These are Google's two launch posts for Bard: "An important next step on our AI journey," published February 6, 2023 under the byline of Sundar Pichai, CEO of Google and Alphabet; and "Try Bard and share your feedback," published March 21, 2023 under the bylines of Sissie Hsiao, VP Product, and Eli Collins, VP Research. They are classified here as foundational on the same basis as ChatGPT: Optimizing Language Models for Dialogue (OpenAI, November 30, 2022) and Introducing Claude (Anthropic, March 14, 2023): they are the primary record of the competitive response that followed the ChatGPT release, and they state what Google claimed for the product and conceded about it at launch. See Bard (Google).
The February 6 announcement
Pichai frames the release within a longer institutional narrative, stating that Google "re-oriented the company around AI six years ago" and that "our Transformer research project and our field-defining paper in 2017, as well as our important advances in diffusion models, are now the basis of many of the generative AI applications you're starting to see today." The post asserts that "the scale of the largest AI computations is doubling every six months, far outpacing Moore's Law"; see Scaling Laws.
Bard is introduced as "an experimental conversational AI service, powered by LaMDA," opened "to trusted testers ahead of making it more widely available to the public in the coming weeks." The stated design is to "combine the breadth of the world's knowledge with the power, intelligence and creativity of our large language models," drawing "on information from the web to provide fresh, high-quality responses."
The initial release used a reduced model: "We're releasing it initially with our lightweight model version of LaMDA. This much smaller model requires significantly less computing power, enabling us to scale to more users, allowing for more feedback." Google states it will combine external feedback with internal testing "to make sure Bard's responses meet a high bar for quality, safety and groundedness in real-world information."
Two further commitments appear in the same post. On Search, Google states that "AI-powered features in Search that distill complex information and multiple perspectives into easy-to-digest formats" would "begin rolling out on Google Search soon," positioning the change as serving queries where "there's no one right answer." On developer access, it announces that "next month, we'll start onboarding individual developers, creators and enterprises so they can try our Generative Language API, initially powered by LaMDA with a range of models to follow." The post names Google Cloud partnerships with Cohere, C3.ai and Anthropic, the last "just announced last week."
The closing section states that Google "was one of the first companies to publish a set of AI Principles" in 2018 and describes continued work with governments, external organizations and communities on standards and best practices.
The March 21 general-access post
Hsiao and Collins announce that "today we're starting to open access to Bard, an early experiment that lets you collaborate with generative AI," beginning "with the U.S. and the U.K." and expanding "to more countries and languages over time." The stated rationale for the widened release is evaluative: "We've learned a lot so far by testing Bard, and the next critical step in improving it is to get feedback from more people."
The post describes Bard as "powered by a research large language model (LLM), specifically a lightweight and optimized version of LaMDA," which "will be updated with newer, more capable models over time." Its explanation of the mechanism is deliberately plain: "You can think of an LLM as a prediction engine. When given a prompt, it generates a response by selecting, one word at a time, from words that are likely to come next. Picking the most probable choice every time wouldn't lead to very creative responses, so there's some flexibility factored in."
The limitations are stated by Google rather than inferred. Because LLMs "learn from a wide range of information that reflects real-world biases and stereotypes, those sometimes show up in their outputs," and "they can provide inaccurate, misleading or false information while presenting it confidently." The post illustrates this with an error in its own screenshot: asked for easy indoor plants, "Bard convincingly presented ideas…but it got some things wrong, like the scientific name for the ZZ plant," which the caption corrects to Zamioculcas zamiifolia rather than the "Zamioculcas zamioculcas" Bard produced. See Sycophancy and Hallucination.
As a mitigation the post describes offering the user "the choice of a few different drafts of its response so you can pick the best starting point," with the option to ask follow-up questions or have Bard try again.
Provenance
Both posts were retrieved August 1, 2026 by direct HTTPS fetch (HTTP 200) and converted to text with a local extractor; Firecrawl was out of credits during this cycle. Both canonical blog.google/technology/ai/ URLs now redirect within blog.google to /innovation-and-ai/ paths; the February 06, 2023 and March 21, 2023 datelines and the bylines are preserved in the served pages. Social-share link lists were removed from the capture as navigation chrome. Image captions are quoted from the served alt text and figure captions.
Relationships
- related: Google DeepMind — the publisher's AI organization.
- related: Bard (Google) — the product the posts announce.
- related: ChatGPT: Optimizing Language Models for Dialogue (OpenAI, November 30, 2022) — the release these posts respond to competitively.
- related: Introducing Claude (Anthropic, March 14, 2023) — the third major assistant launch of the same period.
- related: Sycophancy and Hallucination — the failure the March post concedes and illustrates.
- depends-on: Scaling Laws — the compute-doubling claim in the February post.