Це відео не доступне.
Перепрошуємо.
Does SORA make AI videos using stolen YouTube videos?
Вставка
- Опубліковано 15 сер 2024
- See the full interview here:
• OpenAI's Sora Made Me ...
OpenAI's CTO sat down with journalist Joanna Stern to talk about Sora, the company's new AI that can create videos from simple text prompts. Like all generative AI, Sora needed to be trained with heaps of data. Just where did these videos come from? Good. Question.
Join my Patreon:
/ anoraker
The products I use to make videos:
Camera: amzn.to/3XuhmcG (affiliate)
Camera remote: amzn.to/3RAqrwP (affiliate)
wireless microphones: amzn.to/3xnQe4...)
Mic charging case: amzn.to/3XsX3N0 (affiliate)
Lavier: amzn.to/3ze2LrD (affiliate)
Tripod: amzn.to/4c0Y4jF (affiliate)
Carbon fiber tripod: amzn.to/4c3JuH...)
Teleprompter: amzn.to/3Vv5ZPe (affiliate)
Google's "answers" are often answers to question I didn't ask. Lately they contain nonsense, misleading and sometimes dangerous content. I didn't realize that it was AI content but it makes more sense now.
What's even more scary is an "AI feedback loop", where AI trains itself on its own garbage output, thus reinforcing its own AI hallucinations... Then the ground truth will become even more obscure.
Exabytes of Brute force plagiarism... no references cited. AI could indeed destroy humanity...
Lol, the actual thing that will save the human race will be innovation. Not copyrights patents or anything like that. We need to move forward fast and solve some serious issues, with and without AI.
@@GenuineFlolieThievery is not innovation. In fact, it will detroy humanity due to its high carbon footprint which will contribute to climate change. Congratulation for being big coporations' bootlicker.
I wish she just said: "yes, we used everything that was publicly available. I see no reason not to." instead she made clown of herself
Would have been more honest at least.
But that would be untrue, they are very strict about the quality of data they train the AI on. They don’t train on literally all publicly available data, it makes much more sense to do something like train on the top 10% highest quality of data.
@@default2826 Maybe they are, maybe they just claim they are. This person either really have no clue and should be fired, or she is lying from some reason. Either way, she clowned herself and I lost any interest in this project.
Man, I feel so bad for you, theses AI bros in your comment section have no knowledge, critical thinking or humanity. (Some of them are probably bots) Don't bother to answer to them, believe me they won't even listen to you because the truth hurt them. When I see such comments, I'm scared for what humanity is becoming because let's be honest, many of us are really REALLY dense to the point where we would celebrate and cheer our own destruction and think that massive thievery made by big corporations is innovation and OKAY.
I sincerely hope that this AI will be banned. No one consented for their videos, ilages or anything posted on the internet to be trained to create fake videos. I don't see any good use of it, especially when you think about all the bad consequences, there are just too many. It could be used for fake news, deepfakes, p0rn or even propaganda.
Are we really going to work at McDonald's while machines will do art and create content ? Literally the fun things ? (Of couse, lazy people will say yes because they didn't even bother to pick up a pencil.) That's not the future anyone wants. In fact, I don't want to live in this dystopic world. What a sad world.
I'd be happy with universal basic income as I'm unemployed. Also a computer science engineering graduate😭. All interviews I attended they wanted someone better... maybe AI will take us freshers jobs😞 I hope not atleast
In general, I actually have no issue with the concept of a UBI. We already have some version of it or another in many forms, and I can think of far worse things we already have our tax dollars spent on. But with the way things are going, this will be a necessity and not just a "should we consider this" thing.
Facebook ToS was changed a long time ago to allow AI scraping. You've agreed to it if youre on there.
Nope, you're thinking of Twitter which recently updated its TOS to allow it to use Tweets to train its ai (a big can of worms to be sure).
Facebook never updated its TOS for AI, and has been actively suing companies that scrape Facebook user data to train AI.
And while it admits it does use "public Facebook data" to train its AI (not great), it has an opt-out option (better than nothing, but it should be opt-in).
A lot of this needs to go through courts. Yes companies are going to try and get away with whatever they can until they're told to stop. But going, "oh well" isn't going to help that process.
so it's ok if humans can learn from youtube just not AI ? hmm sounds awkwardly wrong imho
This is called a straw man argument, and isn't even a very good example of one.
@@anoraker perhaps your just looking for an argument - if it's public and your NOT reproducing it but taking inspiration from it then you have NO CASE!
that'snowhowthisworks.gif@@oshinoyoshida
@@anoraker Actually it's a great example. We have managed to distill the human ability to learn down to a series of nodes and biases, and emulate them on computers to achieve uncanny human-like responses and behaviors. We could be on the cusp of creating something that is indistinguishable from consciousness, and people are upset that we want to feed it the same every-day materials that humans consume. The same stuff that influences decisions, inspires creativity, and quite frankly, is in essence, human. This is your opportunity to input a bit of yourself into these systems, and you want to run away because you're worried that AI may emulate that part of you.
Yes. Like every things else Ai.
"Fair use" makes your argument specious.
Except that it doesn’t. See every “fair use” YT video that’s either pulled down or demonetized. Fair use is nowhere near as broad as people think.
I feel that using the word "stolen" is very vague and it's more of a hateful word than anything, why not borrowed, trained on, or inspired? By that logic, an artist being inspired by another artist is considered "stolen". Under US copyright law, an idea, style, etc, can't be copyrighted. If it is, then Palworld would've already been sued by Nintendo for its infringement.
You can only borrow with permission. When you take without asking, that's stealing. If your wallet is visible to me, and I just take it without asking... are you really going to accept saying, "it's hateful to say that I stole your wallet, why not say I borrowed it?" Under US Law, artistic works (whether in video, written, or drawn form) can be copyrighted, and is from the moment of creation. AI isn't trained by "ideas" but by the literal art (visual and written) created by others... without permission.
Let's say the ai watches UA-cam through a screen with a camera, is it ok?
What about the HDMI feed? The stream feed from memory?
That's what humans do with this content, they take input and it alters their brain...
Same with AI.
It watches the content (albeit very fast) and change its structure.
It doesn't steal it, just learns from it.
This could be a valid case by open AI
Except AI isn’t human and doesn’t learn. And it doesn’t do any of that. It copies it, changes it slightly and often just straight up reproduces it.
And the Supreme Court has already ruled that simply using technology to alter how you are stealing content to mimic human behavior doesn’t change that it’s stealing content
Yes I have and do. I've written and studied about this topic extensively. UA-cam isn't my full time job, but knowing about this topic is at this point.
And yes, yes it CAN make copies, and that fact is the subject of SEVERAL lawsuits right now. There are plenty of instances where people have shown AI making replica (or near replica) of both copyrighted written content and copyrighted artistic content.
Those lawsuits are too far to new to know if they will be successful or not. Just lots of pundits taking guesses.
And as far as copying, the Supreme Court actually has a relevant ruling. Using a technology method that accomplishes the equivalent of copyright theft still counts as copyright theft, even if the technology is designed not to do it in a literal sense. @veqv
@@anoraker what you're saying is that if I train and learn to draw like Akira Toriyama (RIP) and draw Sonic The Hedgehog in his style I'm infringing also on the copyright of Akira and not just Sonic's?
What if I draw some new character in his style?
I COULD draw now an exact copy of a dragon ball cell drawing which would be copyright infringement, but that doesn't mean that by learning and being possible to do so, I was violating the material's copyright.
fully agree to that. Besides, there are humans that can exactly reproduce something they've seen, so are those illegal and should be terminated ASAP?
Just because an AI has a way better "memory" than a human does, doesn't mean it should be illegal. It is, in principle, learning the same way.
Besides, an Image AI won't 100% replicate the original, it can be close, yes, but you won't get a 1:1 copy of it, there will be noise or alterations. The same is true for my human example above, it might look the same, but will have some slight alterations.
I get why artists are against generative AIs--and I keep seeing the argument that AI is stealing our artwork--but, from what I understand, that isn't really true.
When an ai generates something, it doesn't copy and past artwork its seen, change the colors or one detail and present it--that isn't how it works.
From what I understand, Trianing is just showing an ai what a dog should look like, then the ai usess the various images it's seen to piece together something that resembles a dog.
It feels like people don't understand how generative ai works, maybe I dont either, but from what I have tried to learn about it--It doesn't seem like stealing. I'm open to both sides tho, I just dont see strong arguments anywhere about how training is actually theft.
using UA-cam videos so what?? If you don't like use your videos to train AI don't share videos
That’s not how it works at all. And do you really want the future where that’s the choice?
If you don't like your car to be stolen, don't park it. or even better, don't drive it anywhere !
Maybe we should read the terms and service of these social media platforms including UA-cam lol. Once it’s on the internet for free it’s for everyone to use my friend . And they not copying your videos exactly how it is. They are just using random videos to feed their ai data to train it, it’s not stealing 😂
Again, no. That's specifically not how it works, and not what Terms of Services say either.
Specifically the TOS of every social media, (and even moreso UA-cam) state that we retain ownership of our content, and all the social network in question has the right to do is show that content to others for the express purpose of viewing ... and viewing only.
Heck, UA-cam goes a step further and specifically says that users of UA-cam only have a right to view content.
"License to UA-cam
By providing Content to the Service, you grant to UA-cam a worldwide, non-exclusive, royalty-free, sublicensable and transferable license to use that Content (including to reproduce, distribute, prepare derivative works, display and perform it) in connection with the Service and UA-cam’s (and its successors' and Affiliates') business, including for the purpose of promoting and redistributing part or all of the Service.
License to Other Users
You also grant each other user of the Service a worldwide, non-exclusive, royalty-free license to access your Content through the Service, and to use that Content, including to reproduce, distribute, prepare derivative works, display, and perform it, only as enabled by a feature of the Service (such as video playback or embeds). For clarity, this license does not grant any rights or permissions for a user to make use of your Content independent of the Service."
Now before you try to twist what the above words mean, you should know that all of it is boilerplate lawyer speak that does nothing more than give Google the right to copy my content onto its servers all around the world, and then display it to users... and for user to "copy it" (that is STREAM it) specifically by playing it through the UA-cam app.
And take extra special note of the line: "For clarity, this license does not grant any rights or permissions for a user to make use of your Content independent of the Service."
So please, take your own advice and read the TOS. Just because you've made something available to view doesn't mean you've given permission to use that content for any other purpose than viewing (especially training AI to copy and reproduce that content).
Actually if AI trained on AI generated content it will create a junk.
You've like... missed the point entirely.
It's called a dreaded AI Feedback loop... Even worse, as that most LLM's are only 70'ish percent accurate, so feedback loops can destroy the truth, and the baseline truths will become obfuscated. I've already seen so many falsehoods in the Google synopsis results... Our future could become an AI hallucination.
No one should care! Let AI be trained on ALL material. Copyrighted or not. The world will get better AI that way - which benefits all of us. We do NOT need "copyright holders" to EXTORT MORE MONEY from those companies - which will 100% passed onto us - the public.
Also: If a PERSON looks at all of Van Gohs paintings - and then imitates him by creating an ENTIRELY NEW painting - but with his influence / technique / style. Then that does NOT de-value his paintings. The same goes for AI. It does NOT copy-paste. It transforms the work. Just like humans - INCLUDING ALL THE CONTENT CREATORS !!!!!!!!!!!!! - do. Thats how inspiration in humans works.
And what’s the incentive to create material in this scenario?
No one should care? How do creators pay their bills then? Exposure won’t buy groceries.
@@MitchellKinard''everyone wants'' no. You and your friend want it, not me and the majority of the population. AI is nothing without humans and our content. It doesn't create ''better content'' but reproduce a bunch of already existing content