Microsoft bakes ChatGPT-like tech into search engine Bing

REDMOND, Wash. (AP) — Microsoft is fusing ChatGPT-like technology into its search engine Bing, transforming an internet service that now trails far behind Google into a new way of communicating with artificial intelligence.

The revamping of Microsoft’s second-place search engine could give the software giant a head start against other tech companies in capitalizing on the worldwide excitement surrounding ChatGPT, a tool that’s awakened millions of people to the possibilities of the latest AI technology.

Along with adding it to Bing, Microsoft is also integrating the chatbot technology into its Edge browser. Microsoft announced the new technology at an event Tuesday at its headquarters in Redmond, Washington.

“Think of it as faster, more accurate, more powerful” than ChatGPT, built with technology from ChatGPT-maker OpenAI but tuned for search queries, said Yusuf Mehdi, a Microsoft executive who leads its consumer division, in an interview.

A public preview of the new Bing launched Tuesday for desktop users who sign up for it, but Mehdi said the technology will scale to millions of users in coming weeks and will eventually come to the smartphone apps for Bing and Edge. For now, everyone can try a limited number of queries, he said.

The strengthening partnership with OpenAI has been years in the making, starting with a $1 billion investment from Microsoft in 2019 that led to the development of a powerful supercomputer specifically built to train the San Francisco startup’s AI models.

While it’s not always factual or logical, ChatGPT’s comes from having ingested a huge trove of digitized books, Wikipedia entries, instruction manuals, newspapers and other online writings.

Microsoft Corp. CEO Satya Nadella said Tuesday that new AI advances are "going to reshape every software category we know," including search, much like earlier innovations in personal computers and cloud computing. He said it is important to develop AI "with human preferences and societal norms and you’re not going to do that in a lab. You have to do that out in the world.”

The shift to making search engines more conversational — able to confidently answer questions rather than offering links to other websites — could change the advertising-fueled search business, but also poses risks if the AI systems don’t get their facts right. Their opaqueness also makes it hard to source back to the original and texts they’ve effectively memorized, though the new Bing includes annotations that reference the source data.

“Bing is powered by AI, so surprises and mistakes are possible,” is a message that appears at the bottom of the preview version of Bing's new homepage. “Make sure to check the facts.”

As an example of how it works, Mehdi asked the new Bing to compare the most influential Mexican painters and it provided typical search results, but also, on the right side of the page, compiled a fact box summarizing details about Diego Rivera, Frida Kahlo and Jose Clemente Orozco. In another example, he quizzed it on 1990s-era rap, showing its ability to distinguish between the song “Jump” by Kris Kross and “Jump Around” by House of Pain. And he used it to show how it could plan a vacation or help with shopping.

Gartner analyst Jason Wong said new technological advancements will mitigate what led to Microsoft's disastrous 2016 launch of the experimental chatbot Tay, which users trained to spout racist and sexist remarks. But Wong said “reputational risks will still be at the forefront” for Microsoft if Bing produces answers with low accuracy or so-called AI “hallucinations” that mix and conflate data.

Google has been cautious about such moves. But in response to pressure over ChatGPT’s popularity, Google CEO Sundar Pichai on Monday that will be available exclusively to a group of “trusted testers” before being widely released later this year.

Wong said Google was caught off-guard with the success of ChatGPT but still has the advantage over Microsoft in consumer-facing technology, while Microsoft has the edge in selling its products to businesses.

Chinese tech giant Baidu also this week announced a similar search chatbot coming later this year, according to Chinese media. Other tech rivals such as Facebook parent Meta and Amazon have been researching similar technology, but Microsoft’s latest moves aim to position it at the center of the ChatGPT zeitgeist.

Microsoft disclosed in January as it looks to fuse the technology behind ChatGPT, the image-generator DALL-E and other OpenAI innovations into an array of Microsoft products tied to its cloud computing platform and its Office suite of workplace products like email and spreadsheets.

The most surprising might be the integration with Bing, which is the second-place search engine in many markets but has never come close to challenging Google’s dominant position.

Bing launched in 2009 as a rebranding of Microsoft’s earlier search engines and was run for a time by Nadella, years before he took over as CEO. Its significance was boosted when Yahoo and Microsoft signed a deal for Bing to power Yahoo’s search engine, giving Microsoft access to Yahoo’s greater search share. Similar deals infused Bing into the search features for devices made by other companies, though users wouldn’t necessarily know that Microsoft was powering their searches.

By making it a destination for ChatGPT-like conversations, Microsoft could invite more users to give Bing a try, though the new version so far is limited to desktops and doesn't yet have an interface for smartphones — where most people now access the internet.

On the surface, at least, a Bing integration seems far different from what OpenAI has in mind for its technology. Appearing at Microsoft's event, OpenAI CEO Sam Altman said the "the new Bing experience looks fantastic” and is based in part on learnings from its GPT line of large language models. He said a key reason for his startup's Microsoft partnership is to help get OpenAI technology “into the hands of millions of people.”

OpenAI has long voiced an ambitious vision for safely guiding what’s known as AGI, or artificial general intelligence, a not-yet-realized concept that harkens back to ideas from science fiction about human-like machines. OpenAI’s website describes AGI as “highly autonomous systems that outperform humans at most economically valuable work.”

OpenAI started out as a nonprofit research laboratory when it launched in December 2015 with backing from Tesla CEO Elon Musk and others. Its stated aims were to “advance digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return.”

That changed in 2018 when it incorporated a for-profit business Open AI LP, and shifted nearly all its staff into the business, not long after releasing its first generation of the GPT model for generating human-like paragraphs of readable text.

OpenAI’s other products include the image-generator DALL-E, first released in 2021, the computer programming assistant Codex and the speech recognition tool Whisper.