TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

The Role of Structure in AI-Native Databases

1 点作者 CShorten超过 1 年前
Hey everyone! We are rolling along with daily episodes of AI-Native Databases before the holiday break! Episode 2 with Bob van Luijt and Paul Groth!<p>This was a fascinating one, diving into the role of structure in data! We begin with the most structured data object out there, the Knowledge Graph (KG), and how LLMs are transforming them. There are two aspects to this: (1) LLMs for KGs, using LLMs to extract relationships or predict missing links to build the KG, and (2) KGs for LLMs, using KGs to provide factual information to the LLM as in RAG. Then there is (3) which sits in the middle of 1 &amp; 2, using Text-to-Cypher&#x2F;SPARQL in order to query KGs. Both useful for humans looking to use KGs (1) and LLMs looking to use KGs (2)!<p>The conversation then takes an interesting turn into whether we need databases at all, and whether we need to structure our data at all? LLMs now give us the ability to structure data on-the-fly! For example, Generative Search describes summarizing search results with an LLM before sending them to the user. This is quite similar to RAG although I think it inspires a bit more of a &quot;search result parser&quot; perspective of the role of the LLM in the pipeline, whereas RAG motivates thinking about the system as a &quot;context provider&quot; to the LLM.<p>I think this is such a fascinating one in the investigation of how generative AI is transforming database systems and what the future of data management will look like! I hope you enjoy the podcast!<p>Link: https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=3ET69F7smk8

暂无评论

暂无评论