Google LLC’s cloud division today rolled out major milestone feature updates to its analytical and database engines BigQuery and AlloyDB that blur the lines between the roles of data and artificial intelligence.

Google Cloud today announced that BigQuery, the company’s serverless data warehouse that allows for the analysis of unstructured data, is getting access to Gemini, Google’s largest and most capable AI model, via the company’s managed service Vertex AI. AlloyDB AI also became generally available, a service that helps developers build high-performance AI applications using up-to-date data.

“In the past, you needed folks with deep machine learning knowledge and deep data science knowledge. But now gen AI is opening up innovation to everyone,” said Andi Gutmans, general manager and vice president for engineering, databases at Google Cloud. “And we know that AI cannot be successful without data.”

With access to Gemini in BigQuery, data analysis can use the AI models to use its advanced reasoning capabilities on extremely large-scale datasets, including unstructured and multimodal data – such as text documents, images, audio, video and more. This can be put to use by healthcare organizations to improve patient care, allow organizations with complex supply chains to understand their workflows and allow retail outlets to fit customer engagement better.

“Because 90%, roughly speaking, of the data out there is unstructured this data is usually not used in enterprise data analytics because you couldn’t work with it in a meaningful way,” said Gerrit Kazmaier, general manager and vice president of engineering, data analytics at Google Cloud. “But now with Gemini Pro in BigQuery, you can do basically all of the rich unstructured data analytics, and combine it with your structured data.”

This also provides enterprise customers access to the powerful generative AI engine behind Gemini to do sentiment extraction, classification, summarization, translation and other capabilities all within the AI model that can be surfaced using BigQuery data, Kazmaier explained.

For example, a call center could use it to extract insights from large volumes of call data to understand customer sentiment in audio messages to better align with what’s happening on the ground. Medical researchers could also use it on bulk numbers of X-rays to empower them to determine if large groups of patient files had similar pathologies that were difficult to discern.

AlloyDB AI, announced in August during Google Cloud Next 2023, is now generally available as well. It is an integrated set of capabilities built into AlloyDB for PostgreSQL designed to help developers build scalable generative AI applications using their own data. Using AlloyDB AI developers can efficiently combine the power of large language models with real-time data.

Google Cloud offers AlloyDB AI as a managed service as well as AlloyDB Omni, which customers can deploy anywhere, in their own cloud or on-premises.

In addition to the AlloyDB announcement, Google Cloud is also bringing vector search capabilities for other popular databases so that developers can build generative AI applications faster.

Vector embedding is a process used in deep learning to create a numerical representation of data that captures the relationship between words, phrases and other types of data. They represent different data types in a multidimensional space so that the semantic or contextual relationship between data points can be mapped such that similar types of data are “close” numerically and dissimilar types are “distant” numerically. As a result, similar types of data cluster together, making it easier to define relationships based on relationships such as two things being types of food, visible in the night sky, place names and so forth.

Vector embeddings are fundamental for semantic search and become even more important for creating generative AI applications when entire sentences, paragraphs and articles can be searched and analyzed according to their context.

Google Cloud databases receiving support for vector search include PostgreSQL, MySQL, AlloyDB, Redis, Spanner, Firestore and Bigtable.

“We do think vectors should be a first-class citizen in every database that really provides benefits around ease of use around price, performance and utility,” said Gutmans. “And you’re going to see us continue to innovate on this front.”

Image: Pixabay