This post was originally published on this site
https://content.fortune.com/wp-content/uploads/2023/07/GettyImages-1532795646-e1690541602590.jpg?w=2048
Google’s new ‘Bard’ chatbot, introduced in response to the A.I. frenzy sparked by OpenAI’s ChatGPT, boasts impressive capabilities including answering questions in seconds and providing audio responses. However, a top Google U.K. executive cautions that Bard can struggle with providing trustworthy information.
“We know people count on Google for accurate information and we’re encouraging people to actually use Google as the search engine” when it comes to cross-checking Bard’s answers, Debbie Weinstein, managing director of Google U.K., said in an interview with BBC aired Friday.
She added that Bard was “not really the place that you go to search for specific information”.
Bard’s homepage acknowledges its “limitations” and concedes “and occasional imperfections”it won’t always get it right,” but omits Weinstein’s advice to cross-check all outcomes using a traditional search engine.
Can you trust any chatbot?
Bard, like other A.I. chatbots, can “hallucinate” or return inaccurate information when asked a question.
Fabricated details and inaccurate answers are a common issue across the industry, even for the newest large language models like OpenAI’s GPT-4.
“No one in the field has yet solved the hallucination problem,” Sundar Pichai, CEO of Google parent Alphabet, told CBS‘s “60 Minutes” program in April.
Bard made a costly error during its demo in February, when a misstated fact caused Alphabet to lose $100 billion in market value.
The bot also made several mistakes when answering basic SAT questions, in response to prompts from Fortune. At the time, a Google spokesperson warned that Bard “may be inaccurate, so double-check information in Bard’s responses.”
One way that people could make Bard better is by giving feedback on whether the bot helped in a given situation, Weinstein suggested.
Where does this leave Google in the A.I. race?
The search giant has been a pioneer in A.I. research, but now faces stiff competition from other tech companies like Microsoft.
In May, a top Google engineer claimed the company had “no secret sauce” and could be falling behind in the A.I. arms race as a result, according to Bloomberg.
Other experts have cautioned against the dangers of scaling A.I. without knowing how to regulate it—including former Google CEO Eric Schmidt.
“We, collectively, in our industry face a reckoning of, how do we want to make sure this stuff doesn’t harm but just helps?” Schmidt said in an interview with ABC News in April.
“Godfather of A.I.” Geoffrey Hinton, who worked at Google and helped lay the groundwork for today’s A.I. technologies, expressed similar concerns in May, saying he quit the tech giant so he could speak freely about “the dangers of A.I.”
Google also faces the steep challenge of balancing its A.I. ambitions with preserving its business in search and advertising—the tech giant’s most lucrative source of revenue.
Google has been contacted by Fortune for comment.
