Google opens Bard to a public waitlist

Google began letting members of the public sign up to use Bard, six weeks after announcing the tool and having it initially open only to a small group of trusted testers. The company said it was “starting to open access to Bard, an early experiment that lets you collaborate with generative AI,” and would begin rolling out access “in the U.S. and U.K. today and expanding over time to more countries and languages.” Users joined a waitlist at bard.google.com rather than receiving immediate access.

The service still ran on “a lightweight and optimized version” of LaMDA, the same conversational model Google had described at the February announcement, though the company said Bard would be “updated with newer, more capable models over time” — language that anticipated the later switch to Google’s PaLM and then Gemini model families. Google positioned Bard as a tool for tasks such as drafting text, brainstorming and explaining concepts, rather than as a direct search replacement.

The gap between announcement and public availability was itself notable. Bard had been unveiled in apparent haste, ahead of Microsoft’s Bing event and just over two months after ChatGPT’s launch, and the rollout was overshadowed almost immediately by a factual error in Google’s own promotional material that erased roughly $100 billion from Alphabet’s market value. The six-week interval before public access reflected Google’s more cautious, staged approach to releasing a conversational system compared with the near-simultaneous announcement-and-ship pattern Microsoft and OpenAI had used with Bing and ChatGPT — a caution partly forced by the criticism the February stumble had drawn, and partly consistent with Google’s historically slower path from AI research to public product before ChatGPT changed the competitive calculus.