Full details of Google's internal coding AI 'GfG': Developed by training Gemini on 1 trillion tokens of internal data, reducing task time by half in some cases

Models✍ TextPublished: Source: ITmedia AI+

#Google

Key point

The paper 'Customizing an LLM for Enterprise Software Engineering' published in October by researchers belonging to U.S. Google and others is a research report summarizing a method to build a coding model specialized in internal system development by having AI learn vast amounts of company-specific development data, and its practical effects.

According to the paper 'Customizing an LLM for Enterprise Software Engineering' published in October by researchers from Google in the U.S. and others, a coding model called 'GfG' specialized in internal system development has been developed. This model was built by training Gemini on 1 trillion tokens of internal data.

The paper is a research report summarizing a method to build a coding model by having AI learn vast amounts of company-specific development data, along with its practical effects. It is reported that for some tasks, the required time was halved.

This item is a summary. The full, checkable version is at the original source:

Read the original at ITmedia AI+ ↗