Google has announced that its Bard AI chatbot will now have the ability to pull real-time data from other Google applications in order to provide more relevant and actionable responses. The chatbot can now access data from Google Docs, Maps, Lens, Flights, Hotels, YouTube, as well as the user’s own personal data. This expansion of capabilities is aimed at building public trust in the technology and addressing concerns about its usefulness and accuracy.
Jack Krawczyk, the product lead for Bard, explained that the chatbot was initially designed to be a creative collaborator, but feedback from users over the past six months has led to the development of more robust and innovative language models. The new version of Bard is the first to not only provide content from the web to support its answers but also to admit when it has made a mistake. This increased transparency and willingness to acknowledge errors is seen as a critical step in building trust with users.
Google is following the lead of other tech companies like OpenAI and Salesforce by integrating its AI chatbot with other apps and services. By accessing real-time capabilities from Google’s other applications, Bard is able to respond to more complex and detailed prompts. Users can now ask questions that require multiple API requests, using natural language requests. For example, a user could ask Bard to show flights and available hotels in Puerto Rico on Valentine’s Day, as well as provide a map of interesting sites near the hotel. The chatbot would then be able to provide a list of flights, hotel options, and recommendations for sightseeing activities.
In terms of transparency, Google is taking steps to improve the AI’s reasoning by linking to the sites it references and introducing a Double Check feature. The Double Check feature allows users to independently audit Bard’s responses by searching the web for supporting information. Contradictory evidence is highlighted, while heavily referenced and supported statements are highlighted as reliable.
Google is also introducing a feature called Bard Extensions, which allows users to grant the AI access to their personal data stored in Google services such as Gmail, Docs, and Drive. This enables Bard to provide specific answers about the user’s daily life, such as retrieving important dates from email conversations or summarizing discussions. It also allows users to collaborate with Bard on tasks like drafting cover letters based on their work experience.
To address privacy concerns, Google emphasizes that user content from Gmail, Docs, and Drive is not seen by human reviewers, used for advertising purposes, or used to train the Bard model. Users have control over their data and can opt in or out of the system at any time. At the moment, the service is only available to non-enterprise users in English, but Google plans to expand its offerings in the future.
Krawczyk believes that harnessing the open web is crucial in communication because it provides context. Users often seek out additional information to deepen their understanding and Bard aims to provide that trusted content through this new experience. With the integration of real-time data from various Google applications and increased transparency, Google hopes to alleviate user concerns and demonstrate the value of AI chatbots in everyday life.