Paul Morgan’s Post

🌌 Unveiling the Future: OpenAI's #Sora Revolutionizes #AI with Unmatched Physics & Emergent Abilities! 🚀 Dive into the future with OpenAI's groundbreaking Sora model! 🌠 Wes Roth takes us on an exhilarating journey, revealing how Sora is not just a leap but a quantum jump in AI video generation. Imagine a world where AI-created videos are indistinguishable from reality. That's Sora for you! 🎥✨ Sora stands out by simulating 3D environments with physics so real, you'll have to do a double-take. From the delicate balance of ice cubes in a glass to the dynamic battles between pirate ships in a cup of coffee, Sora's capabilities are nothing short of magical. 🏴☠️☕️ But here's the kicker: what we see is just the tip of the iceberg. Behind the scenes, Sora is a powerhouse, blending data-driven physics engines with intuitive physics, long-horizon reasoning, and semantic grounding. It's not just creating videos; it's simulating entire worlds, real or fantastical. 🌍🔮 Dr. Jim Fan from #Nvidia hints at the complexity and potential of Sora, suggesting it might even leverage Unreal Engine 5 for its breathtaking simulations. This isn't just about video generation; it's about building the foundation for #AGI through intricate rendering and intuitive physics. 🧠💡 Join Wes Roth as he unpacks the marvel that is Sora, offering a glimpse into the future of AI that OpenAI is shaping. This isn't just an advancement; it's a revolution. And we're here for it. 🌟🛠️ 👍 30 min.⏱here: https://lnkd.in/eq3yCcgr #OpenAI #SoraRevolution #AIFuture #PhysicsSimulation #AGIPotential

Life Will Never Be the Same After OpenAI's Sora OpenAI's Sora, introduced on February 15, 2024, is revolutionizing the way we interact with artificial intelligence. Unlike its predecessors, Sora can transform text prompts into one-minute videos, unlocking new possibilities for creative expression. Sora is not just another AI model; it's a gateway to a new realm of imagination. By seamlessly translating text into dynamic video content, Sora empowers users to bring their ideas to life in ways previously unimaginable. Behind Sora's magic lies a blend of advanced technologies. From the foundational Transformer model, enabling nuanced understanding of language, to Vision Transformer (ViT) and Video Vision Transformer (ViViT), capturing visual context, each component contributes to Sora's ability to create compelling videos. OpenAI's Sora is not merely a tool; it's a catalyst for creativity. By bridging the gap between text and video, Sora opens doors to endless possibilities in storytelling, education, marketing, and beyond. As we embrace this new era of AI-driven creativity, Sora stands as a testament to the power of innovation in shaping our digital landscape. #soraopenai #sora #ai #generativeai #aiautomation #chatgpt #openai

#GenAI Video Accelerates in 2024 via Sora by OpenAI... Creating instant video from text... what are your views? This is all synthetic. Sora's a data-driven physics engine, a simulation of many worlds, real or fantastical. The simulator learns intricate rendering, "intuitive" physics, long-horizon reasoning + semantic grounding, by some denoising and gradient maths. This is no toy! #AI #DataDriven #PhysicsEngine #Simulation

#gameengine #CFD #machinelearning #AI computational fluid dynamics a few years back, I first learned to create fluid animation in game was from NVidia's example using Navier Stoke equation, I didn't fully understand then. this time, hopefully, I can get the full picture :) and in a short amount time, hopefully, within 1 week :) PDE is so fun, see, I just solved one of the simplest, the linear convection both numerically and analytically. to someone who doesn't know anything about PDE, imagine this, a relation between some changes in space and some changes in time :) -- space-time P.S. I uploaded a mp4 version of size 26.4k, LinkedIn complained and asked me to upload files > minFileSize :) P.S. I chose the dt smaller that dx, so, there is computational diffusion effect, which is undesired, but, intentional here to show difference.

Experimenting the creation of living AI / avatar while waiting for GPT4o voices 🙂 The stack: Midjourney (Generate Face) -> XTTS(voice) -> dream talk (Animate Face) -> VideoUpscaling -> NVidia EyeGaze (Eye contact) -> Facemesh 3D (Spark Studio) AI conversation agent: Pi ai

Mixing Houdini's proceduralism to generate synthetic data and train models with the power of running on Conductor on CoreWeave, kickstarts what VFX artists can do with AI/ML. DM me if you want to chat more about this. #sidefx #houdini #ai #ml #conductor #coreweave #vfx

So... cinematic, real time, photorealistic video has arrived sooner than we think. LLMs now have understanding of the real world and physics on a "matrix" level. With things moving foward so quickly, are we near the superintelligence? And if AI can do everything, I mean everything, better than we do, and faster, what will happen to jobs and industries as a whole? for example what will happen to the movie and gaming industries when everyone can generate and customize their own experiences? this is happening way faster than I expected.

As a sucker for well-done explanation videos, I can't recommend this one enough. Very well written, well paced, with on-point animations and probably the best length-to-depth ratio of dozens if not hundreds videos I've seen on the topic so far. If I only had 15 minutes to understand how LLMs and neural nets work I would watch this one.

Add some magic! #MagicTime is a time-lapse video generation model that acts as a metamorphic simulator by learning physical knowledge from real-world time-lapse videos. It introduces several innovations: 🧙♂️ MagicAdapter scheme to decouple spatial and temporal training, allowing pre-trained text-to-video models to generate metamorphic time-lapse videos. 🧙♂️ Dynamic Frames Extraction strategy to adapt to the wide variation range and dramatic object changes in time-lapse videos. 🧙♂️ Magic Text-Encoder to improve understanding of prompts describing metamorphic processes. 🧙♂️ ChronoMagic dataset of curated time-lapse videos and descriptions to facilitate metamorphic video generation. 👉 Be sure to leave encouraging comments to the Magic Time team Shenghai Yuan, Jinfa Huang, Yujun Shi, Yongqi Xu, Ruijie Zhu, Bin Lin, 程鑫华, Li Yuan, Jiebo Luo 🐾 Our comment The physical knowledge encoded within the MagicTime model could prove invaluable in simulating complex systems and interactions within game worlds. This could lead to more accurate and believable physics simulations, resulting in a heightened sense of realism and engagement for players. 🔔 Afraid of missing out on our news? Worry not! Place the emoji 😀 in a comment below, and we'll make sure to tag you once we publish anything new. 👑 Here are our VIP alert subscribers: Vitalii Bortnik, Martyn Redstone, Aliaksei Nastsin, Sergei Sherman, Shapath Das, Guy Sebbag, Eugene Gladovskiy,Illia Shestakov, Albert Nicola , Samuel Wang, Cohi Elchadef Manobela, Adi Beker, Junwoo Yun, Maryanne Collins, Kirill Shamanskiy, Vakhtang Matskeplishvili, Odem Alagem אודם 🎡 , Shahar Butz, Marianna Inozemtseva, Asya Polyak, Guy Galin, Alisa Yurchenko, Adrian Magearu.,Elad Cohen, Chen Offek, Evgenii Raisfeld, Kamil Sawko, Michal Lissak, Eze Vidra, Yash Baid, Cosmin Tanase, Shashikiran R, Erez Rachmil, Meir Peleg, Raz Rosh, Almog Ben Simon #tech #video #generation #AI #innovation #mobilegamedevelopment

"So, are video generation models like Sora world simulators? Maybe in a weak sense, but not necessarily in the way one might think. Their generative process is not conditioned on forward-time simulations of intuitive physics, as one could do with an intuitive physics engine; but it might be shaped by latent representations of key aspects of 3D geometry and dynamics. In that latter (and weaker) sense, it is rather plausible that Sora has a limited “world model”, in the same way as latent diffusion models used for image generation have a – more limited still – “world model”. But we don’t know for sure, not until some research group probes Sora in the right way. OpenAI still occasionally delves into interpretability research, so there is hope; although I would be happy to see interpretability efforts from the broader research community with open weights video generation models like Stable Video. While far less capable than Sora, these models have the merit of being research friendly."