Google Quietly Launches Internal AI Model 'Goose' to Speed Up Engineer Coding

Google is not only committed to applying AI across its various products but also leverages this technology internally to enhance overall operational efficiency. According to a Business Insider report citing internal Google documents, Google is quietly launching a large language model (LLM) named 'Goose.' This marks the first general-purpose LLM approved for coding within Google, exclusively for internal employees to assist in new product development.

Goose is described as a 'descendant of Gemini,' a brand-new large language model trained on '25 years of Google's engineering expertise'! It was jointly developed by Google Brain, DeepMind, and Google's internal infrastructure teams. Consequently, it not only integrates Gemini technology but also possesses a deep understanding of Google's proprietary technologies, with capabilities that likely far exceed the development tools currently provided by Google to its employees.

Goose can answer questions about Google-specific technologies, write code using internal technology stacks, and supports novel development features such as making code changes with natural language prompts. This initiative demonstrates Google's proactive use of internal AI resources to drive innovation and enhance the company's internal technical capabilities.

Another notable feature of Goose is its context length of 28,000 tokens. This length is highly valuable for handling large and complex product development tasks, especially when hundreds of thousands of lines of code need to be written. Context length refers to the amount of text, numbers, and other information an AI model can consider when processing a query. With its more comprehensive understanding, Goose can tackle a wide range of extensive and intricate development tasks.

For Google employees, Goose is likely to become a very practical tool because its design focuses on 'new engineering-related functionalities.' This indicates that Google is continuously enhancing its internal tools to support more efficient and innovative product development processes.