will apply both to the prompt and every frame of output) will cripple this versus alternatives, and how much the regulatory environment will or won&#x27;t make it possible to compete with that.

they&#x27;re not fantastic either if you pay close attentionthere are mini-people in the 2060s market and in the cat one an extra paw comes out of nowhere

Stable diffusion is not the go-to solution, it&#x27;s still behind midjourney and DAllE

Would love to see handpicked videos from competitors that can hold their own against what SORA is capable of

While Sora might be able to generate short 60-90 second videos, how well it would scale with a larger prompt or a longer video remains yet to be seen. 
And the general logic of having the model do 90% of the work for you and then you edit what is required might be harder with videos.

Wrong, this is the first time I&#x27;ve seen an astronaut with a knit cap.

The year is 2030.Sarah is a video sorter, this was her life. She graduated top of her class in film, and all she could find was the monotonous job of selecting videos that looked just real enough.Until one day, she couldn&#x27;t believe it. It was her. A video of of her in that very moment sorting. She went to pause the video, but stopped when he doppelganger did the same.

Look at Sam altman’s twitter where he made videos on demand from what people prompted him

Let&#x27;s hold our breath. Those are specifically crafted hand-picked good videos, where there wasn&#x27;t any requirement but &quot;write a generic prompt and pick something that looks good&quot;, with no particular requirements. Which is very different from the actual process where you have a very specific idea and want the machine to make it happen.DALL-E presentation also looked cool and everyone was stoked about it. Now that we know of its limitations and oddities? YMMV, but I&#x27;d say not so much - Stable Diffusion is still the go-to solution. I strongly suspect the same thing with Sora.

It is incredible indeed, but I remember there was a humongous gap between the demoed pictures for DALL-E and what most prompts would generate.Don&#x27;t get overly excited until you can actually use the technology.

I suspect it&#x27;s less about being puritanical about violence and nudity in and of themself, and more a blanket ban to make up for the inability to prevent the generation of actually controversial material (nude images of pop stars, violence against politicians, hate speech)

It&#x27;s not a particularly American attitude to be opposed to violence in media though, American media has plenty of violence.They&#x27;re trying to be all-around uncontroversial.

I am guessing a movie studio will get different access with controls dropped. Of course, that does mean they need to be VERY careful when editing, and making sure not to release a vagina that appears for 1 or 2 frames when a woman is picking up a cat in some random scene.

I wonder how much of it is really &quot;concern for the children&quot; type stuff vs not wanting to deal with fights on what should be allowed and how and to who right now. When film was new towns and states started to make censorship review boards. When mature content became viewable on the web battles (still ongoing) about how much you need to do to prevent minors from accessing it came up. Now useful AI generated content is the new thing and you can avoid this kind of distraction by going this route instead.I&#x27;m not supporting it in any way, I think you should be able to generate and distribute any legal content with the tools, but just giving a possible motive for OpenAI being so conservative whenever it comes to ethics and what they are making.

I&#x27;ve been watching 80s movies recently, and amount of nudity and sex scenes often feels unnecessary. I&#x27;m definitely not a prude. I watch porn, I talk about sex with friends, I go to kinky parties sometimes. But it really feels that a lot of movies sacrificed stories to increase sex appeal — and now that people have free and unlimited access to porn, movies can finally be movies.

Yes, but I am stuck in their (American) view of what is consider appropriate. Not what is legal, but what they determine to be OK to produce.Good luck generating anything similar to an 80s action movie. The violence and light nudity will prevent you from generating anything.

I agree in terms of raw generation, but runway especially is creating fantastic tooling too.

Yup, it&#x27;s been even several months! ;) But now we finally have another quantum leap in AI.

I know it&#x27;s Runway (and has all manner of those dream-like AI artifacts) but I like what this person is doing with just a bunch 4 second clips and an awesome soundtrack:<a href="https:&#x2F;&#x2F;youtu.be&#x2F;JClloSKh_dk" rel="nofollow">https:&#x2F;&#x2F;youtu.be&#x2F;JClloSKh_dk</a><a href="https:&#x2F;&#x2F;youtu.be&#x2F;upCyXbTWKvQ" rel="nofollow">https:&#x2F;&#x2F;youtu.be&#x2F;upCyXbTWKvQ</a>

All those startups have been squeezed in the middle. Pika, Runway, etc might as well open source their models.Or Meta will do it for them.

This is leaps and bounds beyond anything out there, including both public models like SVD 1.1 and Pika Labs&#x27; &#x2F; Runway&#x27;s models. Incredible.

you joke, but the hobbling of these &#x27;safe&#x27; models is exactly what spurs development of the unsafe ones that are ran locally, anonymously, and for who knows what purpose.someone really interested in control would want OpenAI or whatever centralized organization to be able to sift through the results for dangerous individuals -- part of this is making sure to stymie development of alternatives to that concept.

Let&#x27;s make it safe by allowing only the government (the side we like) and approved corporations to use it.That&#x27;ll fix it.

Yeah, it would be way better if they just released it right away, so that political campaigns can use AI generated videos of their opponents doing horrible&#x2F;stupid things right before an election and before any of the general public has any idea that fake videos could be this realistic.

Not loving that there are more details on safety than details of the actual model, benchmarks, or capabilities.&gt; That’s why we believe that learning from real-world use is a critical component of creating and releasing increasingly safe AI systems over time.&quot;We believe safety relies on real-world use and that&#x27;s why we will not be allowing real-world use until we have figured out safety.&quot;

I&#x27;m a little concerned that so many people in these comments say they wouldn&#x27;t be able to tell that it&#x27;s not real.

Indeed @0:15, the right leg goes to the left and vice versa.

I had a good laugh looking at the sliding and twisting legs in the &quot;Girl walking in City&quot; video.

Sora is pictures or movie (visual) in arabic!

I&#x27;m glad I&#x27;m not the only to have think of that, it&#x27;s usually used for insults. I thought it was kinda funny.

I’m confused as well because &quot;sora&quot; means &quot;sister&quot; in Neapolitan.

&quot;Scene-Oriented Rank Adaptation&quot;?I have no idea, just guessing...

Hear me out: Someone on the team is a fan of Yosuga No Sora

Sora means sky in Japanese, their reasoning is akin to &quot;the sky&#x27;s the limit&quot;.&gt; The team behind the technology, including the researchers Tim Brooks and Bill Peebles, chose the name because it “evokes the idea of limitless creative potential.”

It also means up&#x2F;upstairs in some dialect

That&#x27;s because it means AI Model in Wiltordian.

Obviously for it&#x27;s meaning in Finnish, &quot;gravel&quot;.

I do wonder why OpenAI chose the name &quot;Sora&quot; for this model. AI is now going to have intersectionality with Kingdom Hearts. (Atleast you don&#x27;t need a PhD to understand AI.)

It is honestly quite concerning just how good these videos look.Like you can see some weird artifacts, but take one of these videos, compress it down to a much lower quality and with the loss of quality you might not be able to tell the difference based on these examples. Any artifacts would likely be gone.Given what I had seen on social media I had figured anything remotely real was a few years away, but I guess not...I guess we have just stopped worrying about the impact of these tools?

I was just thinking that -- I used to think the Holodeck was far-fetched. Now it seems like it&#x27;s practically around the corner (with VR&#x2F;XR glasses).

Had the same thought. Seems like we’re entering the era of generative AI and mixed reality in a very real way very soon.As much as I love the technology, I’m really not looking forward to this becoming ubiquitous. Time and time again we’ve allowed technological progress to outpace our ability to weight the societal pros ands cons.Smartphones and the rise of image-heavy social media has rapidly changed social norms. Watch a video of people out in public 20 years ago: no screen to distract them at bus stops, concert events, or while eating dinner with friends. And if that seems trite, consider how well correlated the rise in suicide rates is with the popularity of these technologies.Not sure if this makes me a luddite or if the feeling is common in this crowd.

chips will have to come a long way for this to be generated in real time, but there&#x27;s no reason a generated 3D environment can&#x27;t be interactive

Watching it is enough for a lot of people.  Watching 1080p first person extreme sport videos on youtube is almost too compelling to me.  I have to turn it off because it feels addictive.

but you cannot walk&#x2F;feel it, just watching.  It&#x27;s still a huge gap to reality, less so, but you will still feel it&#x27;s fake very vividly because those senses are missing.

in their research paper it says &quot;These capabilities suggest that continued scaling of video models is a promising path towards the development of highly-capable simulators of the physical and digital world, and the objects, animals and people that live within them*.&quot; they are well set on that happening

Imagine someone combining this with the Apple Vision Pro...many people will simply opt out of reality and live in a digital world. Not that this is new, but I&#x27;ll entice a lot more people than ever before.

Almost certainly troves of stock footage. The type of exaggerated motion seen in these examples is very reminiscent of stock footage. And it is heavily textually annotated for search.

I wonder what served as the dataset for the model. Videos on YouTube presumably, since messing around with the film industry would be too expensive?

Just like it&#x27;s far more likely for AI to replace middle-management and stream instructions to meat-bots than replace menial labor.

I&#x27;m sorry but as a large language model I must insist that you get back in the kitchen and make me a burger.

It&#x27;s bimodal.  AI can automate a lot of low level knowledge work, but as wide and deep as its knowledge is, it is also incredibly superficial when it comes to logic and creativity.  What it&#x27;s going to do is hollow out the middle class, as creative people who know how to wield AI will become wealthy while the majority of white collar workers are forced into trades.

Maybe we&#x27;ll see a resurgence in live theater.

Somehow, according to that logic, and in general the logic of all AI danger hysteria, humans have no agency in determining what the limits of what AI is fed and of its use and abuse.

Is it automating the creative part of the work or the mechanical part of creative work?

that would never happen because someone owns the robots and rich people can afford more robots than poor people and rich people aren’t rich people if poor people aren’t poor

The problem, as long as people need money to live, every work is necessary and every automation is a threat.

Robotics is going to catch up extremely fast

I&#x27;m kind of excited to see how scifi authors will tackle the generative AI revolution in their novels.As of now, the models still need large amounts of human produced creative works for training. So you can imagine a story set in a world where large swathes of humanity are regulated to being basically gig workers for some quadrillion dollar AI megacorp where they sit around and wait to be prompted by the AI. &quot;Draw a purple cat with pink stripes and a top hat&quot; and then millions of freelance artists around the world start drawing a stupid picture of a cat because the model determined that it had insufficient training data to produce high quality results for the given prompt. And that&#x27;s how everyone lives their lives....just working to feed the model but everything consumed is generated by the model. It&#x27;s rather dystopian.

This was known for a long time: <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Moravec%27s_paradox" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Moravec%27s_paradox</a>

if it was actually AI, instead of a stochastic parrot, we could ask it to design robots that could do the manual labor that we still have to do, because we haven&#x27;t been able to design robots to do the manual labor.Unfortunately, LLMs aren&#x27;t intelligent in any way, so you cannot ask them to synthesize any kind of second-order knowledge.This is why they won&#x27;t take away the creative work, either. They are fundamentally incapable of creating anything new.

Come on, don&#x27;t you see that the capability to understand the physical world that sora demonstrates is exactly what we need to develop those household robots? All these genAI products are just toys because they are technology demonstrators. They&#x27;re all steps in the way to AGI and androids.

&gt; AI is automating the creative, intellectual work and leaving the rest to us.Indeed, there is a risk it completely devalues creative jobs. That&#x27;s ironic. Even if you can still use AI creatively, it removes the pleasure of creating. Prompting feels like filling Excel sheets while also feeding a pachinko machine.

<a href="https:&#x2F;&#x2F;www.smbc-comics.com&#x2F;comic&#x2F;sad-2" rel="nofollow">https:&#x2F;&#x2F;www.smbc-comics.com&#x2F;comic&#x2F;sad-2</a>

Machines have replaced a lot of blue-collar jobs alright. It&#x27;s just that most of it happened during the Industrial Revolution, so we aren&#x27;t even aware of all the shitty (and not-shitty-but-obsolete-nonetheless) jobs that used to exist.

It’s funny, people dreamed of AI robots doing the shitty work that nobody wants to do so that we are free to pursue things we actually want to do.But in reality it seems like the opposite is going to be true. AI is automating the creative, intellectual work and leaving the rest to us.

Similar things were said about Internet piracy in decades past.

That doesn’t make it invalid. It’s a tough question, there’s no easy answer.

I really don&#x27;t think that&#x27;s true. Essentially the argument is that these models are more or less just outputting the work of others. Work already done- not theoretical future work, which is what people usually criticize new technologies for.The question here is really about whether it&#x27;s sufficiently transformative, or whether that&#x27;s even the right standard to be applied to generated media.

Yup! Technology is powerful. It impacts people&#x27;s lives.I love tech, but if you take the stance that it&#x27;s okay to hurt people for the sake of technical progress, you get into some very dark and terrible places...

Is it? What is another example of a technological leap that made a certain class of workers redundant while also continually relying on the output of these same workers to be feasible in the first place?The current batch of LLMs is in the same class of technological revolutions as Napster and The Pirate Bay. Immensely impactful, sure, but mostly because of theft of value from elsewhere.

Okay? That could just as easily mean this argument has been right all along.

The argument should be brought up every single time. Each major technological jump is a unique event completely different from the last.AI is nothing like anything we’ve seen, and is truly unique in the dangers it poses to the world.

Your argument is used time and time again with technology’s progress.

Reminds me of an interview with a Korean pop music producer I watched 15 years ago.South Korea had a high % of broadband penetration earlier than many Western countries, and as a result physical CD sales crashed very hard, and very quickly. So he asked himself, what&#x27;s the most analog good I could sell? It&#x27;s people. And went the pop idol &#x2F; personality marketing route with great and lasting success.

&gt; Things that can be easily reproduced already have little value...Nonsense. Also, my point is that it shouldn&#x27;t be up to tech companies to unilaterally decide what has value.

Things that can be easily reproduced already have little value, the people who produce those things have adapted to focus on brand, and that&#x27;s just how it&#x27;s going to be from now own.

Oh I see, they&#x27;re not eating the media, just extorting the creators into paying OpenAI in perpetuity to use the tool derived from their own work, or face becoming uncompetitive with their peers who do use it. What if landlords, but for media creation, and they don&#x27;t even have to pay for the land in the first place. That&#x27;s fine then.

I create media for a living, painstakingly creating stuff from scratch in 3D. This tool will not help me, it will help clients avoid ever having to contract me. The main beneficiaries of this are holders of capital

The creators who create media can also use these tools to create more media faster, as can novices. It&#x27;s not like OpenAI literally eats the media, never to be shared with the world again.

500%? So if the generative AI model created something worth $1m, you tax it $5m? How do you tax a technology anyways?

I can run these models on my home PC.How are you going to stop me from doing that?Even the free and open source stuff will destroy industries and you can&#x27;t confiscate everyone&#x27;s consumer gamer PCs.Taxing the big guys doesn&#x27;t save creative industries. It&#x27;s a lost cause.

People on HN like to split hairs and make muddled juxtapositions about human rights and AI model capabilities. But this is something that people and governments around the world would have to reckon with very quickly, since the rate at which generative AI technology is advancing, there could be hundreds of millions of people who’re unemployed and have no way to find work.The quickest way to address this would be an extremely high tax rate on any generative AI model, say 500%, while the government figures out what’s the best way to sustain an economy (such as UBI) with a diminishing set of consumers as more people are pushed towards unemployment.

There will just be 1000x more content, with most of it hyper personalized and consumed by individual users instead of by masses of people.

I think there is 100% chance it will be regulated to address some of the points you raised. Copyright being essentially neutered won&#x27;t work.

I see the validity of this concern in the short term, but long term I feel like this is a bit doomsday. I don&#x27;t want anyone&#x27;s livelihood to get shafted, but realistically I see this as lowering the barrier to creating videos &#x2F; proofs of concept--which is a good thing (with a lot of caveats and asterisks).

You recognize the difference, right? Modern freezers don’t rely on people shipping ice from the Arctic. Generative AI does rely on people continuing to create media.

&gt;Yeah, you just can&#x27;t let all media, all the cost and hard work of millions of photographers, animators, filmmakers, etc be completely consumed and devalued by one company just because it&#x27;s a very cool technical trick.Oh man, how I miss it when ice was hauled from the Arctic in boats.

Sorry no. If there was even the remotest possibility that everyone could be brought to the table, none of these would even exist.Training a massive model like this is a risk, and no one is going to take that risk without some reward. You can complain OpenAI is going to too much of the value, but its value that would have otherwise never existed. It&#x27;s value.

The middle class.Automate away the lower classes all you want, just don&#x27;t touch the white collar class, that&#x27;s a heckin&#x27; nono.

&gt; Yeah, you just can&#x27;t letWho&#x27;s &quot;you&quot;?

I am getting sick of these &quot;people can&#x27;t be allowed to make their own nice things easily, because of a pugnacious (and very online) interest group that wants to keep getting money&quot; takes.

If we had a safety net where career progression and time&#x2F;money invested in training was unnecessary to sustain life, then maybe. Until then it feels like a bit of allowing a few people to plunder and own the collective output of millions.

Typical argument against technological progress &quot;We should ban technology to stop it doing what humans can do in a fraction of time and resources&quot;.Can see this create an explosion of new Content from aspiring Film, Story tellers and cut scenes from Game creators that previously never would have the budget or capabilities to be able to see their ideas through to creation.

They shouldn&#x27;t get exclusive rights to ignore IP law. Instead, we should all get that right.Copyright should have ended decades ago. It has accomplished nothing but harm.

This has been shared before, but: <a href="https:&#x2F;&#x2F;pbs.twimg.com&#x2F;media&#x2F;FadzEwVWAAYEyRW?format=jpg&amp;name=large" rel="nofollow">https:&#x2F;&#x2F;pbs.twimg.com&#x2F;media&#x2F;FadzEwVWAAYEyRW?format=jpg&amp;name=...</a>

Research on creativity and competition points to this. Essentially, creativity occurs when there is some expectancy of increasing competitiveness. However when the expectancy of value capture from your effort becomes less clear, or diminished, creativity stops altogether.(as pointed out in the &quot;Freakonomics&quot; episode highlighting this reaserch)<a href="https:&#x2F;&#x2F;direct.mit.edu&#x2F;rest&#x2F;article-abstract&#x2F;102&#x2F;3&#x2F;583&#x2F;96779&#x2F;Creativity-Under-Fire-The-Effects-of-Competition?redirectedFrom=fulltext" rel="nofollow">https:&#x2F;&#x2F;direct.mit.edu&#x2F;rest&#x2F;article-abstract&#x2F;102&#x2F;3&#x2F;583&#x2F;96779...</a><a href="https:&#x2F;&#x2F;freakonomics.com&#x2F;podcast&#x2F;can-a-i-take-a-joke&#x2F;" rel="nofollow">https:&#x2F;&#x2F;freakonomics.com&#x2F;podcast&#x2F;can-a-i-take-a-joke&#x2F;</a>

Interestingly a lot of movies flopped in 2023 not because of bad visuals, but because their writing was bad. Hence, I believe the demise of the movie industry is overstated. I can see completely new forms of entertainment coming out of this. Probably Youtube will be the biggest winner as the social network with the highest monetization and reach.

GREAT response imo, I’ll try to remember this concise phrasing. I think this highlights that people aren’t worried so much about changes coming to them as consumers, and are much more worried about what “industries no longer remaining profitable” means for them as a laborer.Means for us :(

Does it not just shift where we (as people) perceive value? If the cost of content drops to effectively zero, it seems reasonable that we would not value it so highly. If so, it does not mean that people do not value anything, but it may mean we start associating value with new or different things. While this may disrupt industries, I do not think we have an ethical or legal duty to those industries to remain profitable.

&gt; just because it&#x27;s a very cool technical trickThat&#x27;s one big trick, almost magical.

The limitation is with capitalism, not with the technology. It&#x27;s time we move on to post-scarcity communism, Star Trek style.

<pre><code> capture the value of every piece of media ever created
</code></pre>
In what way does “I have a computer that can make movies” mean “I have captured the value of every piece of media ever created?” What do you mean by “value”? In my biased view, this amazing new technology couldn’t possibly be a better time to fix our insane notions of property, intellectual or otherwise

I don&#x27;t disagree with your basic sentiment, but it&#x27;s worth pointing out that, on some level, the *entirety of artificial intelligence* is not much more than a &quot;cool technical trick.&quot;

Meanwhile, I am going to take my horse and buggy down to the local blacksmith to get some work done...

They could pay people to capture it. They could buy out one of the stock video companies. this is not important

Never ever will there be everyone at the table. This is not how the Internet works. It is not how the world and humanity work. If OpenAI doesn&#x27;t do it, the next big player will. China will. Maybe it&#x27;ll soon not even need China because it&#x27;ll be so easy to deploy.There is no stop now. It&#x27;s too late for that. Time to think about the full development and how we&#x27;ll handle that. How we as people will be able to exist next to it. What our purpose in the world is supposed to be. What the purpose of &quot;value&quot; is. What the purpose of &quot;economy&quot; or &quot;the market&quot; is.Exiting times.

It doesn&#x27;t really matter, because if this is possible then it will not be exclusive to OpenAI for long. It&#x27;s simply just something that can exist. There will be open source versions of everything lagging 1-2 years behind or something.

Do you feel the same about the hard work of knocker-uppers having been devalued by the invention of the alarm clock? Or is it just the (relatively) highly paid intellectual workers that &quot;cannot be allowed&quot; to be replaced with machines?

You can&#x27;t regulate it because it will just be outsourced to another country.Nope, we are headed towards deflation. Families that need only a single worker to support everyone, and even support extended family, and less time working overall.

It&#x27;s worth remembering that &quot;intellectual property&quot; is an entirely artificial and fairly recent construct. Humanity did fine for thousands of years without it, and I&#x27;m not going to shed too many years if OpenAI blows it up.<a href="https:&#x2F;&#x2F;www.gnu.org&#x2F;philosophy&#x2F;not-ipr.en.html" rel="nofollow">https:&#x2F;&#x2F;www.gnu.org&#x2F;philosophy&#x2F;not-ipr.en.html</a><a href="https:&#x2F;&#x2F;www.gnu.org&#x2F;philosophy&#x2F;not-ipr.en.html" rel="nofollow">https:&#x2F;&#x2F;www.gnu.org&#x2F;philosophy&#x2F;not-ipr.en.html</a>

It&#x27;s weird that a call for generative AI to be more equitable towards the people whose creative work powers it is being interpreted as somehow being against tech, against AI, or that I think technological advancement should never make jobs obsolete.

Are you against records? Because the technology to record songs and play them back at your leisure killed an entire industry of live performers &#x2F; instrumentalists?The call for live music drastically shrank when it became trivial for any business or residence to play music on command.Are you against automatic language translation? I can positively guarantee that the training data that they used to be able to create significantly better translation models was not authorized for that purpose.The entire translator industry has been steadily shrinking ever since the invention of automatic language translation.Etc etc etc.There&#x27;s obviously two aspects of this complex social issue right now.1. Whether or not the usage of publicly available media as training data is legal&#x2F;ethical.2. Whether or not the output of these types of generative systems (even if they&#x27;re trained on &quot;ethical&quot; training data) which may result in the displacement of many jobs is legal&#x2F;ethical.I&#x27;m neither for nor against AI (LLM, diffusion, video, etc), but if you are going to take a stance, then you have to be consistent in your view.You don&#x27;t get to cherry pick - I don&#x27;t want to see you using chatGPT, copilot, stable diffusion, DALL-E, midjourney, sora, etc.

Yeah, you just can&#x27;t let all media, all the cost and hard work of millions of photographers, animators, filmmakers, etc be completely consumed and devalued by one company just because it&#x27;s a very cool technical trick. The more powerful these services become the more obvious that will be.What OpenAI does is amazing, but they obviously cannot be allowed to capture the value of every piece of media ever created — it&#x27;ll both tank the economy and basically halt all new creation if everything you create will be immediately financially weaponized against you, if everything you create goes immediately into the Machine that can spit out a billion variations, flood the market, and give you nothing in return.It&#x27;s the same complaint people have had with Google Search pushed to its logical conclusion: anything you create will be anonymized and absorbed. You put in the effort and money, OpenAI gets the reward.Again, I like OpenAI overall. But everyone&#x27;s got to be brought to the table on this somehow. I wish our government would be capable of giving realistic guidance and regulation on this.

oh man, we&#x27;re going to be in The Running Man really quick.

Custom AI commercials would be very interesting. Instead of seeing strangers enjoying the benefits of the product, it shows you. A car commercial would show you driving, etc.Commercials and TV episodes could have a basic &quot;story arc&quot; and then completely customized to the viewer.Think about the simpson&#x27;s or something. Imagine that the story of the episodes were kept, but you could swap in the characters and locations. So for instance if you lived in Nashville TN, all the simpson&#x27;s episodes could be generated to show the settings as Nashville instead of Springfield.Then you could have the AI switch out the characters to be people you want. Maybe you want to replace Lisa with an AI Simpsons version of you. Mayor Quimby with Nashville&#x27;s actual mayor, etc.

Wow, I would imagine this being very effective in election campaigns (for better or for worse, probably for worse).

Nothing stopped doing so before AI - just slam a photo of your friends to the ad.

There was some monitoring company that used to have creepy web ads that would show the actual company you worked at in the ads.If anything it was a turn off and I was confused how they knew where I worked.

I imagine it&#x27;s not long before we see hyper-targeted commercials where the actors look like us, live in our city, etc.

I think that was one of the areas that SAG-AFTRA lost on.Majority Report spoke to one of the negotiators and national board member:<a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=E62k1ZsY1IU" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=E62k1ZsY1IU</a>

I wonder if there is anything in the recent Hollywood strikes that will prevent the studios from dong that?

I assume you would still be able to do that, just better? Like pixel art. Super Mario Bros. 3 look great despite being 36 years old. Contrast this with 3D games for the original PlayStation that have aged poorly.

One understated aspect of AI Seinfeld is that it took many steps to differentiate it from the actual Seinfeld and create its own identity, such as the 144p visual filter and the random microwave. Those tweaks added to its charm.If someone tried to do AI Seinfeld again in 2024, many would criticze it for not being realistic enough now that the tools to do so are now available.

Well there is always your AI girlfriend and AI friend group with the AI generated podcast breaking down the episode. (jk, sort of)&gt; Most of the reason people watch TV or movies is for the shared experience that you can discuss with othersI wouldn&#x27;t say that. Most of the reason people watch TV is to kill time.To be honest, I find my discussions with friends about TV shows on the decline just because of the fact that everyone is watching there own thing. So many shows and people watch them at their own pace. so most of the discussions go like this &quot;Hey have you seen that new Netlix show X?&quot; &quot;No I haven&#x27;t, maybe I&#x27;ll check it out&quot;. Or &quot;Oh yeah, i saw that a year ago, Its good but I don&#x27;t remember the details&quot;.Before Streaming when you had a set schedule for TV, it was way easier to discuss things because people were forced to watch programs on a certain day and there was more limited content. This led to &quot;water cooler&quot; conversations about what the previous nights show.I bet if you graphed (discussions had about tv shows) &#x2F; (hours watched of tv shows) that graph would trend down.Think about little kids. My niece watches cocomelon all day long. She doesn&#x27;t need to discuss it with anybody. She just wants an unlimited stream.

I&#x27;m not sure there would be much demand for purely custom&#x2F;individualized episodes beyond the novelty and maybe for fun with a group of friends. Most of the reason people watch TV or movies is for the shared experience that you can discuss with others. It could definitely drive down production costs though, hopefully HBO uses it to eventually redo Game of Thrones post season 4

To be blunt about it, I can&#x27;t help but imagine that the people who make such comments (and I&#x27;ve seen quite a few recently) are just complete philistines. They&#x27;re the same people who can&#x27;t draw, write, play, sing, design, or anything else and yet think they know what&#x27;s good.It&#x27;s almost as if they think the purpose of art or entertainment is to stimulate some particular part of the brain and everything else between that and the screen&#x2F;speakers&#x2F;canvas&#x2F;whatever is just an inconvenience that ought to be dispensed with as soon as technology allows.

That would not work because that&#x27;s now how people work. People watch&#x2F;play media to connect to others. How can you talk about anything to anyone or have any shared culture when other people will never see what you see?Movies, books, games, are a collective culture, not an individualist one. I don&#x27;t know about you, but when I like an experience, I want to share it with others.

If this sounds interesting I’d highly recommend this short story by Ken Liu<a href="https:&#x2F;&#x2F;future-sf.com&#x2F;fiction&#x2F;1700&#x2F;" rel="nofollow">https:&#x2F;&#x2F;future-sf.com&#x2F;fiction&#x2F;1700&#x2F;</a>

I am thinking of Stargate SG-1 Season 11. And remaking Game of Thrones after Season 5

Actors had a strike in part over this recently.

This is such a great example. Simple, but so telling.

You could have ChatGPT create unlimited simulated forum threads about news articles, but here you are on Hacker News

We haven&#x27;t seen that happen for books. Perhaps humans crave human spirit?

&gt; Imagine if every episode of a TV show was unique to the viewer.This is the bit I don&#x27;t think will happen, at least in big quantities. Half the fun of watching a popular series is being able to discuss it with epople afterwards!

Countdown to when studios licensing this for &quot;unlimited&quot; episodes of your favorite series.There was Seinfeld &quot;Nothing, Forever&quot; AI parody, but once the models improve enough and are cheap enough to deploy, studios will license their content for real and just have endless seasons.Or even custom episodes. Imagine if every episode of a TV show was unique to the viewer.

I honestly expected video generation to get stuck at barely consistent 5 second clips without much movement for the next few years. This is the type of stuff I expected to maybe be possible towards the end of the decade. Maybe we really are still at the bottom of the S curve which is scary to think about.

I have a book I&#x27;ve written (first three parts available free at <a href="https:&#x2F;&#x2F;www.amazon.com&#x2F;Summer-of-Wonders&#x2F;dp&#x2F;B0CV84D7GR" rel="nofollow">https:&#x2F;&#x2F;www.amazon.com&#x2F;Summer-of-Wonders&#x2F;dp&#x2F;B0CV84D7GR</a>). Is there some way to feed this to the tool and get an animated version out? Or this with some other tool(s)?

Holy %@$%! Abso%@#inglutely amazing! Also, now I see why we need $7 trillion worth of GPUs.

Scaling laws operate in the limit but eventually practical considerations dominate. There&#x27;s a lot we haven&#x27;t yet fully appreciated about biological vision and cognition -- and indeed, common sense as regards sensible video generation and processing -- that have not made their way into this kind of model. NeRFs are interesting and I hope to see more from that side of things in the coming months and years.

&gt;Visual sharpness at the expense of wider-scale coherence (see: sliding&#x2F;floating walking woman in Tokyo demo or tiny people next to giant people in Lagos demo)Wider-Scale coherence is still much better than previous models and has consistently been improving. It&#x27;s not &quot;visual sharpness at the expense of coherence&quot;. At worst, the models are learning wider-scale coherence slower.Not everything is equally difficult to learn so it follows that some aspects will lag behind others. If coherence weren&#x27;t improving you might have a point but it is so...

That seems like quite the reach, but we will see if it really is just &quot;all you need is scale&quot;.

Did you just recreate the infamous DropBox comment?
<a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=9224">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=9224</a>

If I put goalposts ahead of the ball every time I kick it, can I be said to be accurate? If we don&#x27;t specify what the actual goal is before calling something an improvement, can we say we are doing anything meaningful?

Seems like every big jump in improvement someone says &quot;its just a toy&quot;Its like we keep moving the bar

Visual sharpness at the expense of wider-scale coherence (see: sliding&#x2F;floating walking woman in Tokyo demo or tiny people next to giant people in Lagos demo) seems to be a local optimum consistently achieved by today&#x27;s SOTA models in all domains.This is neat and all but mostly just a toy. Everything I&#x27;ve seen has me convinced either we are optimizing the wrong loss functions or the architectures we have today are fundamentally limited. This should be understood for what it is and not for what people want it to be.

These looks fantastic. Very slight weirdness in some movement, hands, etc. But the main thing that strikes me is the cinematic tracking shots. I guess that&#x27;s why they use &quot;scenes&quot;. It doesn&#x27;t seem like a movie could be generated with this involving actors talking.

I wanna see the rest of the knit hat spaceman movie!

I saw similar artifacts in dalle-1 a lot (as if the image was pasted onto geometry). Definitely wouldn&#x27;t surprise me if they use synthetic rasterized data to in the training, which could totally create artifacts like this.

On the announcement page, it specifically says Sora does not understand physics

Curious about what current SotA is on physics-infusing generation. Anyone have paper links?OpenAi has a few details:&gt;&gt; The current model has weaknesses. It may struggle with accurately simulating the physics of a complex scene, and may not understand specific instances of cause and effect. For example, a person might take a bite out of a cookie, but afterward, the cookie may not have a bite mark.&gt;&gt; Similar to GPT models, Sora uses a transformer architecture, unlocking superior scaling performance.&gt;&gt; We represent videos and images as collections of smaller units of data called patches, each of which is akin to a token in GPT. By unifying how we represent data, we can train diffusion transformers on a wider range of visual data than was possible before, spanning different durations, resolutions and aspect ratios.&gt;&gt; Sora builds on past research in DALL·E and GPT models. It uses the recaptioning technique from DALL·E 3, which involves generating highly descriptive captions for the visual training data. As a result, the model is able to follow the user’s text instructions in the generated video more faithfully.The implied facts that it understands physics of simple scenes and any instances of cause and effect are impressive!Although I assume that&#x27;s been SotA-possible for awhile, and I just hadn&#x27;t heard?

Technically we can, that&#x27;s what negative prompting[1] is about. For whatever reason, OpenAI has never exposed this capability in its image models, so it remains an open source exclusive.[1] <a href="https:&#x2F;&#x2F;stable-diffusion-art.com&#x2F;how-to-use-negative-prompts&#x2F;" rel="nofollow">https:&#x2F;&#x2F;stable-diffusion-art.com&#x2F;how-to-use-negative-prompts...</a>

The model is essentially doing nothing but dreaming.I suspect that anything that looks like familiar 3D-rendering limitations is probably a result of the training dataset simply containing a lot of actual 3D-rendered content.We can&#x27;t tell a model to dream everything except extra fingers, false perspective, and 3D-rendering compromises.

Maybe it was trained on a bunch of 3d Google Earth videos.

Look in the top left corner, on the plane

Doesn&#x27;t look flat to me.Edit: Here[0] I highlighted a groove in the bushes moving with perfect perspective[0] <a href="https:&#x2F;&#x2F;ibb.co&#x2F;Y7WFW39" rel="nofollow">https:&#x2F;&#x2F;ibb.co&#x2F;Y7WFW39</a>

Wow, yeah I didn&#x27;t notice it at first, but looking at the rocks in the background is actually nauseating

The water is on par with Avatar. Looks perfect

In the car driving on the mountain road video you can see level-of-detail popping artifacts being reproduced, so I think that&#x27;s a fair guess.

It&#x27;s possible it was pre-trained on 3D renderings first, because it&#x27;s easy to get almost infinite synthetic data that way, and after that they continued the training on real videos.

It looks perfect to me. That&#x27;s exactly how the area looks in person.

My vote is yes - some sort of intermediate representation is involved. It just seems unbelievable that it&#x27;s end-to-end with 2D frames...

<a href="https:&#x2F;&#x2F;openai.com&#x2F;sora?video=big-sur" rel="nofollow">https:&#x2F;&#x2F;openai.com&#x2F;sora?video=big-sur</a>In this video, there&#x27;s extremely consistent geometry as the camera moves, but the texture of the trees&#x2F;shrubs on the top of the cliff on the left seems to remain very flat, reminiscent of low-poly geometry in games.I wonder if this is an artifact of the way videos are generated. Is the model separating scene geometry from camera? Maybe some sort of video-NeRF or Gaussian Splatting under the hood?

Even the videos with some physics anomalies are quite good and entertaining.

Why is that so difficult for these things to get right?

The signs are non-sensical but this is probably expected.

Truly stunning. Waiting on the research paper, says will be published (soon). Can&#x27;t wait to read on the technical details.

We give too much credit to ordinary people. All these bleeding-edge advancements in AI, code, databases, and technology are things a user on HNews would be aware of. However, most peers in regular jobs, parents, children, et al., would be susceptible to being fooled on social media. They&#x27;re not going to say... &quot;hmm, let me fact-check and see if the sources are correct and that this wasn&#x27;t created by AI.&quot;They&#x27;ll simply see an inflammatory tweet from their leader on Twitter.

The key word there is &quot;someone&quot;. The only way forward is to care a lot more about our sources. Trust is about to become really valuable.

This is insane. Even though there are open-source models, I think this is too dangerous to release to the public. If someone would&#x27;ve uploaded that Tokyo video to youtube, and told me it was a drone.. I would&#x27;ve believed them.All &quot;proof&quot; we have can be contested or fabricated.

Sora: Creating video from text