Alphabet Inc’s (GOOGL.O) Google began the public release of its chatbot Bard on Tuesday, seeking users and feedback as it seeks to gain ground on Microsoft Corp (MSFT.O) in a fast-moving race in artificial intelligence technology.
Beginning in the United States and the United Kingdom, consumers can join a waiting list for English-language access to Bard, which was previously only available to approved testers. Google describes Bard as a collaboration experiment with generative AI, a technology that uses past data to create rather than identify content.
The release of ChatGPT, a chatbot from Microsoft-backed startup OpenAI, last year sparked a race in the technology sector to put AI in the hands of more users. The goal is to reshape how people work while also winning business.
Google and Microsoft both made AI announcements two days apart last week. Draft-writing technology is being integrated into word processors and other collaboration software, as well as marketing tools for web developers to build their own AI-based applications.
When asked if competitive dynamics were driving Bard’s rollout, Google senior product director Jack Krawczyk said the company was focused on users. Internal and external testers have come to Bard for help in “boosting their productivity, accelerating their ideas, and really fueling their curiosity,” he says.
Krawczyk demonstrated the site, bard.google.com, to Reuters, showing how the program produces blocks of text in an instant, as opposed to how ChatGPT types out answers word for word.
Bard also included a feature that displayed three different versions or “drafts” of any given answer, which users could toggle between, as well as a button that said “Google it” if a user wanted web results for a query.
Bard, unlike ChatGPT, is not capable of generating computer code, according to Google’s website. Google also stated that it has limited Bard’s memory of previous chat exchanges and that it is not currently using Bard for advertising, which is central to Google’s business model.
Accuracy is still an issue. “Bard will not always get it right,” a Google pop-up warning said during the demo. Last month, a promotional video showed the program incorrectly answering a question, helping to reduce Alphabet’s market value by $100 billion.
During the demonstration, Google pointed out a couple of errors to Reuters, such as how Bard incorrectly claimed ferns required bright, indirect light in response to one query.
When asked for four paragraphs of text in another, Bard produced nine. Krawczyk provided feedback by clicking a thumbs-down button after that response.
“We understand the technology’s limitations, so we want to be very deliberate in the pace at which we roll this out,” he said.
