Alphabet Inc’s Google on Tuesday began the public release of its chatbot Bard, seeking users and feedback to gain ground on Microsoft Corp in a fast-moving race on artificial intelligence technology.
Starting in the US and the UK, consumers can join a waitlist for English-language access to Bard, a program previously open to approved testers only. Google describes Bard as an experiment allowing collaboration with generative AI, a technology that relies on past data to create rather than identify content.
The release last year of ChatGPT, a chatbot from the Microsoft-backed startup OpenAI, has caused a sprint in the technology sector to put AI into more users’ hands. The hope is to reshape how people work and win business in the process.
Just last week, Google and Microsoft made a flurry of announcements on AI, two days apart. The companies are putting draft-writing technology into their word processors and other collaboration software, as well as marketing-related tools for web developers to build their own AI-based applications.
Asked whether competitive dynamics were behind Bard’s rollout, Jack Krawczyk, a senior product director, said Google was focused on users. Internal and external testers have turned to Bard for “boosting their productivity, accelerating their ideas, really fueling their curiosity,” he said.
In a demonstration of the site bard.google.com to Reuters, Krawczyk showed how the program produces blocks of text in an instant, different from how ChatGPT types out answers word by word.
Bard also included a feature showing three different versions or “drafts” of any given answer among which users could toggle, and it displayed a button stating “Google it,” should a user desire web results for a query.
Accuracy, however, is still a concern. “Bard will not always get it right,” a Google pop-up notice warned during the demo. Last month, a promotional video for Bard showed the program answering a question incorrectly, helping shave $100 billion off Alphabet’s market value.
Google highlighted a couple mistakes during this week’s demo to Reuters, for instance saying Bard wrongly claimed ferns required bright, indirect light in response to a query.
Bard also produced nine paragraphs of text when asked for four in another question. After that answer, Krawczyk clicked a thumbs-down feedback button in response.
“We know the limitations of the technology, and so we want to be very deliberate at the pace at which we roll this out,” he said.