Google today is unveiling several updates to Bard, its conversational AI chatbot, designed to make it more intuitive, imaginative, and responsive. The newest capabilities allow Bard to integrate directly with Google apps and services to provide personalized and actionable information. It underscores the platform's growing role in serving both individual and enterprise needs.
The most significant addition is Bard Extensions, launching first in English. This feature lets Bard automatically pull relevant data from Google tools like Gmail, Docs, Drive, Maps, YouTube, and Google Flights and Hotels. For example, Bard can now summarize trip details from emails and integrate real-time travel information into a single conversation.
Importantly, Google says it has prioritized user privacy and trust with Extensions. As explained by Google's Jack Krawczyk, Extensions will only activate when a user expressly opts in, and permissions can be revoked anytime. Personal data from Gmail, Docs, and Drive is not used for Bard's reinforcement learning or exposed to human reviewers. Bard can only access specific content on a per-prompt basis when directed. While this may limit capabilities initially, Google believes focusing on transparency, user control, and building trust is critical.
Additionally, Bard's "Google it" feature has been improved to help users double-check Bard's responses against information on the wider web. Highlighted statements can be clicked to view supporting or contradicting content from Google Search. As conversational AIs become more intricate, ensuring the veracity of their outputs grows ever crucial.
Building on the theme of collaboration, Bard now enables users to continue conversations shared with them through public link sharing. This works similar to ChatGPT's shared links feature and allows individuals to expand upon previously initiated discussions, adding another layer to collective creativity.
Other updates include expanding certain features to over 40 languages, and upgrading to a more advanced underlying AI model.
Google says this newest model draws on reinforcement learning techniques to deliver more intuitive, imaginative responses with greater accuracy. Early users have reported improvements in areas like multilingual conversations, creative brainstorming, and coding assistance.
The enhancements reflect Google's iterative approach of rapidly incorporating user feedback into Bard. With today's updates, the company aims to further distinguish itself from ChatGPT and position Bard as a helpful collaborator. Users can explore the new capabilities at bard.google.com.