According to reports, Google today made a series of updates to its generative artificial intelligence robot Bard, which has the ability to fact-check answers and analyze users’ Google personal data to catch up with competitor ChatGPT. Last year, Microsoft-backed artificial intelligence startup OpenAI triggered an AI technology arms race in the technology industry after it released the generative chatbot ChatGPT. At that time, ChatGPT became the fastest-growing consumer application of all time and is now one of the top 30 websites in the world.
But Google's Bard didn't take off at the same pace. According to data from website analysis company Similarweb, in August this year, Bard had 183 million visits, which was only 13% of ChatGPT’s visits.
In order to make progress in the rapidly developing field of artificial intelligence, Google has upgraded Bard again today and launched three new features. The first is Bard Extensions, which allows users to import data from other Google products.
For example, users can ask Bard to search their files in Google Drive or provide a summary of the user's Gmail inbox. Google senior product director Jack Krawczyk said that currently Bard users can only get information from Google apps, but Google is working with outside companies to connect their apps to Bard in the future.
Another new feature in Bard attempts to alleviate a concern about generative AI: AI illusion. In the past few months, artificial intelligence large-scale language models represented by ChatGPT have attracted worldwide attention. They are surprisingly "smart", but they occasionally say some fictitious people or things, and they maintain consistent confidence. This phenomenon is known as the "AI hallucination."
With this update, users will be able to see which parts of Bard's answers are different from Google search results and which are the same. "We will present Bard in a way that is not particularly assertive, with the goal of building user trust in generative AI by taking on Bard's responsibilities," Krawczyk explained.
The third new feature allows users to invite others to participate in Bard conversations. Just send a public link and the other party can join the Bard conversation.
Google also specifically emphasizes that if users associate Bard with personal data, this information will not be seen by reviewers, will not be used by Bard for ad targeting, and will not be used in Bard models.
Krawczyk concluded: "Based on your feedback, we applied state-of-the-art reinforcement learning techniques to train the Bard model to make it more intuitive and imaginative. Now, whether you want to collaborate on something creative, start in one language and continue using one of more than 40 languages, or seek in-depth coding help, Bard can respond with higher quality and accuracy."