not now, but surely will be one day. the video will end and marques will come, sit on a small chair in middle of a dimly lit room, and say this video was completely AI generated.
@ it's always been possible, but it required the necessary knowledge and a little time. here, it's within everyone's reach, not to mention the fact that AI will continue to progress imagine such tools within the reach of governments
I imagine there will be much more work available for forensic analysis techs to determine the real from the AI, for an appropriate fee. (Bitcoin accepted.)
There is already a lot of stuff that's AI. Dashcam/securitycam videos (easier to fake), stock footage for "interesting" factoids (often blatantly false facts), terrible life hacks, "most satisfying" vids -- There's a lot of AI generated video for clickbait stuff already going around.
Facebook is number one for this , I had to remake a new account and do " show less " of all AI stuff to finally get a usable acc , I don't use the video tab , but that one is filled with Cat videos or AI ( weird ones ) and ... let's not talk about Ai-Jesus , he's like number 1 face of facebook right now with posts like " my kid made this with hand " or " you are good the way you are " ( add girls with 3 legs or two heads from AI )
Soon? It's already infested with so much AI generated crap and the worst part of it is that millions get fooled into believing they are real. And don't even get me started on images.
@@macTijnquality and repetition of patterns. The mountain clip has a relatively same pattern for tree placement. Almost unnatural. It’s scary… but now we have to deal with it. lol.
No matter how impressed i am by the tech behind these innovations, i always come away feeling empty and sad, only able to think of the bad that will come out of these tools.
@@MrPootzen Have you ever drove a fast car ? Fast car with good brakes, you just floor it on a straight then slam those high perf brakes to stop in no time. No one harmed
I too come away feeling sad.... Bcuz I can't make my close friends do stupid things in the prompts as a silly in-meme bcuz the bastards who programmed it won't let me use their likeness. Or can I? Hmmm
Its hilarious how careful and strict OpenAI is with EVERYONES copyright and intellectual property EXCEPT for the millions of people whose work, art and private data this was feasting on to train the model - somehow their rights dont matter one bit, but uploading a Mickeymouse? You crook!
Open ai directly stated “there would be no way for open ai to be profitable without using copyrighted material for free”. Literally admitting to theft.
Humans are trained on other people’s work all the time. Your work is never original and was inspired by your training. AI is no different, and, frankly, the vast majority don’t care how things are made as long as they are good, cheap, and fast.
@@tarettime9392 It's not even close to profitible anyway. They have plundered IP from millions of creators and spent billions for an, as yet, unviable business model.
@@o_o825 Humans are in fact not trained on other people's work like AI. Also, you might be surprised to know that if you pirated or stole a book, you'd still be legally liable even if you used it to learn. Besides, people care. It's why we have all these rules and regulations around labor and sketchy import/export. I think you're just assuming that the general population shares your same disinterest for how the sausage is made. Oh, and software developers are trained on other people's work all the time too. So can we use all of OpenAI's engineering work as if it was public domain, up to and including 'acquiring' their internal material? Or is that only if you're doing it to art?
The internet is already almost dead imo, at least half of TH-cam is ai generated content with ai voices and ChatGPT text. It just gets more fake from now on. Its ridiculous and it’s no more fun for me honestly.
@@tobik2627its very easy to just not open those links and use real internet lol. If youre chronically online and watching fake videos, thats a you issue. This will only make the internet grow. Once the entire population is on it thats when we’ll see major shifts. You need to remember some countries restrict internet to their own websites, while others have a majority population that dont use it daily, its a luxury for many.
Tbh I would only want to use this tool to make the cool imaginary stories that I had as a kid/teenager come to life. I am CS student and mostly in the field of SWE/Devops but always had a passion for art/storytelling/3d design etc... but never went that direction because I just suck at other thing that is not logic. This tool just makes it possible to bring those "dreams" to life without having skill in drawing/designing/modeling in a couple of seconds
the thing that worries me is, why are we automating jobs that everyone wants? Drawing, Photography, Music, Video making, etc... and leaving unchanged jobs that no one wants?
AI is not capable of anything magnificent. By its nature It can only produce the most average representation of something. It is unvieldy, it ignores half of the prompt, it cannot produce something new. For example. try to convince Midjourney to draw a zeppelin falling nose down through the clouds. It just cant. And that giraffe in the video supposed to be centaurus, not just giraffe. Most people lack the ability to produce something "magnificent" also. So no big deal really.
The last part of this video gives me chills down my spine. For example, one Chinese AI video generate tool somehow created scenes from Korea's Joseon dynasty (old korean dynasty) surprisingly well. It turned out that they had used old Korean historical dramas without permission or copyright. This came to light because some results from that tool actually showed Korean broadcasting companies' logos at the top of the videos it created.
@@marwin4348 Fair use is not something you can claim protection for, it's a defense you need to use in court. Whether it's protected is not decided, and in some jurisdiction like the EU not respecting opt-outs is illegal (so funnily enough, those 'not for AI use' disclaimers would be legally binding).
If it makes the tool more accurate about what it generates, then I’m all in favor of training it on copyrighted content. Also, by defending copyright, you’re not protecting the artists, you’re protecting the companies making tons on money off royalties.
Remember the good old days when you looked at an artwork and thought to yourself, "Wow, that artist must have put in a lot of dedication and work hard into this art. Mad respect" Now it's just, "This must be made with AI. Lame AI. Too many of them. Pass."
The problem is not that physics is hard. The problem happens because the algorithms are learning a pixel distribution. It essentially learns to predict the most likely frame coming next, but this doesn’t mean the algorithm actually understands what it is generating. It does not understand logic like us, that’s why it is not coherent in movements. Anything that requires too much rules and logic the AI will likely fail at. What I mean is, this happens to physics but could happen to anything that follows “logic”. The algorithm is not inherently bad at physics, it is bad at logic. It does not even care about physics to start :)
Well said. Sora has a very limited internal model of how the world works which is really just a reflection of its weights. I can see a future type of technology that combines the power of Gen-AI like Sora with a state of the art physical models (like the type that Tesla use for training) and I believe that fusion will be a real leap.
Isnt that the exact same argument for LLMs? It doesnt know what comes next, just predicting the next word. It doesn't understand like we do. But if you use the latest gpt4 or o1 models, they are very good. This will likely end up similar to that. The first LLMs sucked and didn't really make much sense. Until they started to...
This is the most level zero argument. You could say that about your brain. “It’s just optimizing for the release of certain neurotransmitters! It’s just a bunch of neurons firing!”
@@marcuskissinger3842 I’m not saying SORA is bad nor that fitting curves is bad. You missed my point, I’m saying the current methodologies will not be able to reason. Maybe the function they are currently fitting, such as probability density functions (PDF) in this case, are not rich enough to incorporate all mechanics required to reason and understand the real world. We humans function in a different way despite maybe being just another “probabilistic machine” in theoretical terms haha. However, I think we are able to find patterns and reason with more flexibility. I believe AI will reach that point, but not trough LLMs and DDPMs. I have a PhD in Computer Science and believe me I know how impressive SORA results are! I’ve studied DDPMs, they are complex and amazing (although quite computationally expensive haha), but it’s still far from perfect. It will not lead to “logic/reasoning” capabilities :)
"Dead internet theory" was already more of a reality than a theory, but man is it about to get SO MUCH worse and the only thing we're gonna hear is "who could've saw THIS coming?" followed by thousands of "I told you so"
Was about to comment this. I mean it wasn’t even that long ago that I remember seeing the old SORA previews…and now it’s open to public. Within the next year or two we’ll probably get way more indescribable fakes. While the dead internet theory is very real and has been growing over time, this I think is the actual start to a rapid acceleration towards a truly dead internet. On the other hand though I’m curious to see if it can get better. The GPT architecture is still very limited. Apple had a study that basically said it cannot reason. Meaning I don’t think it could have an accurate physics engine built in.
Literally no one except for clickbait headlines will say " who could have seen this coming". Most major news outlets have talked about it. Every single tech outlet has talked about it. Major tech people have commented on it and been interviewed about it. So even the clickbait headlines will still say it happened just as predicted in the actual text.
@SoteriaNavi laughed at that. I was more thinking in the way of media, since they started talking about it this year. To be fair, the dead internet theory becomes fact with he integration of AI into everything, so it makes sense that it became a talking point this year. I wonder how and what the internet will be like in 5 years. What even are facts anymore at that point?
"Cartoons don't have to have realistic movements or physics" - I just went to see a re-release of "My Neighbor Totoro," and there is a behind-the-scenes interview with Miyazaki where he hand-draws and re-draws a short scene of someone cutting bread to ensure the bread loaf moves and deforms in the right way. That's the kind of dedication he shows to his animation. Also worth noting that he called AI animation "an insult to life itself."
Did we watch the same Miyazaki films? There's a whole lot of unrealistic movement/physics that happens in those movies. I'm sure all of it was carefully directed to move "the right way", but there's no need for it to be realistic if it enhances the experience.
@@FunkyJeff22unrealistic doesn’t mean the movement just happens randomly. unrealistic movement in animation happens for a reason to enhance the experience, not like the ai in these videos
There have been over a dozen AI video tools available for almost a year now that blow Sora out of the water, and they also let you generate whatever you want without the "safety" bullshit.
Lew has been using that exact fake plant to review smartphone cameras for ages. It's a classic test he uses along with his beard hair test for the selfie camera.
@@115spt2 Since June 2023 he’s had 3 violations and got away from all of em with a slap on the wrist, not counting this most recent instance where the cops weren’t involved. He’s so lucky he doesn’t even know it. Although he acknowledged and apologized for his behavior in his latest tweet on the matter which is a step in the right direction (and hopefully not just said out of necessity to avoid controversy). Hopefully he keeps his promise and doesn’t put his life or the lives of others at risk again.
If you’re a free user, you get no generations, I think the whole point is to just give people a chance to recognize that it’s AI without restricting the person generating as well
The top comments being so neutral or even positive about this are making me sick. This shit is not a good thing at all, like literally anybody can make any accusation on you and now can just generate fake but convincing-looking "evidence" against you and you'd have no way of disproving it. Like genuinely, please think of the moral and political issues this'll bring. Security footage, news, photographic evidence, ect. All of that will mean nothing now. We'll no longer be able to trust ANYTHING. How that scares none of you is beyond me.
By far the best comment on the subject yet. 👍 Add to this all of the free speech, debate and disagreement censoring laws, bills and regulations and we literally stand no chance against it.
This will be another tool to further distance the people from each other. I thought we are already living a life where people are senseless but oh boy, this will be even bigger step.
If you knew anything about how these systems work, you would know you can in fact detect AI generated videos. Does this mean people won't use it for propaganda purposes? No, of course they will, but the problem is not the technology then, isn't it?
As sick as this all is. I physically struggle to watch these ai videos because they send me down an uncanny valley. Because I know its not real but its trying its best to convince me it's real. I feel uncomfortable. I can't look for too long. Strange, I know lol
Why do people care about this so much??? People speed all the damn time and I know if you had a Lamborghini you wouldn’t be going the speed limit all the time💀💀
13:54 this moment is one of those that make you appreciate the intentionality on every detail of your videos. I was puzzled by that odd choice of background. Suspiciously asymmetrical and now it makes perfect sense. You guys are artists
So if you are caught doing something illegal on a CCTV camera, you can start performing magic with slight of hand in the vid and claim that the video was SORA generated.
Can you talk about the issues of creators content being stolen, with AI tools like these? Worth addressing for a proper review of the situation at hand, and SORA's context. If you do a basic review of technicalities, that's a great way to encourage the strategy of stealing content to be unpunished.
@@ThunderDawg0 " Midjourney developers caught discussing laundering, and creating a database of Artists (who have been dehumanized to styles) to train Midjourney off of. This has been submitted into evidence for the lawsuit. Prompt engineers, your “skills” are not yours " - @JonLamArt, also attaching actual proof of the laundering and what used to be the spreadsheet of over 16k artists before it got restricted
Imagine you are Born without being able to touch or see, but you are forced to draw a Person, Remember that you took the USA flag and mashed It into a ball.
People are going to lose all trust in images or video being real. Which is a strange reality to be entering into. Real images are going to be devalued, and lost in a sea of fakes.
There is a new standard to combat this called C2PA (Coalition for Content Provenance and Authenticity). Although it is early in adoption it has some big names behind it (Adobe, Microsoft). How it works at a technical level goes over my head but I guess we’ll have to see if this can solve that issue, at least enough for the general public to feel comfortable.
My issue with these services is that they’re gonna make money from these models, models trained on scraped data that wasn’t legally protected because our legal system hasn’t yet caught up to the predatory behavior of said models. AI is obviously here to stay, and as a free tool to mess around with I think it’s fine, but when companies start making money off of these mass-internet-scrapers I can’t help but feel uneasy
You can use the same argument against fair-use YT commentation/reaction channels. The point is that they are changing it and "adding their own spin" as you would say, which if those YT channels are okay, then I would have to say this is ok too.
And they're not even showing the original content, it's new content based on some ideas of original content. It's no different than a human looking at something then making an entirely new art piece based on his previous knowledge@@chrisjacobsen1659
One simple method I've discovered to somewhat reliably tell if an image (or video) is AI-generated is to look at the overall distribution of light and dark spots. I remember seeing on Corridor Crew that AI-generated images usually tend to have an equal distribution of light and dark spots/regions throughout. Meaning, that if you see an image or (a video) that, on close inspection seems to have an an equal number of dark and light regions, it's probably AI-generated. I mean, look at that Mt. Fuji video. You have a light blue sky, white clouds and white snow in approximately half the screen and the other half is filled with a dark brown mountain and a dark green forest. So, approximately half the image/video is light-coloured while the other half is dark-coloured. From what I know, it's simply because the AI generates things like images in such a way that it starts from a relatively neutral-coloured box and then increases and decreases the brightness of various regions to create an image. But it still maintains the overall average of the contrast. The guys at Corridor said something (that I don't really remember) about some tool/setting in Photoshop that can average out the colour of the image and easily help you figure it out.
And even that will probably be ironed out in future revisions of the software. It's so crazy that folks are actually wilfully pushing towards a future where nothing is real..
@ylondes9927 the main thing that I hope for is that all of this forces people to look at reality with much more scrutiny, instead of blindly accepting whatever is presented to them as fact
I hate that he plays dumb with the ethics of it. Of course his videos have been stolen, no you can't opt out because once its been used it's in the weights forever, and of course it uses more energy then text generation. Why play dumb tech bro "who just doesn't know" with these big companies instead of actually talking about the problems?
You have no basis for these claims. With a platform as big as MKBHD's you have to be careful while slinging around allegations. For a video focused on just reviewing the capabilities of the software he did a great job highlighting the ethics issue
The proof is simply inherent if you have any knowledge of video generation. It's also why these companies are getting more GPUs and trying to get nuclear reactors. However, I do agree that this video was more review focused.
@@infinitesky1495 the replies here are so lame, especially when Marques previously was AGAINST having ai used for his stuff and was exposing the issues around it. What I mean is why criticize it back then but now suddenly do reviews on the tech and kinda almost praise it? At least being annoyingly neutral on it. Like you said, why play dumb when you 200% know the issues with all of this?
The fact that OpenAl is so cautious and stringent about everyone's copyright and intellectual property with the exception of the millions of people whose labor, artwork, and private information.
I can see theres going to be a whole media law department that deals with such frequent issues. Things are going to take a very weird turn from now on, any visual media can be manipulated.
NOOO TAKE DOWN EVERYTHING WE CAN'T GIVE IT MORE STUFF TO COPY FROM Edit: i don't think its copying anymore, they did it, they made a model able to be trained to make videos (basically the visual-making part of your imagination i think) but good news, they did not make a model for understanding or thoughts or move-logic(how objects move for realistiker videos) yet... yet.
what do you mean by that? it's not stealing intellectual property for a human to learn from the things they see, so why would it be for an AI to do the same?
Wait its gonna be available to everyone soon? This extremely powerful tool. Once it does, there's no going back. Its literally gonna be like opening Pandora's box.
There have already been a number of other video generation tools out since the first Sora announcement, some of those now on par in quality, Runway, Luma, Kling, …
This tool is going to allow for an inundation of propaganda, unless we ban these, totalitarianism is our future. These videos arent good, but the crappy AI generated videos are throwing enough confusion into the mix to make the truth harder and harder to know. Thats perfect for a totalitarian.
As a 3d animator/shader, the giveaway regarding humans is the fact Ai bypasses creating subsurface scattering. The skin looks dead. I’m sure that is around the corner, but these are things we have to do to make characters look more warm and realistic.
It looked like there was some subsurface scattering on the skin in that monkey shot. It's kind of hard to tell with Sora from these included videos. (Also a lot of times, if people are wearing heavy makeup, then there's no actual subsurface scattering on their faces. Looks weird. (Oh god and don't even get me started on those horrific 'beauty filters' people are suing now 🤮🤮🤮🤮🤮))
Walnut milk and other nut milks are the food tech advancements that’s mind-boggling to me! Now with crispr, we can turn nuts into mammals to create milk. 🥛
@@lmbuewe need two things, AI being able to actually think by itself, with quantum computing it could become "complete" together with robots and machines being agile enough to execute what the AI says, when that happens, the world will evolve 100 years (if there was no AI) in only 10
Double the speed limit from where I’m from would get your car impounded, fined big time and loss of license on top going to court in front of a judge with possible jail time. Yeah multiple police body cam footage of you has been uploaded and I’m disgusted.
I've noticed in the freelance market some clients specifically ask for "NO AI". People, even non-creatives, Also - I think high quality unique REAL stock footage will have it's value! Let's hope.
Someone else described AI in a great way. I'm paraphrasing but it went something like: "It's like stepping on a full bus, but none of the people have any faces." Fake, uncanny, nightmarish.
That already happened in 2020, when everyone covered their faces with stinky moist gag rags, because the tv told them they could "catch" a "deadly disease".
As a video artist, I've seen some objectively cool effects leveraging AI. Transitions are where it really shines. Using it's inherent lack of object permanence to create match cuts, or intricate motion matching (like turning a dancer into dancing fire). These can be time consuming effects to create by hand
Agreed. One of my favorite instagram accounts creates videos that are nightmarish hence it isn't supposed to look realistic. The artist deliberately leans into the creepiness so that's why it works so well
@@tbricksable Trends change as fast as tech. Truth is, AI will probably make these kind of effects seem dated as they oversaturate the market. Human creativity is what I'm paid for.
I have yet to see a law or any sort of regulation in the world that properly tackles the use of AI-generated content. Morality is nothing for people (billionaire moguls) who will do anything to maximize profit and cut costs. It's sad honestly, governments have yet to do something about this,
14:10 Pretty messed up,alot of tech reviewers have the exact same plant. Pretty bad they're using content as of it doesn't matter at all.They are literally stealing stuff.
Not really, I have already found use for it at my graphic design job, I can easily curate ideas quickly without going searching for hours on end looking for something. It will also make stock images and videos accessible to everyone now. Creating content has literally never been easier.
@mahjoubadam1742 As an aspiring artist myself I understand that easier and more convenient doesn't always mean better. The merit of these professions lies solely in the fact that the work is performed by people with their own styles and specialties. An author would never ask a large language model for inspiration on a part of their story. Maybe I'm just a purist, and maybe it's different for something like graphic design, but in my eyes AI has no place in any of the arts. The moment AI becomes capable of replicating what humans can do, and the moment you let it, is the moment it becomes advantageous to let it do everything. And that is something that might not be up to the artists themselves. Like I said, the merit lies in one place, and it's important to preserve it. Obviously I can't tell you what to do but that's just how I see it
@@coutherino I wouldn't call you a purist as much a regressionist, LLMs are really nothing but advanced search algorithms, just predicting the next most common word for the current scenario. Maybe it's easier on my soul to use AI because I studied it at my university.
@@mahjoubadam1742 I mean yea, exactly. I think the process you just described is antithetical to the nature of art in the same way sora is. If two things can create works that are virtually indistinguishable from each other (and AI is certainly close) they are only differentiated by the process they take to arrive there. AI is incapable of creating art because it cannot use the process through which art is created and defined. Not that that's what we were talking about, it's just my reasoning behind why AI shouldn't be used to create art. Or rather, can't
@@coutherinobut if you agree that AI generated content is just a mixture of everything it's seen, then using it as a reference make sense, because in references you're not looking for originality (you don't care, since the reference isn't your end product), you're just looking for a particular thing that resonates with you. So, instead of searching for these images/videos, you describe them, and AI gives you back what google would, but more in line with your prompt and use-case. It's not art, I agree, but as a reference it's no worse than using standard references (like 90% of the artists do). In short, if using AI as a reference is bad, then standard usage of references is also bad, in my opinion.
It will become available at some point most likely. But I actually like that EU regulation is cautious with allowing these kinds of tools straight of the bat
If you call standing still fast. Tell me the difference between this and Sora's own demos from a year ago which if anything, looked better. Emperor's new clothes.
@bearimo2867 There is a difference between shown publicly available ai models and ones in the labs . When sora was first shown it was far ahead of any ai video the ones in current labs are far ahead of this version of sora infact the next version was already leaked in a presentation it's really good and almost Hollywood level
@@tambal40 it's nowhere near Hollywood level. 5 second clips that take hours to curate to get right that still have obviously artificial issues like legs blending together, is not Hollywood level. And this is absolutely about as good as it will get. Like chat, there is limited training data and they have already exhausted it for chat, I'd not be surprised if they had done the same for video. AI is one big shell game and too many people are falling for it. The costs nowhere near make the value proposition. AI will crash hard, and fast, soon. The running costs are astronomical, and there is zero profitability to be gained. Open AI is losing more money daily regularly than any company in history. And that's with the advantages it has through partnership with Microsoft. Without them it is screwed. Tick, tock. It's inevitable.
It's a bummer that we already know we're gonna need a market for pre-GenAI data the same way we have a whole marketplace for "low-background" steel. The enshitification of the internet marches boldly forward.
It's not true that the models work by generating images and just stacking them together. You can literally just look up the architecture and see that's not how they work. They take in whole videos to learn how videos work, including the time domain and the relative movements between frames. This is why they learn to simulate physics. They did try just stacking generated images together 4 or 5 generations ago, but they quickly realized that didn't work. There was no consistency and you just had artifacts popping in and out between every frame.
@@jhanninnen yes but the images aren't generated in isolation, there's a time domain that's consistent across all frames, which is what makes a video different from a series of images.
The prompt of this video is: "Create a video of MKBHD reviewing Sora"
Got ‘em
This cannot be true! There is audio in this video.
@ that’s fair though, your not quite right. The audio can always be ADR in afterwards
Woosh @@sascha148
not now, but surely will be one day.
the video will end and marques will come, sit on a small chair in middle of a dimly lit room, and say this video was completely AI generated.
"This is a powerful tool about to be in the hands of millions" - oh boy here we go
Im gonna use your pfp for the main characters bf
Now is a great time to reduce your usage of the internet for informational purposes.
It's almost as if we shouldn't have built the Torment Nexus
With great power comes great responsibility
Don’t worry, you have to pay to try it, and the smallest package is 50/month. It will not be in the hands of millions!
the ability to make CCTV footage is about to change the world in ways i can’t even imagine
It's always been possible. What world changing events are you worried about specifically?
@ it's always been possible, but it required the necessary knowledge and a little time. here, it's within everyone's reach, not to mention the fact that AI will continue to progress
imagine such tools within the reach of governments
I imagine there will be much more work available for forensic analysis techs to determine the real from the AI, for an appropriate fee. (Bitcoin accepted.)
@@michaelberg7201 Just because it's been possible doesn't mean it's always been easy, mass producible, or convincing. That's the difference.
its 2025 time for CCTV footage to be higher Res
Bro needs to drop a video on radar detectors, going 59 in a 25 is crazy
I came here just for this comment 😭
It’s not that crazy. A lot of people do it…
Are you gonna fabricate some story to try and guilt trip us?
yap
little bro can't even drive a car 😂
@@TheUltimateGeminiHasSurvived
Yeah, such a 'guilt trip'....
You tube shorts boutta be a bunch of AI spam
About to be?
people are gonna curse the whole AI thing real soon if they haven't already
Fake Movie trailers are already full of them
There is already a lot of stuff that's AI.
Dashcam/securitycam videos (easier to fake), stock footage for "interesting" factoids (often blatantly false facts), terrible life hacks, "most satisfying" vids -- There's a lot of AI generated video for clickbait stuff already going around.
there is a lot of this on tiktok already
Oh man. Facebook will soon be covered with AI videos I have to explain to my parents are not real...can't wait :P
Facebook is number one for this , I had to remake a new account and do " show less " of all AI stuff to finally get a usable acc , I don't use the video tab , but that one is filled with Cat videos or AI ( weird ones )
and ... let's not talk about Ai-Jesus , he's like number 1 face of facebook right now with posts like " my kid made this with hand " or " you are good the way you are " ( add girls with 3 legs or two heads from AI )
But will you be able to tell yourself?
It already is, but it's going to get worse (probably)
Tho I don't see how this is improvement compared to Runway for these scammers.
my mother is stuck to facebook i have no idea how to help her
Soon? It's already infested with so much AI generated crap and the worst part of it is that millions get fooled into believing they are real. And don't even get me started on images.
Man I’m screwed, I didn’t get any of those first clips right 😭
I got just one of them. And this stuff will just keep improving in quality. I'm certainly a bit more nervous about the future now...
we cooked
@@macTijnquality and repetition of patterns. The mountain clip has a relatively same pattern for tree placement. Almost unnatural.
It’s scary… but now we have to deal with it. lol.
bro im actually same, 😭😭
I only got the second one
Can you stop going 1440p in a 720p zone?
True
No matter how impressed i am by the tech behind these innovations, i always come away feeling empty and sad, only able to think of the bad that will come out of these tools.
bet Marques was mad that he didn't think to blame the video of him nearly killing those kids on AI
@@MrPootzen "nearly killing those kids" bro you are a delusional hater 💀touch grass
@@MrPootzen Have you ever drove a fast car ? Fast car with good brakes, you just floor it on a straight then slam those high perf brakes to stop in no time. No one harmed
I too come away feeling sad.... Bcuz I can't make my close friends do stupid things in the prompts as a silly in-meme bcuz the bastards who programmed it won't let me use their likeness. Or can I? Hmmm
@@Broke.97 actually delusional
i thought the entire video was ai generated lmao
@@johnsalamii you're making assumptions way too fast. faster than MKB-
It is
How? It just came out
it is.
wait for few year😂
*The first video in a while where the comments are actually about the video*
Refreshing
Thank god. That shit was never that funny
And it's less about people wanting him to be held accountable and more about just getting the likes for the joke.
@@Jwats96 That's because you don't have a sense of humor
I think it was more about it being illegal then funny
Gotta change that name to MKB-Caught In-HD
ur corny bro
Its hilarious how careful and strict OpenAI is with EVERYONES copyright and intellectual property EXCEPT for the millions of people whose work, art and private data this was feasting on to train the model - somehow their rights dont matter one bit, but uploading a Mickeymouse? You crook!
great point
Open ai directly stated “there would be no way for open ai to be profitable without using copyrighted material for free”. Literally admitting to theft.
Humans are trained on other people’s work all the time. Your work is never original and was inspired by your training.
AI is no different, and, frankly, the vast majority don’t care how things are made as long as they are good, cheap, and fast.
@@tarettime9392 It's not even close to profitible anyway. They have plundered IP from millions of creators and spent billions for an, as yet, unviable business model.
@@o_o825 Humans are in fact not trained on other people's work like AI. Also, you might be surprised to know that if you pirated or stole a book, you'd still be legally liable even if you used it to learn.
Besides, people care. It's why we have all these rules and regulations around labor and sketchy import/export. I think you're just assuming that the general population shares your same disinterest for how the sausage is made.
Oh, and software developers are trained on other people's work all the time too. So can we use all of OpenAI's engineering work as if it was public domain, up to and including 'acquiring' their internal material? Or is that only if you're doing it to art?
RIP internet, 1994-2025, Thanks for the memories. Ai, generate video memories from my photo album to sad rock music.
Yup
The internet is already almost dead imo, at least half of TH-cam is ai generated content with ai voices and ChatGPT text. It just gets more fake from now on. Its ridiculous and it’s no more fun for me honestly.
@@tobik2627its very easy to just not open those links and use real internet lol. If youre chronically online and watching fake videos, thats a you issue.
This will only make the internet grow. Once the entire population is on it thats when we’ll see major shifts. You need to remember some countries restrict internet to their own websites, while others have a majority population that dont use it daily, its a luxury for many.
Tbh I would only want to use this tool to make the cool imaginary stories that I had as a kid/teenager come to life.
I am CS student and mostly in the field of SWE/Devops but always had a passion for art/storytelling/3d design etc... but never went that direction because I just suck at other thing that is not logic.
This tool just makes it possible to bring those "dreams" to life without having skill in drawing/designing/modeling in a couple of seconds
@@tobik2627Respectfully. I don't agree with you. Your comment reminds me of the homeless men walking around with signs saying: "The End Is Nigh"
The fake plant is crazy 13:44
The Fake Plant Situation Is Actually Insane
@@SuperFilmregisseur its not a coincidence. its trained off the internet.. dont think its not using things exactly as the appear from the internet.
Unbox therapy uses that plant also
@@ZupE891 The Fake Plant Situation JUST Got Worse -》This. Is. BAD
That plant was copied straight from “unbox therapy”
the thing that worries me is, why are we automating jobs that everyone wants?
Drawing, Photography, Music, Video making, etc...
and leaving unchanged jobs that no one wants?
It costs too much to make the robots be able to do all the manual labor. They can't let the masses have too much freedom or else we might wake up.
I know actually!!
@HampsterGirl you know the answer to my question?
well I'll be damned, Now I want to know!
Because it's way cheaper
Not everyone wants to draw, photograph, make music, as well as make videos
“We marveled at our own magnificence as we gave birth to A.I”
- Morpheus
An so man became the architect of it's own demise.
I know man haven't any of these idiots making AI seen any of these films
The Matrix (first one). is a prophecy.
AI is not capable of anything magnificent. By its nature It can only produce the most average representation of something. It is unvieldy, it ignores half of the prompt, it cannot produce something new. For example. try to convince Midjourney to draw a zeppelin falling nose down through the clouds. It just cant. And that giraffe in the video supposed to be centaurus, not just giraffe. Most people lack the ability to produce something "magnificent" also. So no big deal really.
@@kudandanga9400 More like The Second Renaissance from Animatrix
The last part of this video gives me chills down my spine. For example, one Chinese AI video generate tool somehow created scenes from Korea's Joseon dynasty (old korean dynasty) surprisingly well. It turned out that they had used old Korean historical dramas without permission or copyright. This came to light because some results from that tool actually showed Korean broadcasting companies' logos at the top of the videos it created.
It's protected as fair use, they do not need permission and they are not breaking any laws.
Unfortunately it seems the time period where humans could reliably watch recordings of what happened was less than 2 centuries.
@@marwin4348 using images that contain a private company's logo with that logo then showing up on generated images is not fair use
@@marwin4348 Fair use is not something you can claim protection for, it's a defense you need to use in court. Whether it's protected is not decided, and in some jurisdiction like the EU not respecting opt-outs is illegal (so funnily enough, those 'not for AI use' disclaimers would be legally binding).
If it makes the tool more accurate about what it generates, then I’m all in favor of training it on copyrighted content.
Also, by defending copyright, you’re not protecting the artists, you’re protecting the companies making tons on money off royalties.
15:46 I thought he was gonna say "This entire video you just wathced was AI generated"
😂😂😂
lol 😂
Same tbh.
Same I thought that too
AI is still limited to 10 and 15 second short videos.
Can you generate a video of you not driving over the speed limit or is that impossible even for AI?
💀💀💀
That meme is still massive
@waterscod259 you know what else is massive...
Joke would have been funnier if you had said under the speed limit
@@TheValetudinarianwoooahhhh there
I'm at the point where id be more likely to purchase a product with a "certified Ai free" badge
Absolutely. This is so disgusting to replace the art with AI
haha yes! Something that’s specifically AI free.
Already there myself
Remember the good old days when you looked at an artwork and thought to yourself, "Wow, that artist must have put in a lot of dedication and work hard into this art. Mad respect" Now it's just, "This must be made with AI. Lame AI. Too many of them. Pass."
Oh no, not the "Detroit: Become Human" references already surfacing... We're so cooked 🔥
The problem is not that physics is hard. The problem happens because the algorithms are learning a pixel distribution. It essentially learns to predict the most likely frame coming next, but this doesn’t mean the algorithm actually understands what it is generating. It does not understand logic like us, that’s why it is not coherent in movements. Anything that requires too much rules and logic the AI will likely fail at.
What I mean is, this happens to physics but could happen to anything that follows “logic”. The algorithm is not inherently bad at physics, it is bad at logic. It does not even care about physics to start :)
Beautifully put.
Well said. Sora has a very limited internal model of how the world works which is really just a reflection of its weights. I can see a future type of technology that combines the power of Gen-AI like Sora with a state of the art physical models (like the type that Tesla use for training) and I believe that fusion will be a real leap.
Isnt that the exact same argument for LLMs? It doesnt know what comes next, just predicting the next word. It doesn't understand like we do.
But if you use the latest gpt4 or o1 models, they are very good. This will likely end up similar to that.
The first LLMs sucked and didn't really make much sense. Until they started to...
This is the most level zero argument. You could say that about your brain. “It’s just optimizing for the release of certain neurotransmitters! It’s just a bunch of neurons firing!”
@@marcuskissinger3842 I’m not saying SORA is bad nor that fitting curves is bad. You missed my point, I’m saying the current methodologies will not be able to reason. Maybe the function they are currently fitting, such as probability density functions (PDF) in this case, are not rich enough to incorporate all mechanics required to reason and understand the real world. We humans function in a different way despite maybe being just another “probabilistic machine” in theoretical terms haha. However, I think we are able to find patterns and reason with more flexibility. I believe AI will reach that point, but not trough LLMs and DDPMs. I have a PhD in Computer Science and believe me I know how impressive SORA results are! I’ve studied DDPMs, they are complex and amazing (although quite computationally expensive haha), but it’s still far from perfect. It will not lead to “logic/reasoning” capabilities :)
"Dead internet theory" was already more of a reality than a theory, but man is it about to get SO MUCH worse and the only thing we're gonna hear is "who could've saw THIS coming?" followed by thousands of "I told you so"
Was about to comment this. I mean it wasn’t even that long ago that I remember seeing the old SORA previews…and now it’s open to public. Within the next year or two we’ll probably get way more indescribable fakes. While the dead internet theory is very real and has been growing over time, this I think is the actual start to a rapid acceleration towards a truly dead internet.
On the other hand though I’m curious to see if it can get better. The GPT architecture is still very limited. Apple had a study that basically said it cannot reason. Meaning I don’t think it could have an accurate physics engine built in.
it's gonna be even worse when the AI starts getting trained on all the AI slop already posted on the internet. It's like a centipede of shit content
Literally no one except for clickbait headlines will say " who could have seen this coming". Most major news outlets have talked about it. Every single tech outlet has talked about it. Major tech people have commented on it and been interviewed about it. So even the clickbait headlines will still say it happened just as predicted in the actual text.
@fredrikfjeld1575 you have so much more faith for the intelligence of the average person and I am deeply jealous of that
@SoteriaNavi laughed at that. I was more thinking in the way of media, since they started talking about it this year. To be fair, the dead internet theory becomes fact with he integration of AI into everything, so it makes sense that it became a talking point this year. I wonder how and what the internet will be like in 5 years. What even are facts anymore at that point?
So you're saying we should be nicer to technology...
Remember always Please and Thank you!
We're saying he should drive under the speed limit
all youtubers including you are stressed...
Good luck Zac, I wish you all the best
"Cartoons don't have to have realistic movements or physics" - I just went to see a re-release of "My Neighbor Totoro," and there is a behind-the-scenes interview with Miyazaki where he hand-draws and re-draws a short scene of someone cutting bread to ensure the bread loaf moves and deforms in the right way. That's the kind of dedication he shows to his animation. Also worth noting that he called AI animation "an insult to life itself."
Yes, that comment was more based on his lack of understanding of animation principles than anything else.
Did we watch the same Miyazaki films? There's a whole lot of unrealistic movement/physics that happens in those movies. I'm sure all of it was carefully directed to move "the right way", but there's no need for it to be realistic if it enhances the experience.
@@FunkyJeff22unrealistic doesn’t mean the movement just happens randomly. unrealistic movement in animation happens for a reason to enhance the experience, not like the ai in these videos
For every one animation like that there are 20 more that arent even remotely realistic, whether due to budget, laziness, or even stylistic reasons
South park enters the chat
There have been over a dozen AI video tools available for almost a year now that blow Sora out of the water, and they also let you generate whatever you want without the "safety" bullshit.
Got any names?
@@QuestGoblin Kling, Luma, Hunyuan and plenty of others.
that's scary. every AI video should have a stamp that it is an AI video. government should implement something for this...
@@motorepoloh ...why?
@@MatthewSabia So people don't get framed lol. Just one example.
If I remember correctly, that plant can be found in a lot of Unbox Therapy’s earlier videos from a few years ago.
Wow 😮
Doesn't make it any better. It's still scrapping YT for content
its still in his videos lol
Ikea prob 😂
Lew has been using that exact fake plant to review smartphone cameras for ages. It's a classic test he uses along with his beard hair test for the selfie camera.
Who is here after the seeing the body cam footage?
Me lol
Yuh
wheres the speed comments. why yall stopping
mkbhd is a literal monster
Wonder if he got rid of his illegal tints by now...
@@epzo It's the internet. It's old news and nobody cares anymore.
13:33 best part of the video
Sora casually hinting it saw much more videos from you than you saw from it 🌚🌚🌚
The fact that this didn’t end with will smith eating spaghetti is a failure of unimaginable proportions. It’s the perfect baseline for improvement.
Probably not allowed under their terms lol
In the terms any popular celebrity is not allowed to be generated
He actually referenced this in the video at 10:28 - "rewritten to remove reference to public figures"
He mentioned that we can’t do that in 14:56
@@wembathen he could describe a black man eats spaghetti
I hate AI man especially after this comes into public, because YT is gonna be AI trash very very soon
We move into bilibili
@@dimensionsazuma32 bilibili already like some tiktok trash.
ai is already polluting TH-cam
It's mostly trash already.
Only bad artists fear AI replacing them. Make better art.
95mph in a residential neighborhood is crazy you should never drive again
The fact he's done it multiple times and faced barely any repercussions is insane. Dude should have at least a few months of Jail by now.
@@115spt2 a few years would be justice.
you need to go out of your way to drive 95mph in such a place..
thats attempted murder or sth
@@robinspanier7017 lmfao
These kids crying about someone going over the speed limit is hilarious.
@@115spt2 Since June 2023 he’s had 3 violations and got away from all of em with a slap on the wrist, not counting this most recent instance where the cops weren’t involved. He’s so lucky he doesn’t even know it. Although he acknowledged and apologized for his behavior in his latest tweet on the matter which is a step in the right direction (and hopefully not just said out of necessity to avoid controversy). Hopefully he keeps his promise and doesn’t put his life or the lives of others at risk again.
step away from the kangaroo 😂
oi mate, gotta make some lane math
😂😂zesty ahh boy
That video was hilarious
6:05 the "mouse" turns into a kitten
I’m weirded out😟😟
😂
It looks more like just a miniature cat, not a kitten. Or a cat/mouse hybrid.
I wish my mice did that
@@dannywoods17 yep you're right, it has a cat head from the start
What's the point of adding a watermark to the video if you then give the user a "Download without watermark" button?
I presume "download without watermark" will be a premium feature you pay extra for?
Subscription
If you’re a free user, you get no generations, I think the whole point is to just give people a chance to recognize that it’s AI without restricting the person generating as well
@@malsawmzela609 No, because there are no free users for Sora. Either you have a subscription or you can’t use it at all.
Pro plan can download without watermark
"I am above the law"
-MKBHD, probably
The top comments being so neutral or even positive about this are making me sick. This shit is not a good thing at all, like literally anybody can make any accusation on you and now can just generate fake but convincing-looking "evidence" against you and you'd have no way of disproving it. Like genuinely, please think of the moral and political issues this'll bring. Security footage, news, photographic evidence, ect. All of that will mean nothing now. We'll no longer be able to trust ANYTHING. How that scares none of you is beyond me.
By far the best comment on the subject yet. 👍
Add to this all of the free speech, debate and disagreement censoring laws, bills and regulations and we literally stand no chance against it.
🤷♂️ eh if that is the future…so be it
This will be another tool to further distance the people from each other. I thought we are already living a life where people are senseless but oh boy, this will be even bigger step.
also the human removal in modern art culture
If you knew anything about how these systems work, you would know you can in fact detect AI generated videos. Does this mean people won't use it for propaganda purposes? No, of course they will, but the problem is not the technology then, isn't it?
As sick as this all is. I physically struggle to watch these ai videos because they send me down an uncanny valley. Because I know its not real but its trying its best to convince me it's real. I feel uncomfortable. I can't look for too long. Strange, I know lol
I feel that completely, the same with all A.I vids. It’s some weird itch under my skin, I react so weirdly watching things move in ways they shouldn’t
Wow, this is so interesting, I never expected someone feeling like that, but I kinda understand
Yeah, that's exactly what "uncanny valley" means by definition.
Its like experiencing hallucinations, you feel uncomfortable because it feels like losing control.
I feel the same way about modern music. I can't listen too long.
That fake plant was a creepy “My work is being used without consent or payment” moment.
Multiple police bodycam videos showing MKBHD being pulled over for traffic violations have surfaced
Why do people care about this so much??? People speed all the damn time and I know if you had a Lamborghini you wouldn’t be going the speed limit all the time💀💀
@@user_47313 tell that to the people whose lives are endangered
@@user_47313What if he hit someone?
You want a cookie or?
@@user_47313 "People speed all the time"? Does that make it okay?
13:54 this moment is one of those that make you appreciate the intentionality on every detail of your videos. I was puzzled by that odd choice of background. Suspiciously asymmetrical and now it makes perfect sense. You guys are artists
I dont get it
Yeah, I imagine it'll become a common practice to watermark videos in this way to protect copyright.
Say more
I don't get it either
Explain please
So if you are caught doing something illegal on a CCTV camera, you can start performing magic with slight of hand in the vid and claim that the video was SORA generated.
Just always walk funny, and it will look AI generated.
For example driving too fast in a school zone you mean?
Or if you’re caught three times…
"CCTV video of a driver nearby a school"
I raced here as fast as I could
Most underrated comment
I zoomed through a school zone to get here fast enough
Not as fast as MKBD going through a school zone.
Can you talk about the issues of creators content being stolen, with AI tools like these? Worth addressing for a proper review of the situation at hand, and SORA's context.
If you do a basic review of technicalities, that's a great way to encourage the strategy of stealing content to be unpunished.
Seriously, it's too late for that. The genie's out of the bottle, and she's not going back in any time soon, or ever.
Stollen? Never has ai done that, its just butt hurt people.
@@ThunderDawg0
" Midjourney developers caught discussing laundering, and creating a database of Artists (who have been dehumanized to styles) to train Midjourney off of. This has been submitted into evidence for the lawsuit. Prompt engineers, your “skills” are not yours "
- @JonLamArt, also attaching actual proof of the laundering and what used to be the spreadsheet of over 16k artists before it got restricted
Imagine you are Born without being able to touch or see, but you are forced to draw a Person, Remember that you took the USA flag and mashed It into a ball.
@@ThunderDawg0exactly.
People are going to lose all trust in images or video being real. Which is a strange reality to be entering into. Real images are going to be devalued, and lost in a sea of fakes.
there could be like a law that requires any image, vid, or media that utilized AI to indicate it as an icon of sorts
Every corrupt politician caught on video or camera doing something wrong can claim it's fake... I feel like we're entering such dangerous territory
I mean, its the same as it is with text today already (e.g. quotes).
There is a new standard to combat this called C2PA (Coalition for Content Provenance and Authenticity). Although it is early in adoption it has some big names behind it (Adobe, Microsoft).
How it works at a technical level goes over my head but I guess we’ll have to see if this can solve that issue, at least enough for the general public to feel comfortable.
Therre is *NO SPOON*
My issue with these services is that they’re gonna make money from these models, models trained on scraped data that wasn’t legally protected because our legal system hasn’t yet caught up to the predatory behavior of said models. AI is obviously here to stay, and as a free tool to mess around with I think it’s fine, but when companies start making money off of these mass-internet-scrapers I can’t help but feel uneasy
Not your storage to begin with 😂 that's a loop hole so they can share it to any buyers
Waaaa waaaa waaaa
You can use the same argument against fair-use YT commentation/reaction channels. The point is that they are changing it and "adding their own spin" as you would say, which if those YT channels are okay, then I would have to say this is ok too.
And they're not even showing the original content, it's new content based on some ideas of original content. It's no different than a human looking at something then making an entirely new art piece based on his previous knowledge@@chrisjacobsen1659
@chrisjacobsen1659 The difference is that yt channels actually credit the source (most of the times) but the ai can never point out any source.
One of the best videos Marques has ever done. He really understands the future, and his talent shines when he reviews things other than gadgets.
MKBHD: Did you catch it?
Me: YA I DID YA YA
MKBHD: No that was real
Me: I knew that
MKBHD: But you thought about it
Me: Goddammit
Lol 😆 thought the same thing! I knew it was real, but damn I thought about it
We about to remake Game of thrones S8
I just watched game of thrones and I swear I see a comment about it every video I click now 😂
I was thinking now we can just take books and create movies hassle free
a 20 second AI generated video would be better than the whole season
enter the most mind-numbingly derivative era of creation in human history
Not on my watch
Era of AI sloptent
13:40 Sora be watching too much unbox therapy haha definitely part of the source material
One simple method I've discovered to somewhat reliably tell if an image (or video) is AI-generated is to look at the overall distribution of light and dark spots. I remember seeing on Corridor Crew that AI-generated images usually tend to have an equal distribution of light and dark spots/regions throughout. Meaning, that if you see an image or (a video) that, on close inspection seems to have an an equal number of dark and light regions, it's probably AI-generated. I mean, look at that Mt. Fuji video. You have a light blue sky, white clouds and white snow in approximately half the screen and the other half is filled with a dark brown mountain and a dark green forest. So, approximately half the image/video is light-coloured while the other half is dark-coloured.
From what I know, it's simply because the AI generates things like images in such a way that it starts from a relatively neutral-coloured box and then increases and decreases the brightness of various regions to create an image. But it still maintains the overall average of the contrast. The guys at Corridor said something (that I don't really remember) about some tool/setting in Photoshop that can average out the colour of the image and easily help you figure it out.
I see what you mean but that doesn't sound reliable in the long run. That said, it'd be interesting to literally analyze the histograms
And even that will probably be ironed out in future revisions of the software. It's so crazy that folks are actually wilfully pushing towards a future where nothing is real..
@ylondes9927 the main thing that I hope for is that all of this forces people to look at reality with much more scrutiny, instead of blindly accepting whatever is presented to them as fact
Good point, AI does seem to reuse the same exposure levels
But I wonder if that could be prompted away.
I hate that he plays dumb with the ethics of it. Of course his videos have been stolen, no you can't opt out because once its been used it's in the weights forever, and of course it uses more energy then text generation. Why play dumb tech bro "who just doesn't know" with these big companies instead of actually talking about the problems?
hes a product reviewer, not an investigative journalist. those are different skillsets
You have no basis for these claims. With a platform as big as MKBHD's you have to be careful while slinging around allegations. For a video focused on just reviewing the capabilities of the software he did a great job highlighting the ethics issue
The proof is simply inherent if you have any knowledge of video generation. It's also why these companies are getting more GPUs and trying to get nuclear reactors. However, I do agree that this video was more review focused.
@infinitesky1495 I've logged in to stupid TH-cam just to +1 this. well said.
@@infinitesky1495 the replies here are so lame, especially when Marques previously was AGAINST having ai used for his stuff and was exposing the issues around it. What I mean is why criticize it back then but now suddenly do reviews on the tech and kinda almost praise it? At least being annoyingly neutral on it. Like you said, why play dumb when you 200% know the issues with all of this?
The fact that OpenAl is so cautious and stringent about everyone's copyright and intellectual property with the exception of the millions of people whose labor, artwork, and private information.
I can see theres going to be a whole media law department that deals with such frequent issues. Things are going to take a very weird turn from now on, any visual media can be manipulated.
NOOO TAKE DOWN EVERYTHING WE CAN'T GIVE IT MORE STUFF TO COPY FROM
Edit: i don't think its copying anymore, they did it, they made a model able to be trained to make videos (basically the visual-making part of your imagination i think) but good news, they did not make a model for understanding or thoughts or move-logic(how objects move for realistiker videos) yet...
yet.
that plant 💀
what do you mean by that? it's not stealing intellectual property for a human to learn from the things they see, so why would it be for an AI to do the same?
@@tommys234 because of the scale of it. also because piracy IS stealing and they definitely did not pay for all that IP
Marques Speedlee at it again
Can't believe nobody is talking about this, the dude is going to end up killing a kid or someone's parents.
@@115spt2 Because nobody actually cares. It's a tech channel.
Wait its gonna be available to everyone soon? This extremely powerful tool. Once it does, there's no going back. Its literally gonna be like opening Pandora's box.
There have already been a number of other video generation tools out since the first Sora announcement, some of those now on par in quality, Runway, Luma, Kling, …
Teen dating and fatherlessness has opened Pandora's box
Going from the video of AI Will Smith eating food to this is absolutely astonishing. I don’t think any of us are ready for the future lol
This tool is going to allow for an inundation of propaganda, unless we ban these, totalitarianism is our future.
These videos arent good, but the crappy AI generated videos are throwing enough confusion into the mix to make the truth harder and harder to know. Thats perfect for a totalitarian.
And that video was made just last year
It took us only 66 years from the discovery of airplane to landing on the moon.
I can't imagine what ai will be capable of doing in 10 years.
@@malsawmzela609nah it took us almost 300y from the first industrial revolution which is less than 1% of how long humans have been existed
dont overgeneralize not everyone is stupid
As a 3d animator/shader, the giveaway regarding humans is the fact Ai bypasses creating subsurface scattering. The skin looks dead. I’m sure that is around the corner, but these are things we have to do to make characters look more warm and realistic.
depends on what you're generating but i'd hazard there are bigger giveaways on humans especially
Everything is around the corner apparently
Yo thanks for the insight
Yes…but you’re an expert…regular ppl won’t be able to tell the difference
It looked like there was some subsurface scattering on the skin in that monkey shot. It's kind of hard to tell with Sora from these included videos. (Also a lot of times, if people are wearing heavy makeup, then there's no actual subsurface scattering on their faces. Looks weird. (Oh god and don't even get me started on those horrific 'beauty filters' people are suing now 🤮🤮🤮🤮🤮))
Can't wait for the AI generated video of MKBHD addressing his reckless driving!
1:07 Look at her feet - how she walks. This is uncomfortlable to watch
Her legs switch 🤢
You are just jealous of her cool walk😂😂
Good catch
@@ran-ns8ry Look closely. No one is jealous. It's AI generated and her steps feel a little bit weird. That's it.
@@ran-ns8ry And go to 7:27
wild time to be alive. Last 20-25yrs of technological advancements has been nuts.
Walnut milk and other nut milks are the food tech advancements that’s mind-boggling to me! Now with crispr, we can turn nuts into mammals to create milk. 🥛
Kinda feels like we’ve already reached peak human, now the machines take over.
90s tech is actually even more impressive than the last 25 years considering what we have in the 80s and
Feels like things were slowing down over the last decade, or like we’re about to break through a plateau.
@@lmbuewe need two things, AI being able to actually think by itself, with quantum computing it could become "complete" together with robots and machines being agile enough to execute what the AI says, when that happens, the world will evolve 100 years (if there was no AI) in only 10
some youtubers 14:16 put a greenie to humanize their desk setup, that grass look.
Can you AI generate yourself driving under the speed limit?
8:55
GET DOWN HE'S GOT A GU-oh it's a camera
😂
He's gonna shoot with it, though.
And have the cameraman never dies, armor
Ai gooning sessions are gonna be wild 💀
How does bro have 5 likes?💀
i-
gonna be?
jailbroken sora💀🙏🏽
how is that your firs thought T-T
MKBHD - so SORA has been using my materials for weeks now...
😂😂😂😂
Double the speed limit from where I’m from would get your car impounded, fined big time and loss of license on top going to court in front of a judge with possible jail time. Yeah multiple police body cam footage of you has been uploaded and I’m disgusted.
ur crying cause another man violated traffic🤣🤣, grow up
U little sheep😂
Oh god... fake news and online scams are about to get so much worse! 😖 Also, RIP stock footage business
I've noticed in the freelance market some clients specifically ask for "NO AI". People, even non-creatives, Also - I think high quality unique REAL stock footage will have it's value! Let's hope.
Yes as a designer I exclude ai imagery! It’s good for somethings but not everything.
Someone else described AI in a great way. I'm paraphrasing but it went something like:
"It's like stepping on a full bus, but none of the people have any faces." Fake, uncanny, nightmarish.
That already happened in 2020, when everyone covered their faces with stinky moist gag rags, because the tv told them they could "catch" a "deadly disease".
Ok pal, tell that to the 7 million people who have died from COVID worldwide
12:53 Smoky fell on some hard times after all the wildfires in recent years.
What is the reason to tint your car windows?
it looks cool
14:10 that plant was used by so many tech reviewers 🤣
11:05, The way you "shed" a piece of your leg here is hilarious.
I didn't see that initially lmao 😂😂
As a video artist, I've seen some objectively cool effects leveraging AI. Transitions are where it really shines. Using it's inherent lack of object permanence to create match cuts, or intricate motion matching (like turning a dancer into dancing fire). These can be time consuming effects to create by hand
Agreed. One of my favorite instagram accounts creates videos that are nightmarish hence it isn't supposed to look realistic. The artist deliberately leans into the creepiness so that's why it works so well
No principles
Time that you get paid for, and that will change soon if we keep accepting it
Net negative for society
@@tbricksable Trends change as fast as tech. Truth is, AI will probably make these kind of effects seem dated as they oversaturate the market. Human creativity is what I'm paid for.
I wonder if AI can generate a video of you doing the speed limit
Me as an Amateur Photographer. I knew the bird video was real because of that Rolling shutter effect. 😂
Give it time. They'll perfect that too
same, and the birds in the back weren't glitching, after seeing hundreds of AI generated video's you can rather easily still tell them apart
You can see they’ve used warp stabilizer on it lol
@ but you can still see it if you can pixel peep 😄
Wait till they give it the prompt : “with rolling shutter effect”.
there should be more and more discussions on ethical and moral aspects of artificial intelligence, before it's too late.
It’s probably already too late.
most ai generated comment
It already cause too much damage.
If you have to say this then it's already way too late.
I have yet to see a law or any sort of regulation in the world that properly tackles the use of AI-generated content. Morality is nothing for people (billionaire moguls) who will do anything to maximize profit and cut costs. It's sad honestly, governments have yet to do something about this,
14:10 Pretty messed up,alot of tech reviewers have the exact same plant.
Pretty bad they're using content as of it doesn't matter at all.They are literally stealing stuff.
The plant thing was a super creepy reveal of how these companies are, in part, probably stealing.
11:22 Nothing. Nothing at all.
Not really, I have already found use for it at my graphic design job, I can easily curate ideas quickly without going searching for hours on end looking for something.
It will also make stock images and videos accessible to everyone now. Creating content has literally never been easier.
@mahjoubadam1742 As an aspiring artist myself I understand that easier and more convenient doesn't always mean better. The merit of these professions lies solely in the fact that the work is performed by people with their own styles and specialties. An author would never ask a large language model for inspiration on a part of their story. Maybe I'm just a purist, and maybe it's different for something like graphic design, but in my eyes AI has no place in any of the arts. The moment AI becomes capable of replicating what humans can do, and the moment you let it, is the moment it becomes advantageous to let it do everything. And that is something that might not be up to the artists themselves. Like I said, the merit lies in one place, and it's important to preserve it. Obviously I can't tell you what to do but that's just how I see it
@@coutherino I wouldn't call you a purist as much a regressionist, LLMs are really nothing but advanced search algorithms, just predicting the next most common word for the current scenario. Maybe it's easier on my soul to use AI because I studied it at my university.
@@mahjoubadam1742 I mean yea, exactly. I think the process you just described is antithetical to the nature of art in the same way sora is. If two things can create works that are virtually indistinguishable from each other (and AI is certainly close) they are only differentiated by the process they take to arrive there. AI is incapable of creating art because it cannot use the process through which art is created and defined. Not that that's what we were talking about, it's just my reasoning behind why AI shouldn't be used to create art. Or rather, can't
@@coutherinobut if you agree that AI generated content is just a mixture of everything it's seen, then using it as a reference make sense, because in references you're not looking for originality (you don't care, since the reference isn't your end product), you're just looking for a particular thing that resonates with you. So, instead of searching for these images/videos, you describe them, and AI gives you back what google would, but more in line with your prompt and use-case. It's not art, I agree, but as a reference it's no worse than using standard references (like 90% of the artists do). In short, if using AI as a reference is bad, then standard usage of references is also bad, in my opinion.
14:50 to everyone except every person living in europe😅
can we bypass that any way ?
I don’t think so same with Apple intelligence
It will become available at some point most likely. But I actually like that EU regulation is cautious with allowing these kinds of tools straight of the bat
@@Dév0x5762 Why? There is literally no use of not allowing it. I hate the EU. Its completely antidemocratic. Study it. Its a shithole
Sora: generate a first person action cam style video of a Lamborghini driving through a school zone at a respectable speed
please generate a video where you drive 35mph in a 35mph zone
I was about to go to study...thanks Maques
not even joking was about to start writing an essay lmaoo
@@Amir-rl7rqsame here 😂
Same here, .. exam in 8 hours
Exam in 6hrs….
Kinda crazy how Fast Ai is growing, from Chat Gpt and other Ai softwares as well.
It’s still in its infancy, giant paradigm shift incoming
Still not 96mph fast
1:08 wtf the left leg switched with the right lmao 🤣
She was just getting sturdy!
next thing he's gonna say is SORA generated the body cam footage. give this man a drink next time he drives.
AI sure is accelerating real fast.
Almost up to 96mph.
I just farted
If you call standing still fast. Tell me the difference between this and Sora's own demos from a year ago which if anything, looked better. Emperor's new clothes.
@bearimo2867 There is a difference between shown publicly available ai models and ones in the labs .
When sora was first shown it was far ahead of any ai video the ones in current labs are far ahead of this version of sora infact the next version was already leaked in a presentation it's really good and almost Hollywood level
@@tambal40 it's nowhere near Hollywood level. 5 second clips that take hours to curate to get right that still have obviously artificial issues like legs blending together, is not Hollywood level. And this is absolutely about as good as it will get. Like chat, there is limited training data and they have already exhausted it for chat, I'd not be surprised if they had done the same for video. AI is one big shell game and too many people are falling for it. The costs nowhere near make the value proposition. AI will crash hard, and fast, soon. The running costs are astronomical, and there is zero profitability to be gained. Open AI is losing more money daily regularly than any company in history. And that's with the advantages it has through partnership with Microsoft. Without them it is screwed. Tick, tock. It's inevitable.
It's a bummer that we already know we're gonna need a market for pre-GenAI data the same way we have a whole marketplace for "low-background" steel. The enshitification of the internet marches boldly forward.
Oh yea really good point. Though let's be honest - a lot of the text data was already enshitified by bots quiet awhile ago.
Lol, people don’t need AI to post nonsense on the internet.
Our kids won't ever believe generating videos wasn't a thing for us
My boy got a heavy foot?? That collage is crazy 🏎️💨
It's not true that the models work by generating images and just stacking them together. You can literally just look up the architecture and see that's not how they work. They take in whole videos to learn how videos work, including the time domain and the relative movements between frames. This is why they learn to simulate physics.
They did try just stacking generated images together 4 or 5 generations ago, but they quickly realized that didn't work. There was no consistency and you just had artifacts popping in and out between every frame.
Any video is a bunch of images together
@@jhanninnen yes but the images aren't generated in isolation, there's a time domain that's consistent across all frames, which is what makes a video different from a series of images.
10:30 “it will reject things it thinks is under 18” after showing all the memes it made with kids in it😂
Sora didn't creating those videos, those were from another video AI.
MKSpeedHD
OpenAI has to be super careful training on TH-cam tech review videos like MKB's. Their world model might assume all cars drive at 150mph
you should add "Potential child killer in playground zones" in your channel description too as that is obviously one of the things you do, right?
Looks like 59 in a 25
Bro is a speed demon 💀
The upload fuction is so cool😍 Unfortunately it is not available in Europe yet but I guess it's coming soon.. looking forward to it🤩