Sep 19, 2023
Bard Unveils Enhanced Features and Google Integration
Today marks the release of Bard's latest iteration, designed to offer more specialized and versatile responses to user queries. The update comes with integrated functionality for Google's suite of apps and services. This new model aims to make multitasking more straightforward for users by pulling relevant information from various Google tools.
Introducing Bard Extensions
Available in English, Bard Extensions is a new functionality that allows users to interact with Bard in a more comprehensive manner. This feature enables Bard to access and compile data from Google services such as Gmail, Docs, Drive, Google Maps, YouTube, Google Flights, and hotels. It streamlines information retrieval even when the data is scattered across multiple applications.
For instance, someone planning a trip to the Grand Canyon can now direct Bard to collect pertinent dates from an email thread, provide real-time flight and hotel data, show directions via Google Maps, and even suggest YouTube videos regarding activities at the destination—all within a single conversational thread.
Bard emphasizes its commitment to user privacy. If users opt for Workspace extensions, their Gmail, Docs, and Drive content remains confidential, neither reviewed by humans nor used for ad targeting or model training. Users retain full control over their privacy settings and can disable extensions at their discretion.
Verification of Responses
A new "Google it" button allows users to validate Bard's responses more efficiently. Upon clicking this button, Bard scans the response and gauges whether corroborative or refuting information is available online. This feature seeks to enhance the accuracy of information provided by Bard.
Building on Shared Conversations
Another improvement allows users to expand upon existing conversations with Bard. If someone shares a conversation via a public link, the recipient can continue the dialogue or initiate new topics based on the original discussion.
Bard is extending its pre-existing features in the English language, such as image uploads via Lens and search images in responses, to over 40 languages.
This advancement is fueled by the PaLM 2 model, which incorporates state-of-the-art reinforcement learning techniques based on user feedback. The aim is to make Bard more intuitive and resourceful, capable of providing higher-quality responses across a variety of tasks and languages.