The success of generative artificial intelligence (AI) depends on data. But enterprises adopting generative AI (genAI) are struggling to access, manage and activate data of different formats.

“Data is crucial for AI, as it is the foundation upon which machine learning algorithms learn, are grounded, make predictions and improve performance over time,” Google Cloud VP and GM Brad Calder said.

For businesses to fully realize the potential of genAI, “they need to access, manage and activate structured and unstructured data across their operational and analytical systems,” Google Cloud GM & VP of Engineering Andi Gutmans added.

To address genAI data challenges, Google Cloud launched new capabilities for its database and data analytics portfolios that contextualize the new era of genAI and “break new ground from traditional approaches in data warehousing by accessing data and delivering insights in fresh new ways,” Calder said.

The hyperscaler shared the general availability of AlloyDB AI, which is designed to simplify the process of building enterprise-grade genAI applications. The AI database platform can run on-premises or in any public cloud environment and is ideal for transactional, analytical and vector workloads.

AlloyDB is in use at organizations like Character AI, B4A and Regnology, according to Google Cloud. Regnology’s regulatory reporting chatbot, for example, uses AlloyDB as a dynamic vector store that indexes repositories of regulatory guidelines, historical reporting data and compliance documents.

Google Cloud also announced the public preview of vector search capabilities across its different database offerings – Spanner, Memorystore for Redis and Cloud SQL for MySQL. Vector search is a key tool for building accurate and effective genAI applications because it simplifies the process of finding similar search results for unstructured data like text or images.

“For example, vector search can help retailers improve product recommendations, summarize fixes for common customer care support tickets or even help discover trends across large sets of documents,” according to Calder.

In other words, developers can store up to millions of vectors in the same MySQL instances they are already using, and they can store and perform vector searches in the operational database without setting up new systems.

For genAI applications that need high performance, the cloud hyperscaler launched vector storage and search in Memorystore for Redis, which will make each Redis instance capable of storing tens of millions of vectors and performing vector searches at single-digit-millisecond latency. This is ideal for large language model (LLM) semantic caching and other AI recommendation systems.

The new capabilities in Spanner allow developers to scale vector searches for highly partionable workloads, or large-scale workloads that involve billions of vectors and millions of queries per second. Spanner’s exact nearest neighbor search “reduces the search space to provide accurate, real-time results with low latency,” Gutsmans said.

Spanner can scale vector searches for highly partitionable workloads. Large-scale vector workloads that involve billions of vectors and millions of queries per second can be challenging for many systems. These workloads are a great fit for Spanner’s exact nearest neighbor search because Spanner can efficiently reduce the search space to provide accurate, real-time results with low latency.

Google Cloud also enhanced its data analytics portfolio with Gemini 1.0 Pro for BigQuery customers through the hyperscaler’s Vertex AI platform. These new integrations allow data engineers and analysts to use Gemini AI models for advanced multimodal reasoning on data stored in BigQuery.

“This can help healthcare providers improve patient care, make supply chains more efficient and increase customer engagement in telco, retail and financial services,” Calder said.

Google Cloud also launched a new BigQuery integration with Vertex AI for text and speech, which is available in preview. This integration allows enterprises to gather meaningful insights from unstructured data like text documents or audio files, which “unlock[s] new analytics scenarios that combine unstructured data with structured business data,” Calder said. For instance, data analysts can extract insights from call center audio recordings.