Google’s Bard Chatbot Opens to the Public

Google is ready to open the Bard floodgates, at least to English speakers around the world. After two months of testing, access to the AI-powered chatbot no longer is gated by a waitlist.

Google announced the move at its Google I/O developer conference on Wednesday, a week after Microsoft removed the waitlist for its competing Bing chatbot. In addition to opening Bard up to people in 180 English-speaking countries and territories, it added Japanese and Korean chat abilities as part of a 40-language expansion plan.

Chatbots have been around since Eliza from the 1960s, but new artificial intelligence technologies like large language models and generative AI have made them profoundly more useful. LLMs are trained to spot patterns across vast collections of text from the internet, books and other sources, and generative AI can use that analysis to respond to text prompts with human-sounding written conversation.

Read more: Google Reveals Its AI-Powered Search Engine To Keep Up With Competition

It’s a revolution in what computers can offer, combining a wealth of information with a natural interface. Chatbots have shown skills in writing poetry, answering philosophy questions, constructing software, passing exams and offering tax advice.

But modern chatbots also are prone to making up data, and their backers are working hard to keep them from contributing to problems like abuse, misinformation, hacking and sexual harassment. And those are just near-term worries. Today’s AI is powerful enough to trigger fears about wiping out white-collar jobs and undermining civilization.

Despite pioneering some of the technology behind new chatbots, Google was somewhat late to the party. OpenAI’s ChatGPT website sparked a surge in interest. Microsoft, an OpenAI investor, built the underlying GPT-4 technology into its own Bing search engine.

“Google was hesitant to productize this,” said John Hennessy, a Stanford University professor and board member of Google’s parent company, Alphabet, in an April talk.

But now Google is working to catch up with what Bard product leader Jack Krawczyk calls a “bold and responsible approach” intended to balance progress with caution. Google could have expanded to 40 languages now, but limited it to Japanese and Korean to proceed more carefully, he said.

Aligning Bard with human values

“This is part of our commitment to responsibility and alignment and understanding the limitations that we know large language models have,” Krawczyk said. Alignment refers to the principle of making sure AI behavior is aligned with human interests.

In addition to making Bard public, it added options to export Bard chat data to the Google Docs word processing tool and the Google Sheets spreadsheet.

One tricky part of AI chatbots is figuring out where they got their information. That opacity makes it hard to verify facts, attribute information to appropriate sources and generally understand why a chatbot offered the results it did.

Better Bard citations of source material

Google hopes to help with this problem with an improvement coming soon, initially with responses involving programming code.

“Starting next week, we’re going to make code citations even more precise by showing you the specific blocks of code that are being sourced along with any relevant licensing information,” Krawczyk said. “This applies to citing narrative content from across the web as well.”

Also in the works is the ability to handle images, both as the prompt you give to Bard and as part of its responses. And Google will offer an interface that lets developers marry Bard with Google apps and third-party apps. One of the first: Adobe’s Firefly suite of generative AI tools.

Editors’ note: CNET is using an AI engine to create some personal finance explainers that are edited and fact-checked by our editors. For more, see this post.