Sora Developers Explain AI Video Generation
ฝัง
- เผยแพร่เมื่อ 22 พ.ค. 2024
- In this clip, Marques talks with the team leading Sora over at OpenAI!
Watch full episode: • The Rivian R2: Are We ...
Shop the merch:
shop.mkbhd.com
Twitters/X's:
/ wvfrm
/ mkbhd
/ durvidimel
/ andymanganelli
/ adamlukas17
/ ellisrovin
Threads:
Waveform: www.threads.net/@waveformpodcast
Marques: www.threads.net/@mkbhd
Andrew: www.threads.net/@andrew_manga...
David Imel: www.threads.net/@davidimel
Adam: www.threads.net/@parmesanpapi17
Miles: www.threads.net/@carswithmiles
Instagram:
/ wvfrmpodcast
Shop the merch:
shop.mkbhd.com
Join the Discord:
/ discord
Music by 20syl:
bit.ly/2S53xlC
Waveform is part of the Vox Media Podcast Network. - วิทยาศาสตร์และเทคโนโลยี
I started counting the developers fingers just in case
Very convenient placement of their hands now that you mention it… very convenient 😂
😂😂😂
😂😂😂😂
100%, 1st thing i looked for!🤣
😅😅😅
The comment section probably won't help the poor guy to be less stressed next time...
Gotta start somewhere, it’s like a job interview. I bet so many of the people here ragging on him would melt under an interview at McDonalds. He’s representing a company where a single sentence can alter a companies stock in the amount of millions of dollars. We all saw the Fisker situation unfold. MKBHD pulls some massive weight. This small interview isn’t to be taken lightly.
@@caryrabbitWhat is bro talking about - OpenAI doesn't have publically traded shares
@@supersuede91 Right but you get the gist of what he's saying, that any poorly worded or communicated sentence he says can be misconstrued and hyped in the media. That's more pressure than the average employee will ever experience. You know that, and you know what he meant but you want to play semantical games.
literally the first thought at the poor guy in the red, it's nerve-racking being interviewed by Marques for TH-cam! especially when you are the one always behind the screen (developers) OpenAI being nervous to chat with Marques instead of the other way around, I love it haha
Well he was nervous, but who cares, many best people at their jobs are nervous.
Give Aditya a break, he's clearly very anxious in front of a camera! But good on him for at least being there and contributing to the conversation.
There's many secrets of SORA, i think Aditya is worried about saying he's supposed not to say
I thought the guy was a hostage at first.
He deserves a participation trophy
@@punkerverse 🤣
@@boiboiboi1419he’s the tom holland of OpenAi??
Aditya you did great my guy. Don't stress, I've also gotten so nervous in waaaaay less big meetings etc lol. Absolute legend for getting on this interview, contributing and being one of the smartest developers in the world!
Although Aditya seems really nervous here in the video, the fact that he actually decided to expose himself in front of the whole world (despite being really socially anxious) makes him really awesome and brave in my optic. Kudus to him, and brilliant discussion with the Sora team!
or maybe openai talked him into it
Aditya is doing his best to not make the same mistakes as his CTO!!
I think this was done before that interview and then the waveform guys pulled it and re-released it today - probably a Bloomberg exclusive meant they had to.
Leave Aditya alone, he did fine
Yeah everyone is overreactin
He did perfectly fine. What are people talking about?
I read all the hilarious comments about aditya before he even gave his first response thinking he’d be an awkward mf but he came off totally normal and not anxious
Seriously, I was expecting his voice to start cracking and stumbling over his words. Dude answered questions just fine. He just needs to chill out physical stage presence.
Being socially anxious must be terrifying. Aditya is going through it
😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂
At first i thought the red jacket guy was a video generated demonstration
It is. 🤫
My thoughts precisely.
Looks like a real time “Video Realism” Test to see what they can actually get away with at this point.
I say he’s video generated and the Open Ai team is having a real laugh about this currently.
Buddy on the left is one bad day away from being a super villain.
he's already putting thousands of people out of their jobs. Not far off
he's a legend. have some respect
@@matrix_multiply the legend on the left is one bad day away from being a super villain…✨respectfully✨
He is their boss. This video reminds me why I don't like working for other people.
He is the original creator of image generator
What did aditya see…?!? What are they doing to you adityaaaa??
This coment made me laugh out loud 😂
He saw what really happened to Ilya Sutskever…
@@jakejad4427 Fr. All those a's
I morse code his breathing, and he needs help. He watched the news about the BOIENG whistleblower.
lol
"In some ways we see modelling reality as the first step to be able to transcend it". Bro waited and conserved all his energy to drop one of the hardest line and he did.
This line really stood out to me. Transcending reality… Alex Jones has been talking about people in the AI community transcending reality into other dimensions for years.
This felt like an AI generated reboot of The Big Bang Theory
So true😂
Oh yeah
I for one, Love Aditya's whole persona here. He is nervous, he is human!
And this should show people that even the BEST of the best, aren't perfectly just strolling through life. Being nervous to chat with one of the largest most consequential people in consumer tech, knowing he knows what he is talking about so you can't bullshit him much, its nerve-wracking for anyone. And seeing the nerves actually made me feel better about myself and how I hold myself, if he can do it, then so can I.... idk it sounds silly but we need more real people like that on display
Pretty funny talking about being human when commenting on AI gen imagery, the thing that's getting rid of those pesky humans out of the creation of art.
Aditya is probably a legendary programmer. As an IT student I am super jealous of individuals like this. I would happily giveup my social skills to be at his level
Bro looks like he was told about the call right after he’d already started his weekend with a fist full of gummy bears.
Aditya you rock dude. And very heartwarming to read all the nice comments. I've been crazy stressed in important meetings too... I think everybody has... No big deal. Keep up the amazing work with SORA... Cant wait to be able to play with it. It will, without doubt mark a clear before and after for our civilization.
Aditya is prolly just anxious but he looks absolutely terrified, like he’s literally flinching & breathing so deep, who convinced this poor dude to take the interview
i dont get why people cant see being interviewed on youtube for millions to see, by someone like Marques is nerve-wracking for most.
The guy in red jacket was very nervous in the beginning but got a bit more relaxed towards the middle 👍🏻 good for him
look how fast the central guy dodged the “where you train your data”? kek
buddy changed the subject real quick
Focused on the "cool" stuff instead of the ethics. Typical tech bro mindset
Tech companies use a weird logic of they should be unethical because someone else might be unethical later which means it’s perfectly fine if they are now. Plus they know there are no negative consequences for being bad
@@DustinHasVideos as a software eng, this is sooo true and I'm also guilty of this, but my religion is literally the only reason why i keep my ethics ...
he answered it better than the CTO, how did she get that job being so bad at meetings and interviews is a mystery to me
The video was supposed to be about Sora.
But Aditya stole the thunder.
I like that guy. Handled it really well!!
Aditya just want to do his work and that's it. No cameras, presentations, interviews, etc. You did fine Aditya ! 💪💪
“What was it trained on?” “uuuuh we can’t really go into details”
‘TH-cam, we downloaded all of TH-cam’
Congratulations on your success and here's to the incredible innovations yet to come!
Guy on the left is being held hostage.
Maybe slave labor?!
Lmao!!
@@marcusfonseca6673he is their boss and original creator of the image generator
hahaha I thought the same thing
I peed my pants laughing at this comment 😂😂
ADITYA if you are reading this - you are my hero. Instruggle with panic atacs and i would just runn off screen 😂 good job holding your ground. This is a big victory my man!
God bless Aditya. What a legend.
The guy on the left is tweaking bro. Why is he taking so many deep breaths. Damn bro take it easy
lol maybe he ran over or perhaps could be a personal health thing..
Probably nervous
He looks like he was in a straight 16mins of fight or flight.
He doesn't know what's real or fake, Sora f**ed him up!
Probably a health thing and can’t help it. Probably just introverted. Also taking it easy isn’t as easy for some as it is for others
The next big thing would be, if OpenAi can create "environments" like one in Apple vision pro based on prompts and react to our actions in the environment, maybe an environment of the jungle and we can touch plants and stuff
“and stuff”
Though the guy in red jacket was AI generated for a while
lol me too
Because he’s not white?🌚
😂
When I saw the red jacket guy I RAN to the comments.
Marques is gonna pull a classic MKBHD plot twist a week from now telling us THIS video was generated with Sora
the recent interviews with the CTO would explain the nervousness of this team lol
Aditya was normal.... he's just a dude with some important thoughts to share.
I think this technology would allow the creation of a new entertainment medium called “Cyber Dream” which would replicate actual dreaming while awake. It would generate 360 degree video content in VR/Spatial computing but the content can be interacted with by voice to create iterations on initial scenes requested (extrapolating). The difficulty would be consistency, locomotion, sound, etc but dreams are sometimes also inconsistent so you wouldn’t expect it to be perfect.
Man he is having panic attack. Good job not running away. I would run away xD
“Modeling reality as the first step to transcend it”. THAT is the mic drop moment right there. Well said.
So random seeing you here lmao. Stay blessed habibi.
I loved that!
Insert distracting marketing chatchphrase here "................."
That will take a long time. I look forward to seeing that, but it essentially means the model will have to know all laws of physics and know how to apply it to the specific scene. Telling a model to do that will be hard.
So, Aditya, however anxious he was, took the cake at the end with his killer comment! :D
Why does the guy on the left look like he's just seen a risen Freddy Krueger.
Great conversation!! I won't be surprised if what Aditya sees is kinda like when Neo sees the Matrix, that's the only explanation :D
Bruh this is me at every presentation… I feel you bro
I would like for AI to make a movie/Anime where the story can change and adapt to have different endings... would be cool to see differents outcomes of certain situation
Aditya is a looped video created by Sora
😂😂😂😂
The future of video and media production is gonna get weird
Superb questioning & excellent delivery by the leads of Sora.
this why we keep engineers behind monitors lmao
"people are terrified of the advancement in the field of AI" - Aditya is those people
Thank you, Sora Devs. Thank you, Marques. I look forward to making new art and learning new things from the technologies you're advancing.
Aditya was just excited to talk yall, the moment he started talking, you can see him ease up
Aditya is thinking About : Devin AI Release
At the beginning I thought Adytia was made in Sora. But then I realized he was made by Dall-e
Bro aditya having a terrible time
Probably he had to binge eat his lunch to make time for the interview. Eating salad before higher glycemic index foods can regulate energy absorption.
@@GenghisbeardProbably he just had social anxiety and it had nothing to do with what he ate ten minutes before lol
I think they could improve their model even more by training the A.I. on tracking specific object transformations in scenes and stories. Do this for the same set of videos, but tracking a different object for each one.
Then compile them together with a GAN to compare and improve itself.
14:17 ‘In some ways we see modelling reality as the first step to transcending reality.’ - Aditya
Once videos are done, going into spatial video generation would be so amazing
This looks like a poker table... all 3 have things to hide, but with varying degrees of a poker face, lol Increases from left to right ;)
Adding sounds? They are definitely working in that front. They couldn’t hide surprise in their face.
I wish you would've asked them more about the ethics of making this.
Same reason they didn't ask Tesla about their ethics. Don't burn bridges for future collaboration which dilutes truth
Maybe because there are no ethical issues. What ethical issues would there be?
@@codycastinsane take
@@GrantBrees nice argument
So what are the ethical questions?
@@codycast no point in talking about it with someone who already has their mind made up
A novel in and a movie/video series out is what I want.
It would make sense if every company developing AI generative models were required to include a classifier for the generated output. This would enable social media platforms to automatically detect whether the content or its derivatives were generated.
really looking forward to collaborating with SORA
11:14 - Mira Murati said recently in an interview that Sora would be released this year, possibly maybe even within a few months.
Another approach that could be really cool is making 3d modeled decorations and actors with AI and then AI should animate them in a 3d animated scene. So it could be cheap computational to adjust it to desired state (cameras, animations, effects). And the final step should be superresolution rendering. Yeap it's a bit more complicated workflow but for complex projects it will be much more controllable and consistent, cheaper and faster to get what you want.
This is going to be amazing for short establishing shots a b-roll
Aditya Seems Fanstruk ?????
Aditya is not nervous, he has the facial expression of someone who is inwardly focused.
will it be able to generate accurate images tho? for example if i ask it to draw a certain graph it just comes up with a creative version of a graph instead of the values i wanted it to represent; like with stock market data.
It would be useful to be able to feed it audio, and have it sync. Speech in particular. So it doesn’t have to generate the audio, but it will generate characters who speak the audio you feed it. Big ask.
Bro snapped when he said, “In some ways we really see modeling reality as the first step towards being able to be transcendent.”
this sums up those three guys and others just like them ''you either die the hero, or you live long enough to see yourself become the villian.''
First i thought Aditya is AI generated but then i realized he is more human like than most people on camera with their dake expression
Real world applications…take video of a crime scene and, like a classic detective movie, it generates a video to help solve the case complete with bullet trajectory and freeze frames.
I for one am more confident in Sora knowing that’s its Team Lead isn’t some overly salesy type.
14:14 "In some ways we really see modeling reality as the first step to be able to transcend it"
It seems hard for Marques to digest when Aditya speaks out of blue, and it’s even harder for me.
But I can totally feel how nervous Aditya tries to express the technical terms to the public, in a sci-fi fashion. I root for him!
In terms of the protection measures discussed, what if AI videos are screen-captured, circumventing the “metadata” / “classifiers”?
So are patches are specific pixel size? Like 64x64 or smth?
11:13 "Not (available to public) anytime soon". Meanwhile CTO Mira Murati, answering the same question from the WSJ: "definetly this year, but could be a few months"
Imagine a fully AI generated game. Like you give it a prompt and it produces a fully playable 3D game. 🤯
Maybe in 2100
@@dbog5214 hopefully we’re in the Matrix by then
…or out of it depending who you ask
This explains why different patches of sora videos look like they have slightly different perspectives and don’t all move in sync, especially in complex backgrounds like forests and mountains
Aditya is either terrified or created by Sora
Aditya talked to the AGI before attending the meeting 😁
Aditya seems like AI generated
came here to say this lol 🤣🤣
Well actually, he .... is.
Aditya that's it you're a star now !
So are they stealing from creatives with no consent or compensations to train their models like with image generation too?
Aditya said few words but that one line he said just was a slam-dunk!
"public videos or licensed by OpenAI"? that's a pretty vague response.
they didnt say videos in public domain ... public videos can be youtube videos as well. it's insane whats happening here and nobody seems to really care.
@@Robert-ci6sj exactly. Vague.
Well at least that journalist that interviewed OpenAIs CTO cared to ask.
The guy spoke of 'SORA 1' in the past tense. Cool.
Aditya may seems nervous but the transcend reality comment is brilliant.
a next step could be getting it fast enough for real time interactive simulation/video games and then into VR with like 16k resolution per eye ... so 32k real time video and audio generation basically ... in 10 years maybe
As soon as he spoke he was totally fine, come down guys
What was it trained on? We cant say. OPEN stands for what exactly in OPENai?
Can Sora be developed to generate stereo pairs of its clips for a full immersion experience?
The futur of this tech could be realtime brainstorming.
The background is when the matrix changes something 😮
these guys are headlining the art apocalypse
Aditya, I feel the same way around my YT coworkers during important meetings.
I would like to use Sora to fix films with bad endings, or even fix films like Madame Web that would need a massive overhaul to be considered good.
Does sora do their video generation and processing based on copyrighted images/video?
the guy in the red knows something that we all are waiting for and he couldn't hold it and said somthing at 14:17