Google Opens ChatGPT Rival Bard for Testing, as AI War Heats Up

In that post, Mr. Pichai also shared a glimpse of new search engine features that will use AI to answer user queries, and said it would open up some of its AI programs to outside developers. 

Google’s new products come amid a flurry of announcements by rival

Microsoft Corp.

MSFT -0.61%

about its use of AI technologies developed by OpenAI. Microsoft said last month it is making a multiyear, multibillion-dollar investment in the San Francisco AI startup. It said it would be opening up its tools for developers to build upon, and integrating them into services such as its Bing search engine—raising the specter of a new challenge to Google Search’s market power. 

Microsoft says it is planning an event on Tuesday to announce “progress on a few exciting projects,” which are widely expected to include a chatbot-infused version of Bing.

Sam Altman,

chief executive of OpenAI, on Monday tweeted a picture of himself next to Microsoft Chief Executive

Satya Nadella,

saying “excited for the event tomorrow.”

Microsoft’s announcements have led to complaints from investors that Google hasn’t moved quickly enough to release its in-house AI tools, despite being the pioneer of some of the technologies used to build tools such as ChatGPT. Those complaints have been echoed by some of its own researchers, The Wall Street Journal has reported.

In response, Google executives have recently sped up work to review and release artificial-intelligence programs to the general public, while also assigning teams of engineers to work on new ways to integrate new developments into areas such as the core search experience, the Journal reported. 

Google executives have also suggested they haven’t been slow so much as careful with their tools, drawing an implicit contrast to competitors tools, such as ChatGPT, that can spout made-up information in response to some user queries. Google executives say they must test new tools to make sure they don’t show bias, and guard against misuse, concerns shared by many academics.

 “It’s critical that we bring experiences rooted in these models to the world in a bold and responsible way,” Mr. Pichai in his blog post on Monday. “That’s why we’re committed to developing AI responsibly.”

He added that the new external testing period for Bard will be combined with internal research to make sure that it gives responses that meet Google’s “high bar for quality, safety and groundedness in real-world information.” Mr. Pichai said in a separate internal email seen by The Wall Street Journal that Google’s entire staff will get access to Bard next week in order to give feedback as part of a “company-wide dogfood,” or tech-industry slang for testing or using one’s own products.

Google is under the spotlight of regulators in the European Union, U.S. and other parts of the globe. In the EU, policy makers are considering a new AI law that could require companies to conduct risk assessments before launching new tools 

Google says its Bard service is based on its experimental artificial-intelligence program called LaMDA, which stands for Language Model for Dialogue Applications. Google last year suspended an engineer who contended that LaMDA had become sentient—a claim roundly rejected by scientists in the field. 

When OpenAI released ChatGPT late last year, it took off as a viral sensation. 

While it was based on AI tools widely available to researchers, the breadth of its capabilities opened up the possibilities of so-called generative AI—or AI that can create content in response to short user inputs—to a wider audience of potential users. People posted the chatbot’s humorous responses to their queries, like a retelling of the Goldilocks fairy tale in the style of a police blotter, as well as complex computer software coding that it could provide when asked. 

Google offered few details Monday about how Bard will work and what kind of answers it will offer. The company said that Bard would initially use what it called a smaller model of LaMDA that uses less computing power, which will allow the company to make it more widely available. 

Sample queries for Bard include, “What new discoveries from the James Webb Space Telescope can I tell my 9-year-old about?” to which Bard replies three short bullet points of recent discoveries. Other suggested queries in the blog post include, “Plan a friend’s baby shower” and “Compare two Oscar-nominated movies.” 

In Monday’s post, Mr. Pichai also gave a glimpse of some new AI features that he said would soon be integrated into the company’s eponymous search engine. Those features aim to “distill complex information and multiple perspectives into easy to digest formats,” he said. 

The only example query the company provided Monday asks the search engine whether it is easier to learn to play piano or guitar. Google’s answer? Some say piano while others say guitar. 

More examples are expected on Wednesday, when Google has scheduled an event in Paris that will share progress on the new AI-based search tools, the internal email from Mr. Pichai said.

Mr. Pichai also said Google plans next month to start allowing outside developers to start building LaMDA’s generative language capabilities into their own applications, through a new application programming interface, or API, that allows them to query LaMDA as part of their own tools. Eventually the company says it will make a suite of tools. 

That announcement in some ways mirrors Microsoft’s announcement last month that it would start allowing outside developers to build with ChatGPT and other AI tools through its own APIs. 

Write to Sam Schechner at Sam.Schechner@wsj.com