Google announced about the experiment on embedding a large language training model in Chrome. For access to the model from Web applications and browser additions, the API permpt was proposed. that allows you to send requests in a natural language, by analogy with chatbots. It is assumed that a large language model built into the browser will simplify the execution of AI-uses in Web applications and will allow you not to take care of the installation and management of language models.
The conducted experiment is used by the model of gemini nano , the most compact from the family gemini . It is also possible to establish expert models that expand the basic model with additional knowledge that may be required to solve specific problems, as well as for the implementation of skills, such as machine translation and summarizing. Models are performed locally on the user system without contacting external services.
used to perform the Runtime model automatically involves the GPU and NPU system to accelerate working with the model or switches to execution of the model using CPU. Of the advantages of the model on the user’s system, it is noted to maintain the confidentiality of the processed data, the ability to continue to work in Offline-mode in the absence of a network connection or if problems with the quality of communication, reducing delays when sending requests, excluding dependence on external services.
The developed for interaction with the API Prompt model allows you to perform not only the simplest single requests in the natural language, but also organize the attraction of the model to processing and classifying data taking into account the context, take into account the requests and data previously sent in the session, as well as use the model to choose Optimal options (for example, you can ask you to choose an icon from the Emoji list for a specific comment on the site). Additionally, it is planned to develop the API for use in the field of creating content and solving such problems as rewriting in other words, correction and correction of grammar.
In general, two types of API – TASK and Exploratory develops for interaction with the built -in AI model. The first provides access to opportunities to solve certain problems, such as the translation of the text from one language into another (API translation