Google embeds generative AI features in cloud and Workspace
The tech giant's new API enables enterprises to scale and adopt models without their data becoming part of the training ground for large language models, the vendor said.
Google on Tuesday introduced new technology to enable enterprises to use generative AI to create, connect and collaborate.
The tech giant's PaLM API and MakerSuite are new systems with generative AI capabilities embedded in Google Cloud and Google Workspace, its business collaboration suite.
They are available in private preview format to a few select customers for testing.
In Workspace, Google has added generative AI-backed tools to help users save time when drafting text in Google Docs and Gmail.
One new tool helps users combat writer's block. Users can type in their desired topic and the generative AI tool will create a draft based on the topic. The tool will continue to offer edits and suggested topics as dictated by the user.
Meanwhile,as the tech giant competes against Microsoft and others in a fast-moving race to develop enterprise applications using generative AI, it is focusing on responsible AI and generative AI adoption by enterprises.
The PaLM API lets developers experiment with Google's large language models. Developers can access models (such as LaMDA) optimized for applications such as content generation, chat, summarization and classification.
Released alongside the API, MakerSuite enables developers to tune the custom model and export prompts as code in preferred languages and frameworks.
Google's distinguishing factor
PaLM API is about "enabling enterprises to start building their own versions of the [Google generative AI] model," which they can use in their own industry context, said Gartner analyst Chirag Dekate.
"These are not just models for the sake of models," he said, adding that the models were engineered using the latest hardware techniques in Google's cloud infrastructure and data center environments.
Google also is aiming to distinguish itself by assuring enterprises it will not use their data to train underlying foundation models.
With most generative AI models, it is typical for the data model to use user queries for training.
Enterprises can start trusting some of these techniques and building [products] without the underlying concerns of private data becoming part of a larger data model.
Chirag DekateAnalyst, Gartner
However, Google makes the distinction between customer data and the foundation models, which operate on a different data model.
"That distinction is super important," Dekate said. "The reason it's important is so that enterprises can start trusting some of these techniques and building [products] without the underlying concerns of private data becoming part of a larger data model."
Google is one of many vendors seeking to gain the enterprise's trust. Most recently, SambaNova introduced its Suite for Generative AI. And other vendors are taking a similar approach to scale generative AI across enterprises, Dekate said.
Without making the distinction, many enterprise customers would have little choice but to sideline generative AI in their industries, he added.
Competing with Microsoft
To compete with other tech giants like Microsoft, Google Cloud now also has new generative AI capabilities. For example, the Vertex AI platform now includes foundation models for generating text and images.
"They have been preparing for this for a long time, and I think that the competition with Microsoft and OpenAI forced them to release it sooner," said Mike Gualtieri, an analyst at Forrester Research.
Google also introduced Generative AI App Builder, which connects conversational AI flows with pre-made search experiences and foundation models.
The tech giant said it is also expanding its AI ecosystem and specialized programs for its partners, AI-focused software providers, and startups.
Meanwhile, Google added generative AI capabilities to Workspace that support a new rewriting tool in Gmail for when users need to turn notes into a formal email or another scenario that requires communication in a specific tone or voice for the workplace.
Users can also choose the "I'm feeling lucky" option to try out a totally new voice with the tool's help.
"Adoption depends on making AI as natural as possible and meeting the needs of as many of the user base as possible," IDC analyst Wayne Kurtzman said. "Can Google's generative AI become an accepted third-party collaborator at work and home? Then will it be the preferred platform."
"There is a lot of growth to the market, and adjustments to be made by all the vendors, but this is a great start because it fills real needs," Kurtzman added.
While it's too early to say what effect generative AI capabilities embedded in office applications will have on employee productivity, they may not all be positive, according to Kashyap Kompella, founder and analyst at RPA2AI Research.
"Emails may be easier to comprehend, but their number may increase because it's easy to auto-compose one," Kompella said. "Authenticity may be in short supply when text is auto generated."
Being cautious
With all the new capabilities, the tech giant clearly wants enterprises to know that it uses responsible AI principles to guide its work surrounding generative AI, Dekate said.
By focusing on responsible AI, Google is giving enterprises the foundation they need to adopt generative AI, Dekate said.
"These are the kinds of underpinnings needed for broadcast adoption of generative AI," he said. "Trustworthy AI and also ownership of data, ownership of models."
Meanwhile, by making the new technology available initially to hand-picked customers, "they're setting expectations correctly by saying it's going to go out to testers first," Gualtieri said.
"It lets them say, 'hey we have it,' but we're going to test it first.'"
Esther Ajao is a news writer covering artificial intelligence software and systems. TechTarget CX and unified communications news writer Mary Reines contributed to this story.
