SCITECH

Microsoft to limit length of Bing chatbot conversations

OpenAI CEO Sam Altman, left, and Microsoft Chief Technology Officer Kevin Scott at a media event introducing their Bing search engine using AI, at the Microsoft campus in Redmond, Wash. on February 7, 2023. (Ruth Fremson/The New York Times)
 
OpenAI CEO Sam Altman, left, and Microsoft Chief Technology Officer Kevin Scott at a media event introducing their Bing search engine using AI, at the Microsoft campus in Redmond, Wash. on February 7, 2023. (Ruth Fremson/The New York Times)

Microsoft will start limiting conversations with the new chatbot in its Bing search engine to five questions per session and 50 questions per day, the company said on Friday.

Microsoft released a new version of Bing, which combines the search engine with artificial intelligence technology built by OpenAI, a San Francisco startup, with fanfare at an event on its Redmond, Washington, campus less than two weeks ago.

A number of other big tech companies, including Google, are working on similar services. But Microsoft has moved quickly to gain a technology advantage on its competitors, and the company has promised that AI will eventually be built into a wide range of its products.

Microsoft expected its chatbot to sometimes respond inaccurately, and it built in measures to protect against people who try to make the chatbot behave strangely or say harmful things. Still, early users who had open-ended, personal conversations with the chatbot found its responses unusual — and sometimes creepy.

Now people will be prompted to begin a new session after they ask five questions and the chatbot answers five times.

“Very long chat sessions can confuse the underlying chat model,” Microsoft said on Friday.

On Wednesday, the company wrote in a blog post that it “didn’t fully envision” people using the chatbot “for more general discovery of the world, and for social entertainment.” The chatbot became repetitive and, sometimes, testy in long conversations, it said.

Microsoft said its data showed that about 1% of conversations with the chatbot had more than 50 messages. It said it would consider increasing the limits on questions in the future. The company is also looking at adding tools to give users more control over the tone of the chatbot.

This article originally appeared in The New York Times.