“Sophia” an artificially intelligent (AI) human-like robot developed by Hong Kong-based humanoid robotics company Hanson Robotics is pictured during the “AI for Good” Global Summit hosted at the International Telecommunication Union (ITU) on June 7, 2017, in Geneva.
The meeting aim to provide a neutral platform for government officials, UN agencies, NGO’s, industry leaders, and AI experts to discuss the ethical, technical, societal and policy issues related to AI.
The company on Tuesday announced a series of updates to Bard that will give the chatbot access to Google’s full suite of tools — including YouTube, Google Drive, Google Flights and others — to assist users in a wider variety of tasks. Users will be able, for example, to ask Bard to plan an upcoming trip, complete with real flight options. Or a user could ask the tool to summarize meeting notes made in a recent Google Drive document.
The connections to Google’s other services are just some of the improvements to Bard coming Tuesday. Other updates include the ability to communicate with the chatbot in multiple languages, new fact-checking capabilities and a broad update to the large language model that the tool is built on.
The new features mark the biggest update to Google’s Bard in the six months since it was widely released to the public.
The update comes as Google and other tech giants, including Microsoft and ChatGPT maker OpenAI, race to roll out increasingly sophisticated consumer-facing AI technologies, and to convince users that such tools are more than just a gimmick. Google — which earlier this year reportedly issued an internal “code red” after OpenAI beat it to the release of its AI chatbot — is now flexing the power of its other, widely used software programs that can make Bard more useful.
The Google logo is seen on on the company’s European headquarters in Dublin. Ireland
The most popular search engine in the world is about to look different
“These services in conjunction with one another are very, very powerful,” Sissie Hsiao, general manager for Google Assistant and Bard, told CNN ahead of the launch. “Bringing all the power of these tools together will save people time — in 20 seconds, in minutes, you can do something that would have taken maybe an hour or more.”
Previously, Bard had been able to help with tasks like writing essay drafts or planning a friend’s baby shower based on Google’s large language model, an AI algorithm trained on vast troves of data. But now, Bard will draw on information from Google’s various other services, too. With the new extensions, Bard will now pull information from YouTube, Google Maps, Flights and Hotels by default.
That will allow users to ask Bard things like”Give me a template for how to write a best man speech and show me YouTube videos about them for inspiration,” or for trip suggestions, complete with driving directions, according to Google. Bard users can opt to disable these extensions at any time.
Users can also opt in to link their Gmail, Docs and Google Drive to Bard so the tool can help them analyze and manage their personal information. The tool could, for example, help with a query like: “Find the most recent lease agreement from my Drive and check how much the security deposit was,” Google said.
The company said that users’ personal Google Workspace information will not be used to train Bard or for targeted advertising purposes, and that users can withdraw their permission for the tool to access their information at any time.
“This is the first step in a fundamentally new capability for Bard – the ability to talk to other apps and services to provide more helpful responses,” Google said of the extensions tool. It added that, “this is a very young area of AI,” that it will continue to improve based on user feedback.
Bard is also launching a “double check” button that will allow users to evaluate the accuracy of its responses. When a user clicks the button, certain segments of Bard’s response will be highlighted to show where Google Search results either confirm or differ from what the chatbot said. The double check feature is designed to counter a common AI issue called “hallucinations,” where an AI tool confidently makes a statement that sounds real, but isn’t actually based in fact.
“We’re constantly working on reducing those hallucinations in Bard,” Hsiao said. But in the meantime, the company wanted to create a way to address them. “You can kind of think of it as spell check, but double checking the facts.”
Bard will now also allow one user to share a conversation with the chatbot with another person, who can then expand on the chat themselves.
It’s still early days for Bard, which launched in March as an “experiment” and still notes on its website that the tool “may display inaccurate or offensive information that doesn’t represent Google’s views.” But this latest update offers a glimpse at how Google may ultimately seek to incorporate generative AI into its various services.