Dive Brief:
- Google is adding AI assistant Gemini to its Workspace suite of enterprise tools, the company said Wednesday. Users can now access the tool in Docs, Gmail, Meet, Slides and Sheets, in step with the rebrand announced earlier this month, which unified Google’s generative AI efforts under the Gemini moniker.
- The company is rolling out a new experience as part of Gemini for Workspace’s debut, which will feature a standalone space for users to chat with the tool. Conversations are not used for advertising purposes, reviewed by humans or used to train generative AI, Google said.
- Gemini for Workspace is now available in two plans, either as a Gemini Enterprise add-on, replacing Duet AI for Workspace Enterprise, or a new Gemini Business add-on at a lower price point.
Dive Insight:
Google has set a brisk pace for generative AI-related announcements in 2024, with the company’s Gemini rebrand at the beginning of the month followed by the launch of Gemini 1.5 a week later.
As part of Google’s broader generative AI strategy, the company released another family of foundational models, called Gemma, Wednesday.
Alongside the models, Google unveiled a responsible generative AI toolkit "for creating safer AI applications," according to the blog post. The models weights are available in two sizes: Gemma 2B and Gemma 7B.
Open-weight models, which allow developers to build upon and adapt previous work, are currently the subject of a request for public comment from the Department of Commerce’s National Telecommunications and Information Administration. The findings, along with listening sessions and direct outreach, will inform a report offering potential policy recommendations and outlining risks and benefits.
Open models give enterprises the ability to explore and craft a generative AI strategy without breaking the bank, according to Chirag Dekate, VP analyst at Gartner.
“For example, with Gemma models alongside Meta’s Llama 2 and others, you can download, access, experiment, test and validate for free,” Dekate said.
But for enterprises considering whether to use open models for external-facing use cases, Dekate urged caution.
“Trying to create externally facing applications using open weight models could potentially expose enterprises to outsized litigation risks,” Dekate said.
