Google Inc. began publicly releasing its chatbot Bard to attract users and seek feedback in order to compete with Microsoft Corp. in the rapidly evolving race for artificial intelligence technology.
Google said public testing will begin in the U.S. and the U.K., and consumers can join a waiting list for access to “Bard” in English. Previously, testing of Bard was only available to approved testers. Google describes Bard as an experiment that allows collaboration with generative artificial intelligence, a technology that relies on past data to create, rather than identify, content.
Asked if there was competition behind the launch of Bard, Jack Krawczyk, Google’s senior product director, said Google’s focus was on users. He said internal and external testers have turned to Bard to “improve their productivity, accelerate their ideas and really spark their curiosity.
“Bard” begins public beta
In a demo, Krauchick showed how Bud can generate chunks of text in an instant, unlike the way ChatGPT typed in answers verbatim. “Bard” also offers a feature that displays three different versions, or “drafts,” of any given answer, which users can switch between. It also displays a button that says “Google it” if the user wants to get web query results. Google says on its website that, unlike ChatGPT, “Bud” isn’t very good at generating computer code. Google also said it limited “Bud”‘s memory of past chats and is not currently using “Bud” for advertising. Advertising is at the heart of Google’s business model.
Accuracy is still an issue
Accuracy remains an issue, though. “Bard isn’t always right,” Google’s pop-up prompt warned during the demo. Last month, a promotional video showed the program incorrectly answering a question, wiping $100 billion off the market value of Google parent Alphabet.
Google made several mistakes during its presentation on Tuesday. For example, in response to a question, “Bud” falsely claimed that ferns needed bright indirect light to grow. But in fact, ferns have strong shade tolerance.
Also, when asked to write a four-paragraph text, “Bud” wrote nine. When “Bud” finished answering the question, Krawchick gave the answer a thumbs-up as feedback.
“We know the limitations of this technology, so we want to be very careful about how quickly we roll it out,” he said.