Bard is a tool powered by a research large language model (LLM), created as a prediction engine that generates responses by selecting the next word from those that are likely to follow a prompt. As users continue to use it, LLMs become better at predicting helpful responses. However, due to the real-world biases and stereotypes reflected in the information they learn from, their outputs can be inaccurate, misleading or false, and Bard can make mistakes.

source update: Sign up to try Bard from Google