On March 7th, Google announced the addition of a new experimental text embedding model, called Gemini Embedding, through its Gemini API for applications.
 |
| Google has announced the addition of a new experimental text embedding model called Gemini Embedding. (Source: Getty Images) |
In a blog post, Google emphasized that because the experimental embedded model was trained on the Gemini model itself, it inherits Gemini's ability to understand language and nuanced context. The new model also has "incredible" generalization capabilities, delivering superior performance across many fields such as finance, science , law, and search.
Google notes that Gemini Embedding is in the testing phase, and therefore still has limitations in terms of capabilities. The company aims to continue refining and upgrading it so that a stable version can be released to a wider audience in the coming months.
Over the past period, companies including Amazon, Cohere, and OpenAI have offered embedded models through their respective APIs. Gemini Embedding is Google's first embedded model trained on the Gemini AI model.
Embedded models are algorithms trained to encapsulate information into high-density representations in multidimensional space.
Data scientists use embedded models to help machine learning models understand and reason with multidimensional data.
Source
Comment (0)