Google has officially announced the public launch of its latest product, Bard, which is designed to compete with OpenAI’s ChatGPT. Bard is an experimental tool that allows users to collaborate with generative artificial intelligence (AI) via a chatbot feature. The aim of the tool is to increase productivity, accelerate idea generation, and fuel curiosity. Users can ask Bard for tips on how to achieve their goals, explanations of complex topics, and more.

Although Bard is now available to the public, Sundar Pichai, the CEO of Google, has informed the company’s employees in an internal email that “things will go wrong.” However, he has emphasized the importance of user feedback, which is crucial for improving the product and its underlying technology. Google has also been transparent about the experimental nature of the tool, stating that it is still in beta.

Bard is powered by a large language model (LLM), which is a lightweight and optimized version of LaMDA. Google has shared that the LLM will be updated with newer, more capable models over time. The technology works by generating a response to a prompt, one word at a time, and selecting words that are likely to come next. The aim is to provide flexible responses that are not just predictable but also creative and helpful.

Like any technology, LLMs have their limitations. They can learn from biased, stereotypical, or outdated information, which can sometimes show up in their outputs. Bard offers users the option to choose from several different drafts of its responses so that they can pick the best starting point.

Bard is designed to complement Google Search, providing a direct interface to an LLM. Users can easily visit Search to check its responses or explore sources across the web. Google is also planning to integrate LLMs more deeply into Search in the future.

Google has made significant progress with over 80,000 of its employees testing Bard. However, the company believes that the next step is to turn to the public while maintaining a focus on quality and safety through “built-in guardrails.” These guardrails cap the number of exchanges in a dialogue to keep the interactions as helpful and on-topic as possible.

As development continues, Google aims to add additional capabilities, which may include coding, additional language support, and multimodal experiences.