Sora Release Date Deep Dive
Vložit
- čas přidán 22. 05. 2024
- Let's take a deep into the recent Sora interview between Mira Murati and the Wall Street Journal, regarding the release timeline for their highly hyped AI video generator Sora. Just days after the tech leads stated they had "no timeframe" for Sora's launch, OpenAI's CTO gave an interview claiming it could arrive within "a few months."
Did the pressure from AI competitors like Pixverse and other AI startups influence her aggressive projection?
Additionally, I showcase the cutting-edge work of Fable Studios, who created the viral AI-generated South Park episodes. Their latest project "Thistle Gulch" blends interactive fiction with emergent AI storytelling in a compelling Western setting.
LINKS:
Video on Haiper.AI: • New FREE AI Video Gene...
Thistle Gultch Beta Signup: blog.fabledev.com/blog/beta-a...
MKBHD's Sora Interview: • The Rivian R2: Are We ...
Chapters
0:00 Intro
0:33 Contradictory Sora Release Timelines
1:06 New Sora Footage
2:00 How Long Does a Sora Video Take
2:36 Trained Data In Sora
4:03 Sora's Release Date
4:15 Sora's Technical Leads
5:06 AI Video Generation Arms Race
5:29 Is Sora 2 being worked on?
6:08 Technical Details about Sora
6:47 What will Sora Be Used For?
7:50 Intro to Fable Studios' Simulations
8:33 Thistle Gulch - Blending Interactive Fiction & AI - Věda a technologie
For a Free & Available Sora Alternative, check this video out: czcams.com/video/nU7lCnVmcBI/video.html
Your content delivery is easily "digestible." Thank you.
Thank you! Very much something I strive for. I may also cause indigestion, so make sure to have some antacids nearby! Ha!
7:06 - As someone who has been around a long time in the creative world, I agree 100%. Even if what you produce now is junk, nonsense, not something you would share with your best friend, this is how you will learn and be best prepared for the future when Sora (or Stability, Kaiber, etc.) is released, and refined.
That is 100% it! Also, I think there are some great use cases already by generating “raw material” and using it in After Effects or other compositing software to create some pretty unique things.
Maybe not full movies, but certainly interesting videos!
Thanks for this overview / update. What I'm curious about is what SORA video's look like when they are based on an input image, i.e. from MidJourney. It seems to me, that would give people even more control over the look of the video... like that recent EMO lip-sync, singing algorithm. A hybrid of those two could be really powerful for film making with dialogue or a music video.
Very much the same! In the Sora research papers, it indicates that it can be done, but I don’t think we’ve seen examples of image to video yet,
I’m wondering if it doesn’t work quite as well at their T2V model?
Thanks for sharing. Appreciate the work you do.
thank you so much! Really appreciate you as well!!
Thanks for telling me about the gultch sigh up beta I had no idea this existed
I can't wait to get my hands on it-- I've been following it since the South Park thing and have been super excited about the possibilities!
Great video!
I first got to this channel from the Boximator video, wanted to ask if you've seen and covered the recent Emote paper, it's almost flawless..
All the best! ❤
Emote? Or Emotalker? I did Emo, haven’t covered Emote yet! So much Emo!!! Haha
@@TheoreticallyMedia no it's the same :) with the unbelievable lip sync and emotional expressions. I think that was one of the best demos I've seen. And is actively being reversed engineered by the oss community as we speak
I'm just going to repost (quote myself) from what I said to you about your video - "HUGE AI Video Gamechanger!" - on Boxinator from about a month ago, of which you agreed with at the time.
"...the main problem I have with these previews of new AIs shows the crazy cool stuff they can do is that by the time they are released, the already existing ones (Pika, Gen2, etc.) have a good chance of having these same features or even better. So it looks now like they are so much better than the competition, but the competition isn't showing us what they are going to have available to us in 2 to 3 months. They're showing us the great things already available to us."
It seems this could end up being the exact same case with Sora.
I don't think a tech working on the thing calling it "sora 1" means anything more than that internally they're tracking development milestones with a fairly typical versioning scheme, as you'd expect any developers to do anyhow. Like, there's certainly some requirements list set for them and/or by them for a 1.0 or 1.0.0 or however exactly they choose to do it, along with a bunch of intermediate goals they had before being feature complete for this no doubt.
Very happy to have found your channel, a much more measured tone and more substantial research to make the vids than a buch of others in the yt AI space I've previouslly run into!
Oh, I really appreciate that! I know what you mean about some of the other channels. I think there’s often a rush to get a video on a topic out first, obviously to get the view count- but that always comes at the cost of taking the time to research.
I shouldn’t say it’s a straight YT problem either, websites also do it all the time-
But I do strive to get my research reps in, while trying to cover a breaking story, and…y’know, get some dad-jokes in as well!
Appreciate the comment and happy to have you on board!
When the two women in an interview showed up at... 1:43, Perfect editing because until you mentioned it, I thought it was real! As for transparency, see, this is the issue. If companies were to just be bloody transparent about the source of their data, there would be less hate and more concrete grounds for lawmakers to actually make actionable decisions. That would create a less toxic environment in communities and certainly allow for 'ethical progress' (whatever that means). It would also make many of us more comfortable in actually creating, with defined laws as opposed to blurred lines drawn in the sand on a windy day! Easier said than done, I'm sure, but no one is etching the rules in stone; not without vagueness, and not without them applying globally. Great video Tim. 👍 Thanks...off to watch the one I missed about the free video generator!
I can’t wait until Google jumps in hard with theirs, which is sure to be trained on like, every CZcams video ever. And the thing is, we can’t complain. Although I haven’t read the massive T&C document, I’m SURE I’ve agreed they can use me as training data.
Now, my data is pretty limited, Prompt: guy at a desk talking about AI, haha- very limited use there, but man- like, every Mr Beast video? Almost every movie trailer? It’s gonna be crazy. And we all OK’d it somewhere…
thanks a lot for this, and for your advice regarding Sora!
1000%! Hope you’re making something awesome today! No need to wait!
Yeah, I saw that yesterday when the chief tech officer doesn’t even know what their video creators were trained on, that is super shady
Oh, she knows. Or at least has an idea. But at that level, they’ve got to be careful of what they say- since any quote is going to be used against them in the future. Esp considering the lawsuits they’re currently fighting.
I have the feeling (given the body language and tone) this question wasn’t cleared beforehand. Which, good on the reporter for doing if so!
4:11
same here - I'm EXTREMELY skeptical of that time frame
Maybe at the end of the year? By summer? I doubt it. That said, I do think we'll see something LIKE Sora dropping from someone else by then. Someone who maybe isn't as concerned about the guardrails. OAI has a LOT of eyes on it-- but, I can see a small AI Startup (funded with a ton of money) who will just take the leap to get put on the map.
@@TheoreticallyMedia
Mind you, for quite a while, I was a die hard MJ and Stability AI guy......... that is until OpenAI blew my mind with Dall-E 3.
I HIGHLY doubt that any company will drop any AI tool this year that even remotely compares to what we initially got a taste of with those extraordinary SORA videos.
I'm convinced that OpenAI is so far ahead of the competition in the AI imaging / video space. Dall-E 3 has already been out for quite a while and to this day, nothing else comes close in overall capability.
I'm having a blast using Dall-E 3 in Night Cafe where there's easy option for changing the 1:1 ar to portrait or landscape with 1 click as well as easy style prompting using built in custom styles.
You are right, dont wait. There are new videos generators out every week. I keep testing them for my channel. One day Sora will be here and I am sure it will change the entire industry. But get real. While for sure it will provide us with stunning new videos it will not be the ultimate solution.
Exactly it! I see too much chatter lately thinking that Sora is going to be some mega model that will destroy Hollywood. It’s a LONG way from that!
But, still going to be an incredible tool for those who figure out ways to use it in novel ways!
Still gotta put in that elbow grease!
6:45 my close caption says "Bill peoples", same thing when you showed the Sora team ITW, can you help me understand ? thanks!
The caption is supposed to be Bill Peeples. I guess the auto captions made it Bill “Peoples” which, I guess is reasonable,
I’ll try to edit it.
@@TheoreticallyMediaThanks Tim ! I checked straight away to see who he was, and realised he was the guy on the right in the Sora itw video with his name written in big above his head haha. I guess I was too caught by Iditya. (BTW it's Peebles with a B haha)
Exciting times 🚀
I’m cautiously optimistic about Sora. I still think that when we see something that looks like Sora, it won’t be from OpenAi, but rather from someone who is currently grinding away in stealth mode right now
Regarding the timeframe: I can easily see it being a matter of waiting for the boss to give the timeframe estimate to not overstep their role(s). If they were not given explicit directions to reveal an estimate, better to let the boss make the statement. Sounds like a more plausible solution to the timeframe discrepancy. At least to me.
Oh for sure. I also don't discount that within 5 days things might have changed, particularly when you factor in the breaking news about funding for AI Startups. It might have been a case of "Ok, it's go time" within the matter of a day.
Also, just because you announce "maybe in a few months" it doesn't mean you have to stick to it!
woaw ! 4:20 the bit you chose shows Aditya Ramesh having a serious panick attack it seems. I haven't seen the actual itw, people have probably ccmmented a lot about that. The 2 other ones are very pale, must be tough working on Sora
Haha, I noticed that as well! My guess is that he doesn’t do a lot of press, or time in front of a camera. You can legit see the terror and tension in his eyes.
I’ll say, the wide angle lens doesn’t help either. Since he’s on the far left, he’s also catching a little lens distortion, which is certainly doing no favors!
@@TheoreticallyMedia Haha true, maybe he rushed to be on time which would explain his fast breathing which cools down later on. I can relate to be uneasy in front of a camera particularly in their case.
I believe they are offering a "Sora" lite version of the larger model, which explains why the videos demonstrated in the interview feature many hallucinations. This is why they can release it earlier: because the video is of low quality, and high hallucination means it is easier to identify whether it is fake or not. Another reason is that they cannot deploy the larger "Sora" version to the public due to potential backlash from Hollywood, or because they don't have enough GPUs for that. Interestingly, One of OpenAI's newest board members works at Paramount, so maybe that is Hollywood restricting the model's full capabilities.
Agreed on a lot of that. I think we’ll first see a pretty heavily nerf’d free version, if anything.
I still say we’ll see “Sora Like” from a 3rd party before we see it from OpenAI. Likely, by the summer, which is why I think Mira said “in a few months”
I especially agree with you on your comment about not waiting for the right technology to arrive before you decide to start creating. That's how you remain perpetually behind. I think Sora is going to be great for asset creation that compositers can then use along with other methods to do things that have never been seen before... like I'm doing.
Good perspective, I never thought about that.
Untill you see the cost to use it..
Yup. That's a perfect use case here. I've got a video exploring that coming up. Shoot Green Screen (or, heck, just use an AI Roto Tool) and generate backgrounds or environments with the current AI tools.
Actually, there's a pretty cool "all in one" app called SkyGlass that I'm going to dive into-- hopefully next week?
I keep wondering that. I think OAi will release a "lite" version (4 seconds, maybe 720) for free-- but then we'll have a longer 1080 version that will be paid. But, how much will that be? I could legit see up to $99/Mo for full access.
Just a guess...
99/Mo + API costs.@@TheoreticallyMedia
Can sora animate image, or is it just pure prompt/text to animation? If you need to work with image, I think EMO is better, cant wait for its release
I'm sure it can but that will be determined by the iterations of its release
The documentation on Sora indicates it can to Image to Video, but I don't think we've seen any examples as of yet.
👋 looking forward to this
Heya Louis! This one is FOR SURE for you!
They HAVE to release Sora in a few months, simply because of competition.
It would be naive to think that other companies won’t be capable to soon surpass Sora’s current capabilities.
I low key think that was the whole disconnect between Mira and the tech leads. That “in a few months” line sounds very much like it’s coming from behind a desk. A desk that knows the wolves are catching up.
Great advice Tim, use the tools we have and make ART now!
As a musician, you know! You can’t wait to have the perfect piece of gear to write the song!
Guitar players, of course, being notorious for that. “I do it if I had the right pedal!” Haha.
I only say that because I’ve been guilty!
1:31
That's the coolest camera .......I want one! 😉
Seriously! I have no idea what I'd do with it, but I WANT IT!!
SHOW-1 is quite nice model. It's instant lofi film look. Didn't even knew who made it 😂
It’s gotta be Sorny! Or maybe a Pamavision? Gotta hit up some pawn shops to see if I can nab one!
I wasn't expecting such a thorough investigative exposé, great info Tim
I may have gone full 60-minutes on this one! Haha-- Thank YOU!
Not anytime soon could mean a few months as well. Just because somebody says not anytime soon does not mean any longer than a few months. A few months from now is definitely not anytime soon in itself. They could be meaning like anytime within the next few weeks. I don't see anything fishy about that statement, neither does it contradict the other interview
So, it isn’t. I just find it highly unlikely. Mira might also be dodging, while knowing they’ve got a secret event next week where they’re going to drop it.
Same goes for the tech team, but, for sure there is at least a disconnect in terms of messaging that happened over the course of 5 days.
I believe that’s due to the increase in funding by those startups. OAI realizing the wolves are catching up.
I believe the footage that openai is showing is a real thing, I see so many AI mistakes such as cars driving on all sides of the roads, direction etc. you do see glitches but compared to pikalabs or runway this is soo much better, it will get even more amazing including sound effects, voice and lipsync features.. my prediction based on the market and industry developments. what we see happening over and over is that they release the tech in smaller plugins(companies) such as free AI discord image/video/audio/animation generator platforms, where we the people actually collectively train the next beast.(been happeing ever since google/AI) so expect generators in these categories to improve in the next few months, even generators that will combine(all in one) for a better end result, longer stable clips, more control etc. I think the next step then will be end 2024 that we see a first Sora release(alpha/beta) which will rely heavily on the GPT 5 model soon to be released... if we want AI we need to stop complaining about copyrights.. if we look past that creative entitlement(training data) and see it as our collective creative baby that will serve us all, the horizon is broader and brighter than one can imagine.. using AI has boosted my knowledge, the more I use it the more i learn as it teaches me it explains everything and answers all of my questions no matter how much I keep asking... the next generation will be so freaking smart "it aint even funny"😉
Lots to unpack for sure, but my take is that we’ll see something “Sora-Like” before we actually see Sora. And someone will figure out how to optimize the GPUs to be more efficient, and cost less in compute power eventually.
Of course, even when we hit the most amazing video generator of all time, they’ll also be teasing the next version which looks even better! Haha! It never ends!
Sora - in VR - LIVE..evolving...at 120 fps at 4K resolution with multiplayer will be fun '')
Haha. It’ll be Ready Player One, and I’m here for it!
Imagine ai imagines how a gravity device should look like
That interview was funny but the comments section of the WSJ upload are hilarious.
Oh, I’ll have to swing over there! There is sure to be some comedy gold! I’ll save it for weekend reading!
Its been a year since the release of ai image creation and still they don't have a polished version of this
You can easily identify ai images with some minute mistakes
I agree with you, whenever sora gonna release,it isn't gonna be what people are expecting it to be
"Below 1080p videos of max 15sec with major flaws and infinite generation time""
These are my expectations from it
(Sorry for my English,im not a native speaker)
No need to apologize! And I 100% agree! Release Sora is NOT going to be what has been advertised. We're still a little ways out from that-- it will happen, but not this year.
Cool 🙃
Even as a project lead it doesnt mean you are in the know when it comes to release dates. There are people above you making that decision.
Yeah. That’s why I have that bit about being in a meeting when your boss makes some grand announcement, and you and the team are like: “Wait. Wut?”
Has happened to me many times! And by your comment: I believe it has happened to you as well!
I think AISphere is the company that owns PixVerse, no?
y'know, I was wondering that as well. One article referred to them as two separate entities though. It's getting really confusing with all these new players on the scene!
Next, they need to reveal the officila release date for Chat GPT-4 Turbo.
Haha, likely the same time that GPT5 drops! Sam does love a little old school “and one more thing” showmanship.
To be honest, I appreciate it.
How much do you think it will cost? :) Some good AI monthly fees are too expensive
My guess, and this is just a guess, is there will be a limited free version, and a paid version.
Man, that paid version could be anywhere from $20 to $90 a month. That’s totally speculation on my part though
20 dollars is a good fee, I heard from someone, I didn't believe it would be 1000 dollars :)@@TheoreticallyMedia
wait wait wait we won't be able to make 1080p? but somehow this is supposed to put VFX artists out of business. news flash: we work in 4k and can actually do revisions.
Haha. Yeah, I don’t think anyone at ILM is currently concerned. I’ll say, I think there’s a lot of interesting opportunities to create assets with AI for VFX on the pro level, but whole hog? Not for quite a bit.
just not yet
In two years, VFX artists are going to find themselves where several successful Hollywood concept/storyboard artists I know are currently-their jobs have completely dried up in the last few months. It’s scary.
The real useable tools coming out of this movement are the most interesting part indeed! So much advancement in streamlining and integrating photogrammetry for example!@@TheoreticallyMedia
Just you wait , your days are numbered .
🎯 Key Takeaways for quick navigation:
00:29 *🤖 Overview of Sora Details*
- Detailed breakdown of recent Sora developments and interviews.
- New Sora videos showcased, including animations and generated content.
- Discussion on Sora's training data, technical capabilities, and potential release timeframe.
04:13 *🛡️ Doubts on Sora's Release Timeframe*
- Discrepancies between interviews regarding Sora's release timeframe.
- Speculation on technical and safety hurdles before Sora's launch.
- Mention of an arms race in AI funding and competition.
05:36 *💼 Insights into Sora's Development*
- Insights into Sora's development process and potential future versions.
- Discussion on Sora's training data and its relation to the GPT family.
- Emphasis on the interest in artistic use of Sora rather than predefined prompts.
07:58 *🎭 The Simulation's Projects Beyond Sora*
- Overview of projects by The Simulation/Fable Studios beyond Sora.
- Introduction to Thistle Gulch, an AI-driven Western narrative project.
- Details on Saga, an open-source tool for interactive storytelling.
Made with HARPA AI
Thanks! I need to reinstall harpa. It was doing some weird things awhile back, but I think that might have been a chrome issue on my end
@@TheoreticallyMedia Nicely done! Ha!
OpenAI keeps surprising us. In fact, I think the entire AI world keeps shocking us. I'm not saying it's coming soon. But if they release it next week. Boom💥 there you go.
Oh for sure. And I am constantly eating my words. I’ll be happy to if next week OAI tweets out: “Sora is now available, except for you, Tim”- but I just don’t see it in this case.
Personally, I think we’ll see Sora Level coming from a competitor before we see it from OpenAI
Agreed, a competitor popping up could definitely happen.
6:54
That's been a lot of my effort with all image AI's........to guide them to create unusual things that we don't see in the real world
...... kinda like that awesome cheese pirate ship of yours!
I routinely prompt for AI's to create realistic looking things that just wouldn't or couldn't be found on earth. One of the trickiest has been to get a realistic looking Great Wall of China gradually morphing into a giant serpent-like dragon.
With Dall-E 3 and only just VERY recently with MJ have I gotten decent, though sporadic, results. It's apparently a serious challenge still for AI's to get consistently good results with such a concept and I've tried many different prompts.
720p isn’t SD. I’m expect someone commenting on video to know this basic fact! 🤷🏼♂️
SD or Standard Definition is the predecessor of HD or High Definition video and is characterized by the smaller 720 x 480 interlaced screen size 4:3 ratio and 480 lines of vertical resolution (720 x 576 PAL).
Ok, found another source and it looks like there’s an argument about it these days: SD stands for “standard definition” and includes resolutions at or below 480p.
HD stands for “high definition” and includes resolutions at or above 720p.
Still, I call 720 SD these days.
Youre welcome to call 720p whatever you like, but as you’ve just read you’d be mistaken to call it SD!
@@neilmarshall4235 Look at the brains on Brad
If it is so expensive to replace sora at its full capacity, why not just have a enterprise level tier? If it saves a film company a shoot it could be worth hundreds or conceivably thousands of dollars a second to them.
There has been some speculation on that. The problem is, when you hit that level of pro use, the quality of Sora isn’t to those standards. Like, it’s too uncontrollable, and not at 4/8/16k resolution.
It’ll eventually happen, but we’re in a place of growing pains right now. I do think OAI will charge a pretty good amount for higher tier of Sora access when it launches. You’ll have a “lite” free version, and a paid higher tier that will produce better/longer results.
Just a theory though!
Young man. I’ll have you know that if you are in 1958 with a super8 camera, you are shooting a format that wasn’t introduced until the next decade.
I was there.
Haha, been awhile since I’ve been called young man! Thank you! Also, I probably should have looked that up beforehand!
Go Tim
Thank you! I felt like a real grown up reporter for this one!
720p is HD, 1080p is full HD.
Yeah, I messed that up. Apparently a few specs sheets recently are starting to call 720 SD-- but, coming from the "old" days of DVD, I should have known better!
mákina!!!
The MACHINE!!
I want to know the release date 😍
No inside info, but I don’t think we’ll see it until after the US election. Mostly because OpenAi doesn’t want the bad press
@@TheoreticallyMedia waiting 🥲
Look at the flappy towel foot under the guys butt😂 7:19
Haha, I love that video! There's just so much weirdness everywhere! I want it as a screensaver!
These examples were considerably less impressive that almost all the other videos they have shown, I wonder if they are stepping it back a little to allow it to be released, surely pushing videos so early to drill up hype to then take another year + doesn't make business sense?
That’s what I’ve been saying all along. The cherry picking is pretty evident at this point.
Not saying it isn’t impressive, but it clearly also has the same fail rate as all the other text to video generators.
I think they’re starting to set expectations at this point.
I'm not convinced this method of training will ever be a true viable option. I think it will be amazing for inpainting or changing flaws, lighting etc on actual footage. But I reckon we will need to rely on AGI to find the optimal way of understanding things like time, coherence and physics. @@TheoreticallyMedia
Watch Sora dissapoint the astro-high expectations. Create with tools we have now. Being ahead of the curve, is a badge of honor.
That’s the take right there. Some of the Sora expectations I’ve encountered are just set wayyyy to high. This is not going to be “how movies are made from now on” (a legit comment I was told.)
I dont think Sora will be able to keep up with consumer demand. It requires tremendous computing power to generate a video with the quality of Soras video previews. OpenAI even nerfed GPT4 because it could not keep up with consumer demand, and thats just a text model. A video model is 1000x times more compute hungry. I think the Sora released to the public with be nerfed so bad that it will be only slightly better than the current runway and Pika.
Agreed. I think that’s what Mira was implying here. We’ll get 4 seconds of 720p video. It’s really the only way they can possibly serve a user base as large as it is going to be.
I also do NOT think we’re looking at “months”
@@TheoreticallyMedia Of course Hollywood and enterprise/corporate customers may get a backroom deal for the fully unlocked Sora. But us plebs, not so much. So much for "this will kill Hollywood".
There's a lot of coping and pretending that "this is as good as it gets" in this video. You're crazy if you think it ends here. Give it 2 years and come back to this video and see if you still feel the same way.
Oh, for me it’s the opposite, if anything I’m acutely aware of how good everything is going to get. I was just being a little extra critical in this video because of the massive amount of hype and expectation built up around Sora.
I was hearing far too many people saying they wanted to wait to create until Sora was released.
That, in my opinion, is a big mistake.
5:40 three people at table do not look as real as Sora
Ha! Yeah, they really don’t look like they’re comfortable being on camera, do they?
There is no release date. .
Yup. The video is more about the semantics of the release day-- and the clear disconnect between the boardroom and the tech leads.
Personally, I don't think we're getting a full fledged Sora anytime soon. Maybe a nerf'd version in a few months, at best.
It's crazy how hyped everyone is to have an AI slave that will do their bidding. Slavery 2.0 is starting out quickly.
Did you see the robot at the end of the video? Yeah- I do see a point where that’ll be a common household appliance.
Just…remember to be nice to them! We know how it ends when we’re mean!
This information is truly shocking! Lol.
Really tho, it very well could get more affordable. Intel just announced a new chip that's designed to for AI and is apparently really fast. Plus, nvidia has released fairly new tech to double the generation times. Additionally, i reckon if they can all manage to implement the 0shot accuracy claude is showing, we'll see some massive improvements.
I hear as it is right now a 4090 with the right software, can generate real time images at 60fps.
Imagine a generative "portal" into another dimension, on the next generation of phones or ar headsets.
It's real life still but it's not. Lol.
You can have it cell shaded.
Black and white sketched.
Take your pick. Future's going to be crazy.
Thanks for all the info about the other video projects. Good to know. 🤟
Shocking! You won’t believe the 10 ways Sora can make you $6000 a month! Haha
And yeah, chips are only going to get faster and cooler, and (hopefully) more energy efficient.
It’s funny, that’s almost always been the issue. We’ve more or less beaten “size” as a chip problem- I mean, go back 40 years and think of the size of a microchip. And then we started to fight heat with water cooling. Power with improvements in conductivity- and software made these huge leaps. But even as code/software makes these huge leaps, it’s still physical world problems that are a sticky point!
Yes, there is still a lot of ugliness, unnatural, amusing or repulsive video footage. But AI is developing very quickly, and we live in the same amazing time as our ancestors did 120 years ago, at the dawn of aviation.
Agreed! I can fall into the nitpick area fairly easily-- but when I find myself dwelling there, I pinch to remind myself that this is all REALLY AMAZING.
Like, 10 years ago? We never would believe we'd be here!
haiper = sora
This translated is hilarious 😂
Agreed! And we can actually USE Haiper!
It will be behind not so open AI paywall so meh.
Haha, highly likely!
Murati is one of the least competent CTOs I've seen in my life. She barely has any knowledge about anything. How the heck did she become a CTO without having the experience or education in the field? She's not even good at communication. All she does is spewing hot air all the time.
Sora is just another marketing catch. The amount of computing power along with the insane cost will be the two main road blocks for it to become a successful product.
Just like Devin. Everyone's talking about it, but a prompt that runs for 15 minutes burning a few million tokens will cost a fortune and that thing has a success rate of 13%...
I low key agree. I'd be happy to eat my words on Sora's release, but after that initial shock and awe, I'm seeing a lot of emperor's clothes here.
What kills me is seeing people in the comments saying: "I'll start making stuff when Sora releases..."
I think they'll be sorely bummed when it finally drops. If that even happens.
People are legit not understanding what their doing. With these AI, people will start losing jobs and won’t be able to pay for things. And that’s a problem because if we don’t have jobs and can’t have money. That will break our whole system since we will have to make new rules and not have to pay for anything. That is able to cause a lot of chaos. Even if there should be a law for you not to take the whole store for free, it’ll still cause everything to go out of stock with the rapid pace of everyone taking things from stores instead of working hard for it. But the small good side of this is that it’s only art and film making. But for me that is not good at all since I want to become a manga artist. And there are people that want to make films. These people could lose their job or already lost them. And another bad part about this is that people could take advantage of this AI and possibly frame others for crimes. Which can lead to people getting arrested and people can also get hurt if others believe in the fake video footage and attack the person for something they DIDNT do. The creators of this new AI should start thinking about the consequences of creating a super realistic AI. And it’s pretty simple for them to understand all of these reasons and consequences since they are smart enough to understand how to make an AI that is this advanced.
Uncensored
For Sora? Unlikely there
Ooof, "Publicly Available" does not mean it's not copywritten or owned by someone else, it's a good thing OpenAI is generating money b/c they're going to be paying through the nose for lawsuits. I don't think this is going to go the way they expect, ChatGPT & Sora have the "Walmart Problem". So, with Millenials, Zoomers, Alphas, outside of the tech-bro sphere, generated content like Sora is viewed as "Cheap" and like "Oh, you're broke? You couldn't afford to make a proper thing?" Then there's also the current state of entertainment, most people are bored, everything seems derivative, predictable, paint by numbers, b/c creatives aren't driving the work, executives with algorithms are. At first glance I was worried about Sora's capabilities, but it's both imperfect & too perfect in the wrong ways, it'll never substitute the real thing/artists. Even now, only weeks in, Sora stuff seems stale, predictive, a fleeting novelty. Like, I don't think they understand the market & what the majority of humans want, they want things/products produced with care by other humans, entertainment dreamt up by visionaries, AI & Robots while impressive, will fall short.
These sora video’s seem to appear worse and worse… is it just me?
It is not. I think they’ve been slowly setting expectations. That initial shock and awe was pretty cherry picked, I think.
Odd, that's not personally my thought
Same as chatgpts are slowly downgrading
Corse they tweeking it “for safety” and killing iverall creativity and quality in the process. 1st SORA show off was fake. Just a flash for investors. Wi will get this quality not sooner than 2026
@@TheoreticallyMediaor more like the software has higher intensity capabilities that they won't be releasing for use initially or at all.
could y'all also do a report on the lawsuit Elon Musk made 2 week ago, against Microsoft (who had all the OpenAi board members fired and replace them with hand pick microsoft staff, with no Ai expertise), Sam Altman ( evil scientist, manipulate Musk and stole OpenAi from him), gregory brockman, helen toner(advistor of GovAi), adam d'angelo (ceo Quora) OpenAI,inc OpenAi Q* ??
Haha. I’ll try to catch the next volley of that fight! At this point, sadly in YT terms, it’s “old news” but I’ve following it like it’s a sports event! So, next “round” (and there will be another round for sure) I’ll pop a report up!
I’d say…something is bound to happen like, within a few weeks?
But who knows, this being as crazy as it is, it could be tomorrow!
Mira could not even answer the basic questions...like where was your data taken for images and video references....essentially SORA scrapped and stole everything from the web.
I think OAI is being a little more careful these days about training data, but I don’t know 100%. I think she was dodging here because literally, anything she would say would eventually be used against them.
I’m actually surprised the question wasn’t black listed for the interview. It’s possible it was and the interviewer asked anyhow, which if so: props to them!
It's strange, I haven't heard any discussions about how much Sora's annual subscription will cost... I believe they will be very high prices.
Does that mean we will get better porn.
porn is bad. There is no 'better' porn.
It's disturbing that the CTO would pretend to not know whether their flagship product was trained on CZcams videos or not.
oh, I think she 100% knows. She's just not saying...to be honest, any answer she provides there will be eventually skewed in a lawsuit. Best bet is to plead the 5th!
Just saying... Her body language when asked those questions tell me she would be terrible at poker.
On a separate note, ya know when you watch a super heavy CGI movie, and you kinda disconnect from it because you know it isn't real? I guess we need to prepare to feel that way with every fricken video on the internet soon. All prank videos, nature videos, history videos, movies, politics videos, you name it, all belief will be cast aside. Prepare for the death of youtube and a return of going to visit friends and family to get a slice of reality. I know most people won't believe this, but when the other 6.5 million people on the planet can suddenly make cinema quality videos, shit's gonna get spooky real fast.
Haha. Yeah, we need to invite Mira over for poker night!
On the larger note: yeah, we’re going to get flooded for sure, but the Algo of CZcams (and others) are really, really, good at filtering out junk.
So, if it’s low effort, low quality videos- they’ll just get buried and no one will ever see them.
Kind of like how you probably don’t see those Minecraft videos made by a 12 year old. You know the ones with like, 4 views.
Stuff that will rise will be interesting shorts or maybe history videos, with good information/stories, assisted by AI. Like, if I were a channel doing Roman history, I’d be all over making footage of famous battles for my videos!
seqs
Mira’s zero F given approach is so refreshing
Don’t get me wrong, I appreciate it. That was a solid stonewall! Just “NOPE you aren’t baiting me!”
By that you mean doing exactly what all pretentious rich people do?
That bull was rubbish, The nose ring was in its mouth
Yeah, it doesn't track well in 3d space either. At one point, the ring appears in front of a china dish. I mean, that said-- it is still pretty crazy that this exists. 4 years ago this would have been a fever dream. So-- while I complain-- I still have to remind myself that its really impressive.