Urgent.News

What's breaking now, across thousands of outlets.

AI

AI needs fresh data. Vector indexing is getting in the way

It’s not a revelation to say AI workflows are changing enterprise infrastructure. What gets less attention, though, is how they’re The post AI needs fresh data. Vector indexing is getting in the way appeared first on The New Stack .

AI needs fresh data. Vector indexing is getting in the way

AI workflows are transforming enterprise infrastructure, but the shift is also impacting data processing and search. As organizations create generative AI applications, retrieval-augmented generation systems, and AI agents, they produce a vast number of vectors from unstructured data. The challenge now is to find ways to quickly locate the right information within this growing volume of vectorized data.

As Rachael Wade, global tech alliances manager at Elastic, notes, "For customers who want to search across petabytes of data, they may notice that it takes days to weeks to index." This indexing process is a significant bottleneck for AI applications relying on current data. The HNSW (Hierarchical Navigable Small World) graph is employed during indexing to connect similar vectors and enable efficient search.

However, building a vector index at enterprise scale requires substantial computational resources and numerous mathematical operations. As enterprises lack the necessary computing power, indexing becomes slower, leaving AI systems with outdated data. Manas Singh, technical product manager for accelerated ML and vector search at Nvidia, highlights that indexing speed is now the main constraint for maintaining search currentness.

GPUs are emerging as a solution for accelerated vector indexing, as they excel in handling large-scale, high-intensity workloads. Elastic and Nvidia have collaborated on integrating GPU-accelerated vector indexing for Elasticsearch, utilizing Nvidia's cuvs-java library for faster index building and embedding retrieval.

Written by urgent.news from The New Stack's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at thenewstack.io →

More in AI

More from Wednesday 18 November →