Google LLC today announced new capabilities for customers to harness their data in the cloud for use with generative artificial intelligence.

The additions include new tools from the company’s Google Cloud unit with BigQuery Studio into a single interface that simplifies data workflows.

Data has become the golden standard for generative AI, and companies are awash in it from a multitude of sources at massive scale. They need to be able to sift through it, organize it and prepare it for use in in engineering, analytics and predictive analysis with AI models in order to provide for business needs.

Announced Google’s cloud conference Google Cloud Next 2023, BigQuery Studio, now in preview, provides customers a single interface for data analytics in Google Cloud that allows engineers and developers to work on data without having to switch between tools. It also allows teams to edit SQL, Python, Spark and other languages to run analytics at petabyte scale without additional infrastructure.

“Unstructured data is roughly 80% of the data out there and it’s largely never analyzed,” Gerrit Kazmaier, general manager and vice president of analytics at Google Cloud, told SiliconANGLE in an interview. “There’s a whole dark area of enterprise data that’s not used today, yet generative AI is incredibly promising in unlocking that data and relating it with structured data.”

With powerful tools such as BigQuery Studio, Kazmaier explained, data analytics teams can take unstructured data from numerous sources such as customer conversations, audio and video, and prepare it and then ingest it into the AI models. “When you look at AI projects, and what makes them work well, is that most AI projects are really data projects in disguise,” Kazmaier said.

The problem is that before data can be pulled into an AI model, it first needs to be prepared and there’s a lot of work that needs to be done to it. Kazmaier said the paradigm of BigQuery Studio was all about “bringing AI to your data instead of bringing data to your AI.”

The biggest way to provide that is to allow AI and data teams to collaborate meaningfully. With BigQuery Studio, the analytics and AI teams not only get access to a full interface that puts all their tools in one place, but it also abstracts away the infrastructure so they don’t need to worry about it.

Google built Studio with an integration with Colaboratory Enterprise, which allows users to put all their data into Notebooks, collaborative spaces that they can share with other engineers, data scientists and colleagues. They can then take the data and scripts that they build and run AI models with them in a shared environment.

To further the process of bringing AI to data, Google Cloud brought access to foundational generative AI models into BigQuery through Vertex AI, Google’s source for AI models, including its own PaLM 2. That will allow data teams to remove complexity and scale out simple SQL queries and build their own machine learning models using insights generated by AI through the data they have in BigQuery. The same access will allow AI engineers and data scientists to do AI tuning using BigQuery data with vector and semantic search for retrieval of unstructured data, delivering new opportunities to reveal insights in the data.

Google brings conversational AI to data analytics

In order to provide users a better experience when building and creating analytics experiences in its products, Google has brought its most recent conversational AI experience into BigQuery and Looker in preview. Looker is its business intelligence and data analysis platform.

Data scientists spend a lot of time writing SQL and Python code in BigQuery, explained Kazmaier, and that can become extremely tedious. The mental workload of looking through inherited code and trying to understand a workflow or an ecosystem can be taxing. In particular, going through documentation and adjusting code to fit a new dataset can make data preparation more difficult.

Duet AI is Google’s most recent foray into an embedded generative AI assistant with conversational capabilities that it has integrated into many of its products, including Google Workspace. The AI service, which made its debut in Google’s Workspace office productivity platform in the spring, will be the star of the show at Cloud Next, with applications in new areas ranging from database access to security.

In Workspace, Duet AI acts as an assistant within Gmail, Docs and Slides, taking conversational prompts, producing images and text, organizing spreadsheets and more. Since it’s powered by Google’s PaLM 2 AI model, it can generate and understand computer code on request.

“It has all of the abilities to generate and optimize SQL code, it can help you understand and describe code,” said Kazmeier. “It acts as an intelligent assistant that works with you right in the context of your work. We also went one step further and also made Duet AI in BigQuery capable of analyzing your data.”

Duet AI can examine the underlying databases and queries being run and provide insights and provides statements such as, “This is a good way to analyze the following metrics,” such as insights into how to analyze customer records. Kazmaier pointed out that the AI model would serve data scientists by helping them avoid confirmation bias and avoid blind spots in their own data.

In Looker, the addition of Duet AI is designed to make business analytics more accessible to less technical users by providing them a conversational interface to help them build their own queries and dashboards.

In the past, business users relied heavily on existing template or analysts to build them what they needed. Now, they can simply ask Duet AI to make them a dashboard using conversational language and in essence produce a report based on what they need. If the first result isn’t what they need, they can keep refining it with further queries until they have what they want.

Users can ask Duet AI to generate intelligent summaries of metrics and produce formula. With the AI to assist users can rapidly create code based on their prompts that can do calculations for graphs, visuals and other dashboard elements that would have taken a lot of work in the past.

Perhaps users are not looking for a dashboard, so Duet AI can help users create slide decks that help users present the metrics and data for their coworkers instead. The objective behind the AI assistant is to provide a conversational interface that will unlock opportunities for users in Looker to operate the tool as easily as they would by just talking to it.

It also allows users to interrogate their business data in the same way that someone might ask Google a question. This way, users have a tool that behaves in the manner of a data analyst capable of providing intelligent commentary about their data.

Jumpstarting developers into the cloud

Since getting started with a new platform can be difficult, Google announced a way for new developers to get a start with Google Cloud called Jump Start Solutions that will help users get started building. Within the portal developers can quickly find pre-built samples that they can deploy directly within their Google Cloud console including three-tier web apps, secure CI/CD pipelines, load-balanced virtual machines, dynamic websites, an analytics lakehouse or a generative AI app to summarize large documents or an AI-powered image processing app.

All of these solutions are designed with the best practice principles and can be launched with one click and are designed to be an easy starting point. Each comes with its own estimated cost, comprehensive reference architecture and tutorials to get any prospective developer started on a project.

Image: Pixabay