Google wants AI chatbot Bard to help it reach billions of users

FILE PHOTO: FILE PHOTO: A sign is pictured outside a Google office near the company's headquarters in Mountain View, California

By Jeffrey Dastin

NEW YORK (Reuters) -Google's experimental chatbot Bard is a path to developing another product with two billion users, a director said on Thursday at the Reuters NEXT conference in New York.

Bard, which lets consumers brainstorm and fetch information with the help of new artificial intelligence, is laying the ground work for Google to attract still more customers, its Product Lead Jack Krawczyk said in an interview.

Among the opportunities is the company's plan to liven its timer-setting, command-fulfilling Google Assistant with Bard's human-directed suggestions. Connecting these products, first via mobile devices in the coming months, will introduce AI to more people, said Krawczyk.

"We think that opens a completely new path," he said.

Google's outlook underscores ambitions for AI at its parent Alphabet, which to date has six products each attracting billions of users, among them its search engine and YouTube.

It also reflects growing competition. Amazon.com has vowed to upgrade its Alexa aide with similar generative AI, while OpenAI recently added voice commands to ChatGPT, along with other agent-like capabilities.

The Google Assistant lives on more than one billion devices, according to company data.

The way consumers gather information may be evolving. Bard's web traffic grew 2% in October to 8.7 million, though chief rival ChatGPT grew at a faster pace, Bank of America analysts said Thursday citing Similar Web data. Google Search traffic fell 0.4%, the analysts' note said.

Krawczyk said his mandate is to improve Bard's helpfulness rather than sweat chances for monetization like a subscription model or ads, opportunities that may present themselves. Retention has varied over time with the introduction of faster and double-checked responses, he said.

But Bard has faced challenges. It reportedly invented non-existent messages when a user asked it to analyze content in his Gmail inbox, an example of AI's known tendency to "hallucinate" when asked for facts.

On Wednesday, Bard also stumbled after people wanted to use it in "record numbers," Krawczyk said. OpenAI was reporting a major outage in ChatGPT on the same day.

"It appeared that many language-model products were down around the same time," said Krawczyk, who did not name rivals specifically. "It was interesting to see behavior patterns shifting across the internet."

He said, "We're still learning how to most effectively accommodate large swaths of people," adding it was a challenge "over a span of a couple of minutes." He and OpenAI each said they have resolved their respective issues.

The surge reminded Krawczyk of what Google Search faced upon news of pop star Michael Jackson's death in 2009, he said.

(Reporting By Jeffrey Dastin in New York; Editing by Daniel Wallis)