Загрузка видео...

Не удалось загрузить видео

На главную

The architecture of thought. GPT-2's neural pathways visualized in topological 3D space. Language generation as pure mathematical movement.

michieldoteth's profile picture

michielh.eth

5,105 subscribers

21,966 просмотров • 1 год назад •via X (Twitter)

Наука и технологииОбразование
LIVE

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

What if anatomy explorers felt alive? This 3D dog, including the skeleton, organs and rig, was generated with ai and reacts to the cursor in real time with head tracking and tail movement. - Used GPT Images 2 for consistency - Omma AI for 3D generation and code using Three.js Lmk if you’d like to try it!
0:14

What if anatomy explorers felt alive? This 3D dog, including the skeleton, organs and rig, was generated with ai and reacts to the cursor in real time with head tracking and tail movement. - Used GPT Images 2 for consistency - Omma AI for 3D generation and code using Three.js Lmk if you’d like to try it!

Gábor Pribék

204,613 просмотров • 2 месяцев назад

one 2d photo --> 3d gaussian splat quick test with Echo-2 by SpAItial -- these 3d scene generation models are getting better! already at a sufficient quality to serve as a virtual set / backdrop in your 3d tool of choice
0:17

one 2d photo --> 3d gaussian splat quick test with Echo-2 by SpAItial -- these 3d scene generation models are getting better! already at a sufficient quality to serve as a virtual set / backdrop in your 3d tool of choice

Bilawal Sidhu

23,511 просмотров • 2 месяцев назад

Introducing Kaleido💮 from AI at Meta — a universal generative neural rendering engine for photorealistic, unified object and scene view synthesis. Kaleido is built on a simple but powerful design philosophy: 3D perception is a form of visual common sense. Following this idea, we formulate rendering purely as a sequence-to-sequence generation problem, successfully unifying neural rendering with the architecture principles behind modern language and video models. Unlike traditional neural rendering methods, Kaleido learns 3D purely in a data-driven way, without explicit 3D representations or structures. It acquires spatial understanding directly through large-scale video pretraining, then multi-view 3D data finetuning, inspired by how LLMs acquire textual common sense from large corpora before specialising in domains like coding. Through extensive ablations, we progressively modernised the architecture design and training strategies and tackled key scaling challenges in sequence-to-sequence generative rendering, arriving at a design that’s simple, versatile, and scalable. Kaleido significantly outperforms prior generative models in few-view settings, and remarkably is the first zero-shot generative method matches InstantNGP-level rendering quality in multi-view settings. We view Kaleido also as an alternative step towards world modeling that flexibly spans a spectrum of “realities": with many views, it faithfully reconstructs grounded reality; with fewer views, it imagines plausible unseen details. 🔗 Explore more results and paper:
0:12

Introducing Kaleido💮 from AI at Meta — a universal generative neural rendering engine for photorealistic, unified object and scene view synthesis. Kaleido is built on a simple but powerful design philosophy: 3D perception is a form of visual common sense. Following this idea, we formulate rendering purely as a sequence-to-sequence generation problem, successfully unifying neural rendering with the architecture principles behind modern language and video models. Unlike traditional neural rendering methods, Kaleido learns 3D purely in a data-driven way, without explicit 3D representations or structures. It acquires spatial understanding directly through large-scale video pretraining, then multi-view 3D data finetuning, inspired by how LLMs acquire textual common sense from large corpora before specialising in domains like coding. Through extensive ablations, we progressively modernised the architecture design and training strategies and tackled key scaling challenges in sequence-to-sequence generative rendering, arriving at a design that’s simple, versatile, and scalable. Kaleido significantly outperforms prior generative models in few-view settings, and remarkably is the first zero-shot generative method matches InstantNGP-level rendering quality in multi-view settings. We view Kaleido also as an alternative step towards world modeling that flexibly spans a spectrum of “realities": with many views, it faithfully reconstructs grounded reality; with fewer views, it imagines plausible unseen details. 🔗 Explore more results and paper:

Shikun Liu

22,332 просмотров • 9 месяцев назад

Redbelly Testnet is LIVE! Our partners are onboarding billions of dollars in Real World Assets, establishing Redbelly Network as the market leader in the #RWA space. Join the compliant RWA movement: $RBNT #Blockchain #Testnet
0:18

Redbelly Testnet is LIVE! Our partners are onboarding billions of dollars in Real World Assets, establishing Redbelly Network as the market leader in the #RWA space. Join the compliant RWA movement: $RBNT #Blockchain #Testnet

Redbelly Network

214,088 просмотров • 2 лет назад

NEURALINK: WHERE YOUR THOUGHTS LITERALLY CONTROL THE WORLD AROUND YOU It’s not imagination. It’s movement, precision, and control, all from the mind alone. •⁠ ⁠Neuralink bypasses broken pathways, restoring motion through direct brain signals •⁠ ⁠Thought turns into action in under 100ms: faster than most natural reflexes •⁠ ⁠Robotic limbs respond instantly, moving like seamless extensions of your body •⁠ ⁠Users play games, browse, and type with pure intention, no movement required •⁠ ⁠This is more than tech; it’s the foundation of human-AI symbiosis The mind-machine merge has started, rewriting what it means to be human forever. Source: xAI, Massachusetts Institute of Technology (MIT)
0:14

NEURALINK: WHERE YOUR THOUGHTS LITERALLY CONTROL THE WORLD AROUND YOU It’s not imagination. It’s movement, precision, and control, all from the mind alone. •⁠ ⁠Neuralink bypasses broken pathways, restoring motion through direct brain signals •⁠ ⁠Thought turns into action in under 100ms: faster than most natural reflexes •⁠ ⁠Robotic limbs respond instantly, moving like seamless extensions of your body •⁠ ⁠Users play games, browse, and type with pure intention, no movement required •⁠ ⁠This is more than tech; it’s the foundation of human-AI symbiosis The mind-machine merge has started, rewriting what it means to be human forever. Source: xAI, Massachusetts Institute of Technology (MIT)

Mario Nawfal

99,457 просмотров • 6 месяцев назад

3D Character Artist Quaternius released the Universal Animation Library 2, a collection of more than 130 animations covering various movements, such as melee and armed combos, parkour movement, farming, zombie locomotion, and more. Details:
0:10

3D Character Artist Quaternius released the Universal Animation Library 2, a collection of more than 130 animations covering various movements, such as melee and armed combos, parkour movement, farming, zombie locomotion, and more. Details:

80 LEVEL

40,842 просмотров • 5 месяцев назад

Midjourney v8.2 + GPT Image 2 + Seedance 2.0 Ricochet - Prompts in the replies. I've been using 𝕏Caliber 's new RunTitan platform for some of my Seedance 2.0 generations. This isn't sponsored, I'm paying for it myself. It has a fast, clean interface and the generation costs are very reasonable. I'd definitely recommend checking it out.
0:15

Midjourney v8.2 + GPT Image 2 + Seedance 2.0 Ricochet - Prompts in the replies. I've been using 𝕏Caliber 's new RunTitan platform for some of my Seedance 2.0 generations. This isn't sponsored, I'm paying for it myself. It has a fast, clean interface and the generation costs are very reasonable. I'd definitely recommend checking it out.

Kōda

21,481 просмотров • 25 дней назад

AMAZING SARS-COV-2 ! The virus can ENTER and FUSE (leading to the creation of syncytia), using both RAFT-DEPENDENT (LIPID RAFT as CHOLESTEROL) and RAFT-INDEPENDENT pathways. Let me explain in simple terms...
0:15

AMAZING SARS-COV-2 ! The virus can ENTER and FUSE (leading to the creation of syncytia), using both RAFT-DEPENDENT (LIPID RAFT as CHOLESTEROL) and RAFT-INDEPENDENT pathways. Let me explain in simple terms...

Emmanuel

87,276 просмотров • 2 лет назад

nano banana 2's understanding of dimension and 3D space is so crazy now, its like a personal designer in your pocket step 1: go to -> edit image -> select nano banana 2 -> upload floor plan -> type prompt (reply here if you want my prompt)
0:19

nano banana 2's understanding of dimension and 3D space is so crazy now, its like a personal designer in your pocket step 1: go to -> edit image -> select nano banana 2 -> upload floor plan -> type prompt (reply here if you want my prompt)

el.cine

59,888 просмотров • 4 месяцев назад

Empty bottle caught in aerodynamic drag. As the truck moves, it creates a low-pressure volume of space behind it — roughly shaped like a 3D teardrop — inside of which is less air pressure, which pulls the bottle in a perpetual spin behind the truck.
0:11

Empty bottle caught in aerodynamic drag. As the truck moves, it creates a low-pressure volume of space behind it — roughly shaped like a 3D teardrop — inside of which is less air pressure, which pulls the bottle in a perpetual spin behind the truck.

Massimo

129,218 просмотров • 1 год назад

Spider-Man 2’s New Fast Travel Powered by the PS5’s SSD is insanely impressive and is one of those things that can only be done on the PS5. Loading screens in real new generation games are a thing of the past
0:18

Spider-Man 2’s New Fast Travel Powered by the PS5’s SSD is insanely impressive and is one of those things that can only be done on the PS5. Loading screens in real new generation games are a thing of the past

Synth Potato🥔

4,848,813 просмотров • 2 лет назад

New vibecode project: 3D world generation on your phone ✨ Take a photo of a room or space, and transform it into a new style. Then generate a world you can explore. Made in Cursor using two new APIs - Flux.2 Klein (BlackForestLabsAI - Unofficial) and Marble (World Labs). How it works 👇
0:15

New vibecode project: 3D world generation on your phone ✨ Take a photo of a room or space, and transform it into a new style. Then generate a world you can explore. Made in Cursor using two new APIs - Flux.2 Klein (BlackForestLabsAI - Unofficial) and Marble (World Labs). How it works 👇

Justine Moore

42,665 просмотров • 6 месяцев назад

⚠️ SPOILER ZAYNE'S 5TH INTERCOSTAL SPACE! • • • • • • • • • so actually in japanese dub, the local doctor asked if we were zayne's family member instead of S.O. so if we're taking into context of S.O. in another meaning here, the doctor thought of us as zayne's WIFE! 🤭
0:12

⚠️ SPOILER ZAYNE'S 5TH INTERCOSTAL SPACE! • • • • • • • • • so actually in japanese dub, the local doctor asked if we were zayne's family member instead of S.O. so if we're taking into context of S.O. in another meaning here, the doctor thought of us as zayne's WIFE! 🤭

𝒔𝒖𝒛𝒖

21,829 просмотров • 1 год назад

A TEAM JUST DEPLOYED 15 AUTONOMOUS LOOP AGENTS FROM A SINGLE PROMPT USING APPLIED GRAPH ENGINEERING Most developers still manually hardcode multi agent systems, writing separate logic for every individual task. Graph engineering changes this by using a central topological map to spin up all 15 nodes simultaneously. A single 200 word input generates the architecture, routing 120 unique pathways between agents instantly. Instead of failing under conflicting instructions, these loop agents self correct via continuous state sharing. Managing a 15 node mesh requires high token throughput, making this dependent on strict low-] latency API tiers. See exactly how this automated multi agent graph architecture actually operates in real time ↓
0:15

A TEAM JUST DEPLOYED 15 AUTONOMOUS LOOP AGENTS FROM A SINGLE PROMPT USING APPLIED GRAPH ENGINEERING Most developers still manually hardcode multi agent systems, writing separate logic for every individual task. Graph engineering changes this by using a central topological map to spin up all 15 nodes simultaneously. A single 200 word input generates the architecture, routing 120 unique pathways between agents instantly. Instead of failing under conflicting instructions, these loop agents self correct via continuous state sharing. Managing a 15 node mesh requires high token throughput, making this dependent on strict low-] latency API tiers. See exactly how this automated multi agent graph architecture actually operates in real time ↓

slash1s

289,590 просмотров • 2 дней назад

SpaceX is becoming one of the biggest solar-power operators in space. Gen 1 Starlink: ~10 MW in orbit. Gen 2: ~100 MW. Gen 3: targeting ~1,000 MW. That's roughly a 10x jump every satellite generation. Most people think of Starlink as an internet network. Meanwhile, thousands of satellites are turning orbit into a rapidly growing solar-power grid. SpaceX Starlink
0:16

SpaceX is becoming one of the biggest solar-power operators in space. Gen 1 Starlink: ~10 MW in orbit. Gen 2: ~100 MW. Gen 3: targeting ~1,000 MW. That's roughly a 10x jump every satellite generation. Most people think of Starlink as an internet network. Meanwhile, thousands of satellites are turning orbit into a rapidly growing solar-power grid. SpaceX Starlink

Mario Nawfal

68,969 просмотров • 1 месяц назад

"Nobody has the right to live protected from offense, insult, or hurt feelings. It's an occupational hazard of living in society. If you really can't take it… become a hermit." 19-sec clip – pure, unfiltered truth bomb 👇 Still waiting for the "safe space" generation to hear this one.
0:19

"Nobody has the right to live protected from offense, insult, or hurt feelings. It's an occupational hazard of living in society. If you really can't take it… become a hermit." 19-sec clip – pure, unfiltered truth bomb 👇 Still waiting for the "safe space" generation to hear this one.

Camus

30,281 просмотров • 6 месяцев назад