Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

Humans can see in high-res, high-FPS in real-time. Why can't VLMs? Introducing AutoGaze: ViTs/VLMs "gaze" only at key video regions! Up to 4-100x token savings, 19x speedup, and enables scaling to 4K-res 1K-frame videos. 📄 🌐 🤗 (1/n)🧵

baifeng_shi's profile picture

Baifeng

2,336 subscribers

162,275 Aufrufe • vor 5 Monaten •via X (Twitter)

Wissenschaft & Technologie
LIVE

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

Introducing The Matrix --- a foundation world model for generating infinite-length, hyper-realistic videos with real-time, frame-level control: - Infinite-length video generation - 720p high-quality rendering - Real-time, frame-level control at 16 FPS - Generalization to real-world video control 🔗Blog: 📄Paper: 💻Code & Playable Demo: Coming soon! Key Innovation: A brand new technique called the shift-window denoise process model, enabling auto-regressive generation for diffusion and consistency models in real-time. Special thanks to project leader Ruili Feng and the entire Matrix team for their dedication and hard work over the year-long project.
1:00

Introducing The Matrix --- a foundation world model for generating infinite-length, hyper-realistic videos with real-time, frame-level control: - Infinite-length video generation - 720p high-quality rendering - Real-time, frame-level control at 16 FPS - Generalization to real-world video control 🔗Blog: 📄Paper: 💻Code & Playable Demo: Coming soon! Key Innovation: A brand new technique called the shift-window denoise process model, enabling auto-regressive generation for diffusion and consistency models in real-time. Special thanks to project leader Ruili Feng and the entire Matrix team for their dedication and hard work over the year-long project.

Hongyang Zhang

178,403 Aufrufe • vor 1 Jahr

Introducing VL-JEPA: Vision-Language Joint Embedding Predictive Architecture for streaming, live action recognition, retrieval, VQA, and classification tasks with better performance and higher efficiency than large VLMs. • VL-JEPA is the first non-generative model that can perform general-domain vision-language tasks in real-time, built on a joint embedding predictive architecture. • We demonstrate in controlled experiments that VL-JEPA, trained with latent space embedding prediction, outperforms VLMs that rely on data space token prediction. • We show that VL-JEPA delivers significant efficiency gains over VLMs for online video streaming applications, thanks to its non-autoregressive design and native support for selective decoding. • We highlight that our VL-JEPA model, with an unified model architecture, can effectively handle a wide range of classification, retrieval, and VQA tasks at the same time. by Delong Chen (陈德龙) Mustafa Shukor Théo Moutakanni Willy Jade Lei Yu Tejaswi Kasarla Allen Bolourchi Yann LeCun Pascale Fung
1:51

Introducing VL-JEPA: Vision-Language Joint Embedding Predictive Architecture for streaming, live action recognition, retrieval, VQA, and classification tasks with better performance and higher efficiency than large VLMs. • VL-JEPA is the first non-generative model that can perform general-domain vision-language tasks in real-time, built on a joint embedding predictive architecture. • We demonstrate in controlled experiments that VL-JEPA, trained with latent space embedding prediction, outperforms VLMs that rely on data space token prediction. • We show that VL-JEPA delivers significant efficiency gains over VLMs for online video streaming applications, thanks to its non-autoregressive design and native support for selective decoding. • We highlight that our VL-JEPA model, with an unified model architecture, can effectively handle a wide range of classification, retrieval, and VQA tasks at the same time. by Delong Chen (陈德龙) Mustafa Shukor Théo Moutakanni Willy Jade Lei Yu Tejaswi Kasarla Allen Bolourchi Yann LeCun Pascale Fung

Pascale Fung

90,144 Aufrufe • vor 8 Monaten

In my previous post about the #Nioh3 demo, I mentioned that when the game isn’t consistently holding 60 FPS or 120 FPS, several issues can show up. Here’s a clearer breakdown of what's going on. The main reason is that the game speed and physics are tied to the frame rate. When you enable the 120 FPS cap, the engine start overreacting and pushes CPU usage very high, even if the game is still running at around 60 FPS (as shown in the first example). In the second clip, you can see the uneven, jittery camera movement I mentioned before. Notice how much smoother 60 FPS and 120 FPS look compared to the unlocked side. Also, camera movement at 60 FPS is slightly faster than at 120 FPS. In the third clip, player movement is slower when the frame rate is unlocked. This doesn’t happen all the time, but it shows up often enough to be noticeable. Because of how the Katana Engine behaves, the game is clearly designed around 60 FPS. Running at 120 FPS is possible, but it’s only recommended if your system can maintain that target almost all the time, which isn’t easy to achieve. There’s also an alternative workaround where you select the 60 (locked) option and enable Frame Generation (DLSS or FSR 3), as shown in the last clip. The downside is that DLSS Frame Generation tends to show the same stuttery look as when the frame rate isn’t holding a fixed target, likely due to Reflex keeping the frame rate slightly below target. FSR Frame Generation, on the other hand, looks much smoother and works better here.
1:17

In my previous post about the #Nioh3 demo, I mentioned that when the game isn’t consistently holding 60 FPS or 120 FPS, several issues can show up. Here’s a clearer breakdown of what's going on. The main reason is that the game speed and physics are tied to the frame rate. When you enable the 120 FPS cap, the engine start overreacting and pushes CPU usage very high, even if the game is still running at around 60 FPS (as shown in the first example). In the second clip, you can see the uneven, jittery camera movement I mentioned before. Notice how much smoother 60 FPS and 120 FPS look compared to the unlocked side. Also, camera movement at 60 FPS is slightly faster than at 120 FPS. In the third clip, player movement is slower when the frame rate is unlocked. This doesn’t happen all the time, but it shows up often enough to be noticeable. Because of how the Katana Engine behaves, the game is clearly designed around 60 FPS. Running at 120 FPS is possible, but it’s only recommended if your system can maintain that target almost all the time, which isn’t easy to achieve. There’s also an alternative workaround where you select the 60 (locked) option and enable Frame Generation (DLSS or FSR 3), as shown in the last clip. The downside is that DLSS Frame Generation tends to show the same stuttery look as when the frame rate isn’t holding a fixed target, likely due to Reflex keeping the frame rate slightly below target. FSR Frame Generation, on the other hand, looks much smoother and works better here.

BenchmarKing

17,820 Aufrufe • vor 7 Monaten

So I heard we need more data for robot learning :) Purely real world teleop is expensive and slow, making large scale data collection challenging. I’ve been excited about getting more data into robot learning, going beyond just real-world teleop data. To this end, we’ve been scaling up data generation with RL in realistic simulations generated on the fly from crowdsourced videos. Enables realistic data collection, much more cheaply than purely real world teleop. Importantly, data collection becomes even*cheaper* with more environments, allowing training with over 100x more data. Transfers to real robots for generalizable manipulation. A 🧵 (1/N)
0:41

So I heard we need more data for robot learning :) Purely real world teleop is expensive and slow, making large scale data collection challenging. I’ve been excited about getting more data into robot learning, going beyond just real-world teleop data. To this end, we’ve been scaling up data generation with RL in realistic simulations generated on the fly from crowdsourced videos. Enables realistic data collection, much more cheaply than purely real world teleop. Importantly, data collection becomes even*cheaper* with more environments, allowing training with over 100x more data. Transfers to real robots for generalizable manipulation. A 🧵 (1/N)

Abhishek Gupta

13,358 Aufrufe • vor 1 Jahr

** SEGA GENESIS / MEGADRIVE Software Scaling + Rotation *** Sega Megadrive can't scale backgrounds and rotate sprites in software at the same time ?? - hold my beer !! After Scaling I've been tinkering with Rotating sprites in software and I think it looks pretty sweet ! In the video 4 * 32x32 sprites are rotated through 64 steps at 30 FPS in software, it could be 60 FPS and more rotating objects if we were not doing the scaling at the same time but even blast processing has its limits ;-) Rotation at 30 FPS is far smoother than you typcially see in even in pre-rendered rotations anyhow ( a 64 frame 32x32 rotation is 32kb - not small ). Challenges were make it fast : very fast as this needs to work at pixel level, not double pixel like scaling. The megadrives packed pixel format (2 pixels per byte) makes that a little bit tricky . When I started the algorithm was 3x slower than now. Make it sensible : has to be usable in game , had to present a rom size reduction over pre-rendered animation. Make it flexible : so we can rotate frames in the background using up idle cycles (end of frame) and store those frames to vram for very heavy cpu scenes, so we can still have very smooth rotation with very heavy cpu demands still. Thanks to my team Carsten666 for the GFX and CYBERDEOUS - Crouzet Laurent for the marvelous tune. #SegaGenesis #SegaMegadrive #SGDK
0:32

** SEGA GENESIS / MEGADRIVE Software Scaling + Rotation *** Sega Megadrive can't scale backgrounds and rotate sprites in software at the same time ?? - hold my beer !! After Scaling I've been tinkering with Rotating sprites in software and I think it looks pretty sweet ! In the video 4 * 32x32 sprites are rotated through 64 steps at 30 FPS in software, it could be 60 FPS and more rotating objects if we were not doing the scaling at the same time but even blast processing has its limits ;-) Rotation at 30 FPS is far smoother than you typcially see in even in pre-rendered rotations anyhow ( a 64 frame 32x32 rotation is 32kb - not small ). Challenges were make it fast : very fast as this needs to work at pixel level, not double pixel like scaling. The megadrives packed pixel format (2 pixels per byte) makes that a little bit tricky . When I started the algorithm was 3x slower than now. Make it sensible : has to be usable in game , had to present a rom size reduction over pre-rendered animation. Make it flexible : so we can rotate frames in the background using up idle cycles (end of frame) and store those frames to vram for very heavy cpu scenes, so we can still have very smooth rotation with very heavy cpu demands still. Thanks to my team Carsten666 for the GFX and CYBERDEOUS - Crouzet Laurent for the marvelous tune. #SegaGenesis #SegaMegadrive #SGDK

Shannon Birt

19,405 Aufrufe • vor 1 Jahr

1/ 🧠Humans are the best robot data source! 2/ 👓Human egocentric video is rich in quantity, but poor in quality. 3/ Beyond scaling data, smarter representation and architecture matter just as much. 4/ Want an open-source framework to train your own learn-from-human-data robot policy? 🚀We introduce HumanEgo: Zero-Shot Robot Learning from Minutes of Human Egocentric Videos⬇️ ✦ Zero-Shot Human-to-Robot Transfer ✦ Robot-Data-Free ✦ Just 30 min of data per task ✦ Collect by Anyone, Anytime, Anywhere ✦ Deploy on Any Robot, Any Camera, Any Environment ✦ Open-Source & Easy-to-Implement Let's squeeze every bit of signal out of human data! 🌐 Website: 📄 Paper: 💻 Code: 📹 Video: 🧵 1/n
1:28

1/ 🧠Humans are the best robot data source! 2/ 👓Human egocentric video is rich in quantity, but poor in quality. 3/ Beyond scaling data, smarter representation and architecture matter just as much. 4/ Want an open-source framework to train your own learn-from-human-data robot policy? 🚀We introduce HumanEgo: Zero-Shot Robot Learning from Minutes of Human Egocentric Videos⬇️ ✦ Zero-Shot Human-to-Robot Transfer ✦ Robot-Data-Free ✦ Just 30 min of data per task ✦ Collect by Anyone, Anytime, Anywhere ✦ Deploy on Any Robot, Any Camera, Any Environment ✦ Open-Source & Easy-to-Implement Let's squeeze every bit of signal out of human data! 🌐 Website: 📄 Paper: 💻 Code: 📹 Video: 🧵 1/n

Zhi (Leo) Wang

110,687 Aufrufe • vor 3 Monaten

NEW! URGENT REAL-TIME STOCK MARKET FORECAST You can't close your longs or sell your calls over the next couple days because every single day is going to gap up. By the time you can buy calls, the price is already too high to make any money. You got to buy the day before and hold through all the futures after hours volatility. In these three videos I explain why. Here's part one and two. Part 3 is in the next tweet.
1:08

NEW! URGENT REAL-TIME STOCK MARKET FORECAST You can't close your longs or sell your calls over the next couple days because every single day is going to gap up. By the time you can buy calls, the price is already too high to make any money. You got to buy the day before and hold through all the futures after hours volatility. In these three videos I explain why. Here's part one and two. Part 3 is in the next tweet.

Wall Street NYC Quant. bitcoin-fund-manager.com

133,496 Aufrufe • vor 5 Monaten

Introducing Einstein Sales & Service Agents on Salesforce’s Agentforce Platform. Natively integrated with Sales and Service Clouds, you can now scale your sales teams and elevate your customer support with seamless collaboration between humans and AI. For Sales: Automate lead nurturing & personalize engagements. For Service: Deliver high-quality personal service with up to 70% case deflection. Build and customize for any channel with real-time data, integrated APIs, and low-code tools—all in one place. Find out how humans with AI drive customer success at #DF24:
1:23

Introducing Einstein Sales & Service Agents on Salesforce’s Agentforce Platform. Natively integrated with Sales and Service Clouds, you can now scale your sales teams and elevate your customer support with seamless collaboration between humans and AI. For Sales: Automate lead nurturing & personalize engagements. For Service: Deliver high-quality personal service with up to 70% case deflection. Build and customize for any channel with real-time data, integrated APIs, and low-code tools—all in one place. Find out how humans with AI drive customer success at #DF24:

Salesforce

13,634 Aufrufe • vor 2 Jahren

Borderlands 4 Performance Summary in 2 minutes: 0:00 - Native Res Badass Settings the game runs bad 0:30 - Upscaling Performance Mode Badass Settings, the game barely manages to hit 60FPS 1:20 - Frame Gen turned on and now you have 100 FPS Some people will turn all these settings on and be happy with it but in my honest opinion these are used as crutches to make the game look as if it performs good. Borderlands 4 has a certain art style that doesn't look like it needs these kind of settings to be turned on to play at a high framerate but sadly it does and you can defend this if you want but the future of gaming looks bleak, using this tech to not optimize the games further will just result in input latency increase as every game will be delivered with frame-gen turned on in the devs mind.
1:53

Borderlands 4 Performance Summary in 2 minutes: 0:00 - Native Res Badass Settings the game runs bad 0:30 - Upscaling Performance Mode Badass Settings, the game barely manages to hit 60FPS 1:20 - Frame Gen turned on and now you have 100 FPS Some people will turn all these settings on and be happy with it but in my honest opinion these are used as crutches to make the game look as if it performs good. Borderlands 4 has a certain art style that doesn't look like it needs these kind of settings to be turned on to play at a high framerate but sadly it does and you can defend this if you want but the future of gaming looks bleak, using this tech to not optimize the games further will just result in input latency increase as every game will be delivered with frame-gen turned on in the devs mind.

NikTek

359,686 Aufrufe • vor 11 Monaten

🚀🚀🚀Introducing HY World 1.5 (WorldPlay)! We have now open-sourced the most systemized, comprehensive real-time world model framework in the industry. In HY World 1.5, we develop WorldPlay, a streaming video diffusion model that enables real-time, interactive world modeling with long-term geometric consistency, resolving the trade-off between speed and memory that limits current methods. You can generate and explore 3D worlds simply by inputting text or images. Walk, look around, and interact like you're playing a game. Highlights: 🔹Real-Time: Generates long-horizon streaming video at 24 FPS with superior consistency. 🔹Geometric Consistency: Achieved using a Reconstituted Context Memory mechanism to dynamically rebuild context from past frames to alleviate memory attenuation 🔹Robust Control: Uses a Dual Action Representation for robust response to user keyboard and mouse inputs. 🔹Versatile Applications: Supports both first-person and third-person perspectives, enabling applications like promptable events and infinite world extension. 👉🏻Try it now: 🌐Project Page: 🔗Github: 🤗Hugging Face: 📄Technical Report:
1:06

🚀🚀🚀Introducing HY World 1.5 (WorldPlay)! We have now open-sourced the most systemized, comprehensive real-time world model framework in the industry. In HY World 1.5, we develop WorldPlay, a streaming video diffusion model that enables real-time, interactive world modeling with long-term geometric consistency, resolving the trade-off between speed and memory that limits current methods. You can generate and explore 3D worlds simply by inputting text or images. Walk, look around, and interact like you're playing a game. Highlights: 🔹Real-Time: Generates long-horizon streaming video at 24 FPS with superior consistency. 🔹Geometric Consistency: Achieved using a Reconstituted Context Memory mechanism to dynamically rebuild context from past frames to alleviate memory attenuation 🔹Robust Control: Uses a Dual Action Representation for robust response to user keyboard and mouse inputs. 🔹Versatile Applications: Supports both first-person and third-person perspectives, enabling applications like promptable events and infinite world extension. 👉🏻Try it now: 🌐Project Page: 🔗Github: 🤗Hugging Face: 📄Technical Report:

Tencent Hy

189,802 Aufrufe • vor 8 Monaten

Midjourney Video is finally here, and it's very very good Initial observations: - Image to video conversion only for now - Four 24fps generations at 480p per job - Works with any aspect ratio image Two motion settings available: - Motion Low: More ambient movement - Motion High: More dramatic movement - Can upload custom images (including non-Midjourney generated) - Can extend video 4 times, up to 20 sec total - Extensions are very seamless - Cost efficiency: ~$4 for 20 videos (compared to $3 per single video in Veo) - Generation speed notably faster than competitors Two operation modes: - Auto: Runs with existing prompt - Manual: Allows prompt modification - Command key enables scrubbing feature - Resolution: Currently 480p Extensions follow additive approach: - Keep original prompt text - Add new instructions rather than replacing - Removing original prompt text reduced effectiveness
0:20

Midjourney Video is finally here, and it's very very good Initial observations: - Image to video conversion only for now - Four 24fps generations at 480p per job - Works with any aspect ratio image Two motion settings available: - Motion Low: More ambient movement - Motion High: More dramatic movement - Can upload custom images (including non-Midjourney generated) - Can extend video 4 times, up to 20 sec total - Extensions are very seamless - Cost efficiency: ~$4 for 20 videos (compared to $3 per single video in Veo) - Generation speed notably faster than competitors Two operation modes: - Auto: Runs with existing prompt - Manual: Allows prompt modification - Command key enables scrubbing feature - Resolution: Currently 480p Extensions follow additive approach: - Keep original prompt text - Add new instructions rather than replacing - Removing original prompt text reduced effectiveness

Nick St. Pierre

181,846 Aufrufe • vor 1 Jahr

Fast View Synthesis of Casual Videos paper page: Novel view synthesis from an in-the-wild video is difficult due to challenges like scene dynamics and lack of parallax. While existing methods have shown promising results with implicit neural radiance fields, they are slow to train and render. This paper revisits explicit video representations to synthesize high-quality novel views from a monocular video efficiently. We treat static and dynamic video content separately. Specifically, we build a global static scene model using an extended plane-based scene representation to synthesize temporally coherent novel video. Our plane-based scene representation is augmented with spherical harmonics and displacement maps to capture view-dependent effects and model non-planar complex surface geometry. We opt to represent the dynamic content as per-frame point clouds for efficiency. While such representations are inconsistency-prone, minor temporal inconsistencies are perceptually masked due to motion. We develop a method to quickly estimate such a hybrid video representation and render novel views in real time. Our experiments show that our method can render high-quality novel views from an in-the-wild video with comparable quality to state-of-the-art methods while being 100x faster in training and enabling real-time rendering.
0:24

Fast View Synthesis of Casual Videos paper page: Novel view synthesis from an in-the-wild video is difficult due to challenges like scene dynamics and lack of parallax. While existing methods have shown promising results with implicit neural radiance fields, they are slow to train and render. This paper revisits explicit video representations to synthesize high-quality novel views from a monocular video efficiently. We treat static and dynamic video content separately. Specifically, we build a global static scene model using an extended plane-based scene representation to synthesize temporally coherent novel video. Our plane-based scene representation is augmented with spherical harmonics and displacement maps to capture view-dependent effects and model non-planar complex surface geometry. We opt to represent the dynamic content as per-frame point clouds for efficiency. While such representations are inconsistency-prone, minor temporal inconsistencies are perceptually masked due to motion. We develop a method to quickly estimate such a hybrid video representation and render novel views in real time. Our experiments show that our method can render high-quality novel views from an in-the-wild video with comparable quality to state-of-the-art methods while being 100x faster in training and enabling real-time rendering.

AK

20,668 Aufrufe • vor 2 Jahren

Monster Hunter Wilds big performance patch is out on PC. I spoke with the user who originally discovered the DLC bug and they confirmed that it seems to be fixed. More improvements I noticed from some quick testing (continued): • Better 1% lows for smoother gameplay with less hitching and stutter • Lower VRAM usage (10+ GB -> 7.5GB) • Higher FPS in some cases • CPU optimizations • High res texture pack is actually usable now • Lowered the install size by 45GB with texture pack • Some users report that their new frame rate WITHOUT FG = old frame rate WITH FG • More graphics options to tweak Overall, this update seems to have done the trick. Still not 100% perfect but MUCH better than it was before. The recent Steam reviews have also been very positive so it seems that most agree. Another patch is coming in February with even more optimizations for both PC and console. Settings and PC specs at the end - this was the default recommended settings at 4K DLSS Performance, Frame Gen off, 16x AF (roughly the High preset). Finally, Capcom didn't credit anyone in the Patch Notes so shoutout to de_Tylmarande for originally bringing the DLC bug to people's attention. #MonsterHunter #MonsterHunterWilds
3:29

Monster Hunter Wilds big performance patch is out on PC. I spoke with the user who originally discovered the DLC bug and they confirmed that it seems to be fixed. More improvements I noticed from some quick testing (continued): • Better 1% lows for smoother gameplay with less hitching and stutter • Lower VRAM usage (10+ GB -> 7.5GB) • Higher FPS in some cases • CPU optimizations • High res texture pack is actually usable now • Lowered the install size by 45GB with texture pack • Some users report that their new frame rate WITHOUT FG = old frame rate WITH FG • More graphics options to tweak Overall, this update seems to have done the trick. Still not 100% perfect but MUCH better than it was before. The recent Steam reviews have also been very positive so it seems that most agree. Another patch is coming in February with even more optimizations for both PC and console. Settings and PC specs at the end - this was the default recommended settings at 4K DLSS Performance, Frame Gen off, 16x AF (roughly the High preset). Finally, Capcom didn't credit anyone in the Patch Notes so shoutout to de_Tylmarande for originally bringing the DLC bug to people's attention. #MonsterHunter #MonsterHunterWilds

KAMI

294,625 Aufrufe • vor 7 Monaten

Mira Murati says the next AI interface is not a chatbot. It is a model that can listen, see, interrupt, and respond in real time. "The models that we work with today, they're very turn-based." "You talk, they talk, then they go off and think." "While they're thinking, it's almost like they're deaf and blind. They cannot perceive anything else about what's going on." Her new frame: interaction models. "They're not turn-based. They're more like time-based interaction." "They're continuously taking in audio, text, video, and continuously providing output." "We cut this up in chunks of 20 milliseconds... this enables you to catch things like interruptions and simultaneous speech." That is the product shift: from AI as a prompt box to AI as a high-bandwidth collaborator.
1:33

Mira Murati says the next AI interface is not a chatbot. It is a model that can listen, see, interrupt, and respond in real time. "The models that we work with today, they're very turn-based." "You talk, they talk, then they go off and think." "While they're thinking, it's almost like they're deaf and blind. They cannot perceive anything else about what's going on." Her new frame: interaction models. "They're not turn-based. They're more like time-based interaction." "They're continuously taking in audio, text, video, and continuously providing output." "We cut this up in chunks of 20 milliseconds... this enables you to catch things like interruptions and simultaneous speech." That is the product shift: from AI as a prompt box to AI as a high-bandwidth collaborator.

Karl Mehta

191,763 Aufrufe • vor 2 Monaten

Most blockchains weren’t built for real world finance. We sat down with Daniel Marin, CEO of Nexus, to break down why how Nexus has built a purposeful Layer 1 for verifiable finance. We cover the Nexus Exchange, $USDX as the native dollar, predictable execution powered by zkProofs, and how stablecoin yield can create sustainable onchain revenue models. Timestamps: 0:00 – The Problem: Why Ethereum Isn’t Built for Traditional Finance 0:53 – Introducing Nexus: A Purpose-Built Layer 1 for Verifiable Finance 4:44 – How Zero Knowledge Scaling Enables Low Latency Trading 6:24 – Tokenization on Ethereum vs High Frequency Verifiable Finance 8:30 – Nexus Timeline: From Stanford to 3M+ Users 12:34 – USDX Explained: A Native Stablecoin Backed by T-Bills 16:16 – How Nexus Captures Stablecoin Yield & Creates a Value Flywheel 26:15 – What’s Coming Next: Token Campaigns & Mainnet Roadmap
29:04

Most blockchains weren’t built for real world finance. We sat down with Daniel Marin, CEO of Nexus, to break down why how Nexus has built a purposeful Layer 1 for verifiable finance. We cover the Nexus Exchange, $USDX as the native dollar, predictable execution powered by zkProofs, and how stablecoin yield can create sustainable onchain revenue models. Timestamps: 0:00 – The Problem: Why Ethereum Isn’t Built for Traditional Finance 0:53 – Introducing Nexus: A Purpose-Built Layer 1 for Verifiable Finance 4:44 – How Zero Knowledge Scaling Enables Low Latency Trading 6:24 – Tokenization on Ethereum vs High Frequency Verifiable Finance 8:30 – Nexus Timeline: From Stanford to 3M+ Users 12:34 – USDX Explained: A Native Stablecoin Backed by T-Bills 16:16 – How Nexus Captures Stablecoin Yield & Creates a Value Flywheel 26:15 – What’s Coming Next: Token Campaigns & Mainnet Roadmap

Crypto Coin Show

10,669 Aufrufe • vor 6 Monaten

ZoriaGen is officially live. an AI video generator powered by your image. Upload any image, describe the action, and AI will transform it into a dynamic video in seconds. When you sign up, you get a free opportunity to create one video in the Basic version so you can try the platform and see how it works. $ZORIA Premium Features Users with premium access unlock additional capabilities: - Create videos without a watermark - Access to advanced generation and premium platform features - Priority render queue - High-quality video up to 1080p - Video length up to 10 seconds $ZORIA Token Burning🔥 All ZORIA tokens used for video generation will be sent to a burn wallet and permanently removed from circulation. We will also add an interactive burn dashboard, where anyone will be able to see in real time: - the total number of tokens burned - the burn history - the impact on the circulating supply We keep building on BNB Chain Special thanks to the legend CZ 🔶 BNB for his legendary photos.😅
2:54

ZoriaGen is officially live. an AI video generator powered by your image. Upload any image, describe the action, and AI will transform it into a dynamic video in seconds. When you sign up, you get a free opportunity to create one video in the Basic version so you can try the platform and see how it works. $ZORIA Premium Features Users with premium access unlock additional capabilities: - Create videos without a watermark - Access to advanced generation and premium platform features - Priority render queue - High-quality video up to 1080p - Video length up to 10 seconds $ZORIA Token Burning🔥 All ZORIA tokens used for video generation will be sent to a burn wallet and permanently removed from circulation. We will also add an interactive burn dashboard, where anyone will be able to see in real time: - the total number of tokens burned - the burn history - the impact on the circulating supply We keep building on BNB Chain Special thanks to the legend CZ 🔶 BNB for his legendary photos.😅

Zoria

12,786 Aufrufe • vor 5 Monaten

Two of the people most responsible for scaling the transformer are now betting on a next act. Jerry Tworek ran the Reasoning 🍓 team at OpenAI. rohan anil was a pre-training lead on Gemini after years at Google Brain and Anthropic. They just started to find what comes next. Their core argument: (1) models are trained in the lab but deployed in the real world and can't keep learning once they leave; (2) AI research is done by humans today but models will be able to explore and uncover new advances more rapidly and systematically (controversial but timely w this week's petition). The conversation covers: — why Jerry expected AGI in 2025 and what changed his mind — the two kinds of learning from experience, and why RL only captures one — the computational depth problem baked into today's architectures — why the biggest labs can't afford to look for a transformer replacement — the kernel competition where humans + $100K of coding agents found a 60x speedup no frontier model comes close to — a definition of AGI you can actually test: a model that improves itself with no human in the loop 00:00 Introduction 01:46 Appreciating Transformers 02:44 Scaling Hits Limits 04:54 Why Architecture Matters 05:32 RL Reality Check 07:32 Test Time Learning 09:52 Economics Of Scaling 12:47 Why Start A Company 14:24 Rohan On Transformers 19:11 Computational Depth Problem 20:32 When Transformers Top Out 23:22 Beyond Reinforcement Learning 26:41 Optimization And Efficiency 34:24 Building An Automated Lab 39:45 Kernel Automation Roadmap
49:11

Two of the people most responsible for scaling the transformer are now betting on a next act. Jerry Tworek ran the Reasoning 🍓 team at OpenAI. rohan anil was a pre-training lead on Gemini after years at Google Brain and Anthropic. They just started to find what comes next. Their core argument: (1) models are trained in the lab but deployed in the real world and can't keep learning once they leave; (2) AI research is done by humans today but models will be able to explore and uncover new advances more rapidly and systematically (controversial but timely w this week's petition). The conversation covers: — why Jerry expected AGI in 2025 and what changed his mind — the two kinds of learning from experience, and why RL only captures one — the computational depth problem baked into today's architectures — why the biggest labs can't afford to look for a transformer replacement — the kernel competition where humans + $100K of coding agents found a 60x speedup no frontier model comes close to — a definition of AGI you can actually test: a model that improves itself with no human in the loop 00:00 Introduction 01:46 Appreciating Transformers 02:44 Scaling Hits Limits 04:54 Why Architecture Matters 05:32 RL Reality Check 07:32 Test Time Learning 09:52 Economics Of Scaling 12:47 Why Start A Company 14:24 Rohan On Transformers 19:11 Computational Depth Problem 20:32 When Transformers Top Out 23:22 Beyond Reinforcement Learning 26:41 Optimization And Efficiency 34:24 Building An Automated Lab 39:45 Kernel Automation Roadmap

Sonya Huang 🐥

232,559 Aufrufe • vor 1 Monat

We believe we’re the first robotics company to demonstrate a robot peeling an apple with dual dexterous human-like hands. This breakthrough closes a key gap in robotics, achieving bimanual, contact-rich manipulation and moving far beyond the limits of simple grippers. 🧵↓ Today’s AI models (VLMs) are excellent at perception but struggle with action. Controlling high-degree-of-freedom hands for tasks like this is incredibly complex, and precise finger-level teleoperation is nearly impossible for humans. Our first step was a shared-autonomy system: rather than controlling every finger, the operator triggers pre-learned skills like a “rotate apple or tennis ball” primitive via a keyboard press or pedal. This makes scalable data collection and RL training possible. How does the AI manage this? We created "MoDE-VLA" (Mixture of Dexterous Experts). It fuses vision, language, force, and touch data by using a team of specialist "experts," making control in high-dimensional spaces stable and effective. The combination of these two innovations allows for seamless, contact-rich manipulation. The human provides high-level guidance, and the robot executes the complex in-hand coordination required. This work paves the way for robots that can safely handle delicate tasks in human environments. Want the full technical details? 📄 Read the full research paper: Visit us at NVIDIA GTC Booth #1838, Hall 3 to learn more! #Robotics #AI #DexterousManipulation #VLA #NVIDIAGTC Nancy Villicaña NVIDIA GTC
2:51

We believe we’re the first robotics company to demonstrate a robot peeling an apple with dual dexterous human-like hands. This breakthrough closes a key gap in robotics, achieving bimanual, contact-rich manipulation and moving far beyond the limits of simple grippers. 🧵↓ Today’s AI models (VLMs) are excellent at perception but struggle with action. Controlling high-degree-of-freedom hands for tasks like this is incredibly complex, and precise finger-level teleoperation is nearly impossible for humans. Our first step was a shared-autonomy system: rather than controlling every finger, the operator triggers pre-learned skills like a “rotate apple or tennis ball” primitive via a keyboard press or pedal. This makes scalable data collection and RL training possible. How does the AI manage this? We created "MoDE-VLA" (Mixture of Dexterous Experts). It fuses vision, language, force, and touch data by using a team of specialist "experts," making control in high-dimensional spaces stable and effective. The combination of these two innovations allows for seamless, contact-rich manipulation. The human provides high-level guidance, and the robot executes the complex in-hand coordination required. This work paves the way for robots that can safely handle delicate tasks in human environments. Want the full technical details? 📄 Read the full research paper: Visit us at NVIDIA GTC Booth #1838, Hall 3 to learn more! #Robotics #AI #DexterousManipulation #VLA #NVIDIAGTC Nancy Villicaña NVIDIA GTC

Sharpa

20,429 Aufrufe • vor 5 Monaten