Is This the End of LLMs? This Stanford Paper Could Make Nvidia's Memory Chips Unnecessary
Publicado el 8 nov 2026 · 11:22
Me gusta: 23 milComentarios: 1 mil+24h: +14 milVistas/h promedio: 6 mil
El rendimiento en cifras
Cifras a fecha de 16 nov 2026, 23:00 UTC.
Is This the End of LLMs? This Stanford Paper Could Make Nvidia's Memory Chips Unnecessary de Alex - a1crew tiene 1 M vistas, 8 días después de publicarse el 8 de noviembre de 2026.
Ahora recibe 6 mil vistas por hora.
Está en las tendencias de YouTube de 1 país, como mejor #44 en Suecia.
Su mejor día hasta ahora fue el 10 nov, con 919 mil vistas, y se ve sobre todo hacia las 19 UTC.
| Vistas | 1 M | 1.128.567 |
|---|---|---|
| Vistas en las últimas 24 horas | +14 mil | +14.382 |
| Vistas por hora ahora | 6 mil | 5537 |
| Veces lo habitual del canal | 56× | 55,7× |
| Mejor día | 10 nov · 919 mil | 919.027 |
| Hora más fuerte | 19 UTC | |
| Países en tendencias | 1 | |
| Mejor puesto | #44 · Suecia | |
| Me gusta | 23 mil | 23.238 |
| Comentarios | 1 mil | 1277 |
| Publicado | 8 de noviembre de 2026 | |
| Duración | 11:22 | |
| Categoría | Science & Technology | |
| Canal | Alex - a1crew | |
| Suscriptores del canal | 16 mil |
Vistas
1 M
+14 mil · 24 h
Vistas/h ahora
6 mil
× lo habitual
56×
Aún no hay suficientes lecturas.
Vistas por día
919 mil
mejor día, 10 nov
En tendencias de
1
mejor #44 en Suecia
Cuándo se ve
19
su hora más fuerte, tu hora
Dónde aparece1
- #44Suecia9 nov
Del registro de YouTube
Campos del video tal como los devuelve la API de YouTube.
- Publicado
- 8 nov 2026, 11:15
- Duración
- 11:22
- Categoría
- Science & Technology
- Idioma del audio
- inglés (en)
- Idioma del título
- inglés (en)
- Subtítulos
- Sí
- Definición
- HD
- Formato
- 2D
- Licencia
- Standard YouTube
- Contenido con licencia
- Sí
- Se puede insertar
- Sí
- Hecho para niños
- No
- Restringido por edad
- No
- Promoción pagada
- No
- Estadísticas públicas
- Sí
Temas
- Knowledge
- Technology
Título y descripción en 1 idiomas
- inglés
Grabado en
Canada · 56.130, -106.347
Etiquetas (27)
- Next Concept Modeling
- NCM
- Stanford AI research
- Nvidia
- HBM
- high bandwidth memory
- memory bandwidth
- KV cache
- GPU memory
- VRAM
- AI hardware
- LLM inference
- how LLMs work
- token by token generation
- parallel text generation
- nested dropout
- conditioning corruption
- test-time scaling
- Minimum Bayes Risk
- MBR
- small language models
- AI reasoning
- AI Sudoku
- consumer GPU
- frontier AI
- local AI
- AI research explained
Descripción
What if a single Stanford research paper could make Nvidia's most expensive memory chips unnecessary? The AI industry has hit a wall, and it isn't the one everyone talks about. Today's language models write one token at a time, so a 1,000-word answer means 1,000 passes through the entire model, and every pass hammers memory bandwidth. That's why the KV cache eats your GPU memory and why High Bandwidth Memory became the most expensive bottleneck in AI hardware. In this video we look at Next Concept Modeling (NCM), a new architecture that generates the whole answer in parallel and needs 32 passes instead of 1,000. We unpack the three technical secrets that make it work, from nested dropout to conditioning corruption to test-time scaling with Minimum Bayes Risk, and a Sudoku experiment where two models of exactly the same size got very different results. Then the big question: if AI stops needing massive memory bandwidth and only needs raw compute, what happens to the hardware monopoly, and could frontier-level AI run on consumer-grade silicon? Do you think consumer graphics cards will finally run frontier models? Original Stanford paper: https://arxiv.org/html/2610.14162v1 00:00 Can one paper beat Nvidia's memory chips? 00:29 How language models write today 00:58 1,000 words, 1,000 passes 01:31 The processor that just waits 01:43 The KV cache problem 02:02 Expensive memory chips 02:17 Enter Next Concept Modeling 02:35 Borrowed from image generation 02:52 Text in layers 03:31 One pass per layer 03:56 1,000 passes vs 32 04:32 Secret #1: nested dropout 05:02 Subscribe 05:29 Learning true concepts 06:04 The vulnerability: compounding errors 06:31 Secret #2: conditioning corruption 06:56 An AI that fixes its own mistakes 07:15 Can a tiny model compete? 07:30 Secret #3: test-time scaling 07:59 Minimum Bayes Risk 08:10 20 juniors vs 1 senior 08:39 The Sudoku test 09:09 The billion-dollar hardware flip 09:24 No free lunch 09:58 Frontier AI on consumer hardware 10:09 Is token-by-token a relic? 10:24 Will it scale? 10:41 Recap 11:05 Your turn #AI #LLM #Nvidia #MachineLearning #AIHardwareMostrar todo
What if a single Stanford research paper could make Nvidia's most expensive memory chips unnecessary? The AI industry has hit a wall, and it isn't the one everyone talks about. Today's language models write one token at a time, so a 1,000-word answer means 1,000 passes through the entire model, and every pass hammers memory bandwidth. That's why the KV cache eats your GPU memory and why High Bandwidth Memory became the most expensive bottleneck in AI hardware. In this video we look at Next Concept Modeling (NCM), a new architecture that generates the whole answer in parallel and needs 32 passes instead of 1,000. We unpack the three technical secrets that make it work, from nested dropout to conditioning corruption to test-time scaling with Minimum Bayes Risk, and a Sudoku experiment where two models of exactly the same size got very different results. Then the big question: if AI stops needing massive memory bandwidth and only needs raw compute, what happens to the hardware monopoly, and could frontier-level AI run on consumer-grade silicon? Do you think consumer graphics cards will finally run frontier models? Original Stanford paper: https://arxiv.org/html/2610.14162v1 00:00 Can one paper beat Nvidia's memory chips? 00:29 How language models write today 00:58 1,000 words, 1,000 passes 01:31 The processor that just waits 01:43 The KV cache problem 02:02 Expensive memory chips 02:17 Enter Next Concept Modeling 02:35 Borrowed from image generation 02:52 Text in layers 03:31 One pass per layer 03:56 1,000 passes vs 32 04:32 Secret #1: nested dropout 05:02 Subscribe 05:29 Learning true concepts 06:04 The vulnerability: compounding errors 06:31 Secret #2: conditioning corruption 06:56 An AI that fixes its own mistakes 07:15 Can a tiny model compete? 07:30 Secret #3: test-time scaling 07:59 Minimum Bayes Risk 08:10 20 juniors vs 1 senior 08:39 The Sudoku test 09:09 The billion-dollar hardware flip 09:24 No free lunch 09:58 Frontier AI on consumer hardware 10:09 Is token-by-token a relic? 10:24 Will it scale? 10:41 Recap 11:05 Your turn #AI #LLM #Nvidia #MachineLearning #AIHardware
◆ Calculado por viralshooter · el resto de las cifras proviene de la API de YouTube