Gemini, Google’s AI, is now available in Spanish for Drive, Gmail, Documents and Sheets

Google Drive:
Integration with Drive allows you to synthesize information from multiple files, highlight key points in long documents, and find specific data without having to manually explore each file.

Spreadsheets:
Gemini helps organize and analyze data, automatically generate formulas, and present clear summaries of tables and graphs.

Easy access and configuration

To use the side panel in Spanish, users only need to set the preferred language in their Google account. Once this setting is made, the icon “Ask Gemini” It will automatically appear in the interface of the mentioned applications. The addition of Gemini in Spanish represents a significant advance for Spanish-speaking users looking to maximize their productivity with the help of artificial intelligence.

1 hour of video, 11 hours of audio and 700,000 words

Demis Hassabis, CEO of Google DeepMind, which developed the new model, compares its enormous data input capacity to a person’s working memory, something he explored years ago as a neuroscientist. “The nice thing about these basic capabilities is that they unlock sort of ancillary things that the model can do,” he says.

Google claims that Gemini Pro 1.5 can ingest and make sense of one hour of video, 11 hours of audio, 700,000 words or 30,000 lines of code at a time, several times more than other AI models, including OpenAI’s GPT-4which ChatGPT works with. The company has not revealed the technical details of this feat. Hassabis maintains that one of the uses of models capable of handling large amounts of text, tested by Google DeepMind researchers, is to identify important points in Discord discussions with thousands of messages.

Gemini Pro 1.5 is also more capableat least for its size, based on the model’s score in several popular benchmark tests. The new model takes advantage of a technique previously invented by Google researchers to achieve greater performance without requiring more computing power. The technique, called expert mixing, selectively activates the parts of a model’s architecture that are best suited to solving a given task, making its training and execution more efficient.