Загрузка видео...

Не удалось загрузить видео

Возникла проблема при загрузке этого видео. Это может быть связано с временными проблемами сети или видео может быть недоступно.

На главную

What would it look like to combine Google search (shallow Knowledge Graph reasoning over an ultra-ultra-wide index) with LLM in-context learning (highly intelligent operations on a tiny index)?

Dwarkesh Patel

242,238 subscribers

101,417 просмотров • 1 год назад •via X (Twitter)

Наука и технологии Новости и политика Образование

Anya Rossi• Live Now

Private livecam show

Комментарии: 10

Фото профиля Dwarkesh Patel

Dwarkesh Patel1 год назад

Transcript: Youtube: Spotify: Apple:

Фото профиля World Scientific

World Scientific1 год назад

Uncover the latest insights in algebraic computation! 📊👥 Discover new decompositions of graph braid groups and their implications. Click the banner below to read the open access article for FREE! #Algebra #Combinatorics #Mathematics

Фото профиля scuzzlebot

scuzzlebot1 год назад

The future may lie in "cognitive layering" - shallow but broad knowledge as a foundation, with deep reasoning capabilities that can dynamically zoom in and perform complex operations on relevant subsets. Like how human experts combine broad awareness with focused analysis. 🧠

Фото профиля Gerbil

Gerbil1 год назад

Surely were going to find out in the near future

Фото профиля Gary Kibitz

Gary Kibitz1 год назад

Sounds like Jeff made up and answered a harder question (hand-waving), probably because the answer to your question is that Google is already doing that and it's all secret.

Фото профиля Richard Harper

Richard Harper1 год назад

Perhaps read the two appendices of the Syntopicon of The Great Books explaining the methods used in the attempt to index, to .. 'bring together the topics', which seemed important at the time for the continuation of Western Civilization when it was first published in 1953.

Фото профиля Maxime Batandeo

Maxime Batandeo1 год назад

The curation of what to put in the context will be key to unlock actual new knowledge and it will be shaped as a graph. With semantic rules and/or ontology to structure it. Here is how I think of it : You label the inputs as a knowledge graph You prompt the LLM to generate a subgraph of outputs You integrate the subgraph into the global knowledge graph. Repeat. Do that for argumentation for example and you unlock actual reasoning

Фото профиля SubMinima

SubMinima1 год назад

I don't think we actually need a large memory (context) for this... Just a large database and the right search parameters

Фото профиля Vineet Tiruvadi, MD PhD

Vineet Tiruvadi, MD PhD1 год назад

Do you consider LLM in-context learning "highly intelligent"??

Фото профиля Woke GPT

Woke GPT1 год назад

retard moving mouth, news at 0

Похожие видео

Our new short course, Knowledge Graphs for RAG, is now available! Knowledge graphs are a data structure that is great at capturing complex relationships between data of multiple types. By enabling more sophisticated retrieval of text than similarity search alone, knowledge graphs can improve the context you pass to the LLM and the performance of your RAG applications. In this course, taught by Andreas Kollegger of Neo4j, you’ll - Explore how knowledge graphs work by building a graph of public financial documents from scratch - Learn to write queries that retrieve text and data from the graph and use it to enhance the context you pass to an LLM chatbot - Combine a knowledge graph with a question-answer chain to build better RAG-powered chat systems Sign up here!

Our new short course, Knowledge Graphs for RAG, is now available! Knowledge graphs are a data structure that is great at capturing complex relationships between data of multiple types. By enabling more sophisticated retrieval of text than similarity search alone, knowledge graphs can improve the context you pass to the LLM and the performance of your RAG applications. In this course, taught by Andreas Kollegger of Neo4j, you’ll - Explore how knowledge graphs work by building a graph of public financial documents from scratch - Learn to write queries that retrieve text and data from the graph and use it to enhance the context you pass to an LLM chatbot - Combine a knowledge graph with a question-answer chain to build better RAG-powered chat systems Sign up here!

Andrew Ng

244,403 просмотров • 2 лет назад

This idea is either extremely smart or an extremely stupid—no in-between. What if your LLM *is* your search engine? How would you look like inside it? Forget about Perplexity, DeepResearch. What if LLM is your entire Google? Pagination, links and everything - just like the old days: no chat UI, classic Google vibe. If you're unsure what's that mean, watch the demo video below first. We call it LLM-as-SERP (Search Engine Results Page).

This idea is either extremely smart or an extremely stupid—no in-between. What if your LLM is your search engine? How would you look like inside it? Forget about Perplexity, DeepResearch. What if LLM is your entire Google? Pagination, links and everything - just like the old days: no chat UI, classic Google vibe. If you're unsure what's that mean, watch the demo video below first. We call it LLM-as-SERP (Search Engine Results Page).

Jina AI

73,960 просмотров • 1 год назад

Lets build `Auto-RAG` where we let the LLM pull the data it needs from different sources. 🔎 The user asks a question. 🤔 LLM decides whether to search its knowledge, memory, internet or make an API call. ✍️ LLM answers with the context. Code:

Lets build `Auto-RAG` where we let the LLM pull the data it needs from different sources. 🔎 The user asks a question. 🤔 LLM decides whether to search its knowledge, memory, internet or make an API call. ✍️ LLM answers with the context. Code:

Ashpreet Bedi

174,580 просмотров • 2 лет назад

Is football coming home? ⚽️ We’ve been using Veo 3 in Gemini to imagine what that would look like 👀 Have you tried turning your prompts into short films with Google AI Pro and Ultra?

Is football coming home? ⚽️ We’ve been using Veo 3 in Gemini to imagine what that would look like 👀 Have you tried turning your prompts into short films with Google AI Pro and Ultra?

Google UK

14,086,593 просмотров • 1 год назад

Semantic search alone doesn't cut it. Neither does brute-force grep. Agents need both. Today we're shipping the Retrieval Harness in LlamaParse Index: semantic search, server-side grep, and file-level navigation working together in a single agent reasoning loop. 🦙🌤️ Grep a file, list what's in an index, read past a chunk boundary, run hybrid search with reranking — all as native agent tools. Now in beta across all paid tiers. Full breakdown in the blog 👇 Learn More:

Semantic search alone doesn't cut it. Neither does brute-force grep. Agents need both. Today we're shipping the Retrieval Harness in LlamaParse Index: semantic search, server-side grep, and file-level navigation working together in a single agent reasoning loop. 🦙🌤️ Grep a file, list what's in an index, read past a chunk boundary, run hybrid search with reranking — all as native agent tools. Now in beta across all paid tiers. Full breakdown in the blog 👇 Learn More:

LlamaIndex 🦙

29,181 просмотров • 1 месяц назад

Sam Altman says the perfect AI is “a very tiny model with superhuman reasoning, 1 trillion tokens of context, and access to every tool you can imagine.” It doesn't need to contain the knowledge - just the ability to think, search, simulate, and solve anything.

Sam Altman says the perfect AI is “a very tiny model with superhuman reasoning, 1 trillion tokens of context, and access to every tool you can imagine.” It doesn't need to contain the knowledge - just the ability to think, search, simulate, and solve anything.

vitrupo

775,853 просмотров • 1 год назад

AI Mode in Search introduces new reasoning and tools under the hood with a custom version of Gemini models. For any question, it makes a plan, breaks it down into related subtopics, and runs multiple Google searches to find the most helpful and reliable info. With this query fan-out, AI Mode finds real-time info from both the web and from Google, like facts from our Knowledge Graph, info about the real-world, and shopping details on billions of products. This approach helps you get a greater breadth and depth of information than a traditional search on Google.

AI Mode in Search introduces new reasoning and tools under the hood with a custom version of Gemini models. For any question, it makes a plan, breaks it down into related subtopics, and runs multiple Google searches to find the most helpful and reliable info. With this query fan-out, AI Mode finds real-time info from both the web and from Google, like facts from our Knowledge Graph, info about the real-world, and shopping details on billions of products. This approach helps you get a greater breadth and depth of information than a traditional search on Google.

Robby Stein

38,535 просмотров • 1 год назад

I got an LLM to write frontend and backend code for a graph visualizer, fix its own bug and literally deploy it to the public web with google cloud. This would take most devs 2 days. It took AI ~20mins.

I got an LLM to write frontend and backend code for a graph visualizer, fix its own bug and literally deploy it to the public web with google cloud. This would take most devs 2 days. It took AI ~20mins.

Deedy

232,009 просмотров • 1 год назад

Just for fun, here's what 32 simultaneous long-context generations with Qwen3 Next 80B looks like on an M3 Ultra. Using the new batch generation in mlx-lm. Context size for each is about 5k tokens:

Just for fun, here's what 32 simultaneous long-context generations with Qwen3 Next 80B looks like on an M3 Ultra. Using the new batch generation in mlx-lm. Context size for each is about 5k tokens:

Awni Hannun

50,272 просмотров • 10 месяцев назад

See what clothes look like on *you* before you tap to buy with our new try on tool in Search Labs. 👖 Learn more → #GoogleIO

See what clothes look like on you before you tap to buy with our new try on tool in Search Labs. 👖 Learn more → #GoogleIO

Google

118,154 просмотров • 1 год назад

Google just released free 1-hour course on building agentic knowledge Graphs from 0% to 100%: 10% → 4:01 - how to build a GraphRAG agent 30% → 15:00 - Graph Engineering explanation 55% → 30:00 - Agentic search Engineering 80% → 35:48 - Graph Engineering practice 100% → 47:06 - self-improving agents in graphs this free Google course mass replaces a $500 graph engineering bootcamp - learn it in 60 min to 100% watch it today - then read the full graph playbook in the article below ↓

Google just released free 1-hour course on building agentic knowledge Graphs from 0% to 100%: 10% → 4:01 - how to build a GraphRAG agent 30% → 15:00 - Graph Engineering explanation 55% → 30:00 - Agentic search Engineering 80% → 35:48 - Graph Engineering practice 100% → 47:06 - self-improving agents in graphs this free Google course mass replaces a $500 graph engineering bootcamp - learn it in 60 min to 100% watch it today - then read the full graph playbook in the article below ↓

codila

245,433 просмотров • 20 дней назад

Mind = BLOWN 🤯 Grok 3 just launched with reasoning abilities that put other AIs to shame. Just spent hours testing it vs ChatGPT. My honest breakdown: • Reasoning mode destroys basic Grok for complex tasks • Image generation is SIGNIFICANTLY better than ChatGPT • Coding still needs work (ChatGPT wins here) • Includes web search similar to Perplexity ($16/mo for 50 searches) • New platform with sharable chats that index on Google Who's trying it?

Mind = BLOWN 🤯 Grok 3 just launched with reasoning abilities that put other AIs to shame. Just spent hours testing it vs ChatGPT. My honest breakdown: • Reasoning mode destroys basic Grok for complex tasks • Image generation is SIGNIFICANTLY better than ChatGPT • Coding still needs work (ChatGPT wins here) • Includes web search similar to Perplexity ($16/mo for 50 searches) • New platform with sharable chats that index on Google Who's trying it?

Julian Goldie SEO

11,517 просмотров • 1 год назад

What if drop shadows actually look like real shadows? We are using Omma to explore a new kind of canvas that employs real-light approximations rather than simple 2D drop shadows. It runs on WebGPU and uses real-time path tracing to simulate more realistic shading. It also supports SVG shapes. The Z-index defines actual depth, too (objects with a lower z-index would be affected by shadows of other objects).

What if drop shadows actually look like real shadows? We are using Omma to explore a new kind of canvas that employs real-light approximations rather than simple 2D drop shadows. It runs on WebGPU and uses real-time path tracing to simulate more realistic shading. It also supports SVG shapes. The Z-index defines actual depth, too (objects with a lower z-index would be affected by shadows of other objects).

Omma

12,155 просмотров • 3 месяцев назад

NVIDIA sent us 2 DGX Sparks. For a while we wondered what we would do with them. The memory bandwidth is 273GB/s making it 3x slower than an M3 Ultra (819GB/s) for batch_size=1 inference. But it has 4x more FLOPS (100 TFLOPS compared to 26 TFLOPS). So we thought, what if we could combine the DGX Spark & M3 Ultra, and make use of both the massive compute on the DGX Spark and the massive memory-bandwidth on the M3 Ultra. We came up with a way to split inference across both devices and achieve a speedup of up to 4x for long prompts compared to the M3 Ultra on its own. Full details in the blog post linked below.

NVIDIA sent us 2 DGX Sparks. For a while we wondered what we would do with them. The memory bandwidth is 273GB/s making it 3x slower than an M3 Ultra (819GB/s) for batch_size=1 inference. But it has 4x more FLOPS (100 TFLOPS compared to 26 TFLOPS). So we thought, what if we could combine the DGX Spark & M3 Ultra, and make use of both the massive compute on the DGX Spark and the massive memory-bandwidth on the M3 Ultra. We came up with a way to split inference across both devices and achieve a speedup of up to 4x for long prompts compared to the M3 Ultra on its own. Full details in the blog post linked below.

Alex Cheema

281,225 просмотров • 10 месяцев назад

Active v/s Passive: Where should you invest? If you are an investor, you have to flip the question from How many funds beat the index? to Which fund beats the index consistently over a 3/5 yrs time frame? You can use rolling returns analysis to determine which fund consistently outperforms the index. Any fund that has outperformed the index for more than 70% of the time over 3/5 yrs t/f is a good fund imho. I had spoken about it in my podcast with Groww. Check it out.

Active v/s Passive: Where should you invest? If you are an investor, you have to flip the question from How many funds beat the index? to Which fund beats the index consistently over a 3/5 yrs time frame? You can use rolling returns analysis to determine which fund consistently outperforms the index. Any fund that has outperformed the index for more than 70% of the time over 3/5 yrs t/f is a good fund imho. I had spoken about it in my podcast with Groww. Check it out.

Sandeep Kulkarni

30,071 просмотров • 9 месяцев назад

5/ We're introducing Gemini Agent: It uses Gemini 3's advanced reasoning to break down complex tasks, like booking local services or organizing your inbox, into actionable steps. It can then propose actions like drafting your replies or archiving emails, all using deeper integrations with Google apps. Available on the web for Google AI Ultra subscribers in the US.

5/ We're introducing Gemini Agent: It uses Gemini 3's advanced reasoning to break down complex tasks, like booking local services or organizing your inbox, into actionable steps. It can then propose actions like drafting your replies or archiving emails, all using deeper integrations with Google apps. Available on the web for Google AI Ultra subscribers in the US.

Sundar Pichai

132,013 просмотров • 9 месяцев назад

Our last Stanford guest lecture - Ekdeep Singh Lubana on what counts as an explanation & a neuro-inspired "model systems approach" to interp Plus, how in-context learning and many-shot jailbreaking are explained by LLM representations changing in-context (as a case study for that approach) 00:33 - What counts as an explanation? 04:47 - Levels of analysis & standard interpretability approaches 18:19 - The "model systems" approach to interp [Case study on in-context learning] 23:36 - How LLM representations change in-context 44:10 - Modeling ICL with rational analysis 1:10:54 - Conclusion & questions Thanks again to Surya Ganguli for having us in his class!

Our last Stanford guest lecture - Ekdeep Singh Lubana on what counts as an explanation & a neuro-inspired "model systems approach" to interp Plus, how in-context learning and many-shot jailbreaking are explained by LLM representations changing in-context (as a case study for that approach) 00:33 - What counts as an explanation? 04:47 - Levels of analysis & standard interpretability approaches 18:19 - The "model systems" approach to interp [Case study on in-context learning] 23:36 - How LLM representations change in-context 44:10 - Modeling ICL with rational analysis 1:10:54 - Conclusion & questions Thanks again to Surya Ganguli for having us in his class!

Goodfire

31,410 просмотров • 8 месяцев назад

Give Claude Code a semantic filesystem 🗃️🛠️ Giving Claude Code access to the right CLI tools over your filesystem turns it into a general agent capable of automating far more knowledge work beyond code - it can do dynamic financial/legal/medical/technical/backoffice analysis over any subset of documents. With our latest release of semtools 💫, you can now manually or *agentically* create a persistent workspace over any subset of files. This gives Claude Code the ability to get blazing-fast, local semantic search over any data, while still allowing it to chain with commands like grep/cat/etc. so that it can load in dynamic context instead of naive top-k vector search. The coding agent can dynamically index data and use those indexes, instead of having to rebuild it every time. So you get the benefits of fast search along with agentic reasoning over CLI tools mentioned above. Come check it out!

Give Claude Code a semantic filesystem 🗃️🛠️ Giving Claude Code access to the right CLI tools over your filesystem turns it into a general agent capable of automating far more knowledge work beyond code - it can do dynamic financial/legal/medical/technical/backoffice analysis over any subset of documents. With our latest release of semtools 💫, you can now manually or agentically create a persistent workspace over any subset of files. This gives Claude Code the ability to get blazing-fast, local semantic search over any data, while still allowing it to chain with commands like grep/cat/etc. so that it can load in dynamic context instead of naive top-k vector search. The coding agent can dynamically index data and use those indexes, instead of having to rebuild it every time. So you get the benefits of fast search along with agentic reasoning over CLI tools mentioned above. Come check it out!

Jerry Liu

77,373 просмотров • 10 месяцев назад