Google: A new Gemini model swallows texts and videos to find answers

How long will it take you to read it? war and peaceLeo Tolstoy's massive work of over 1,000 pages?

Google's new artificial intelligence (AI) model takes about one minute.

The company filed on Thursday Gemini 1.5It is an upgrade to the artificial intelligence models that it announced two months ago.

With a simple question, a user can ask the tool to analyze much more data than before: 30,000 lines of code, 11 hours of audio or an hour of video.

These numbers are for the Pro, the mid-range model of the Gemini 1.5, which can now hold 1 million packets of data, known as tokens.

This means it outperforms the Claude 2.1 by five times, the model produced by US-based Anthropic and considered the closest competitor, Google says.

“A new way to search”

Google's investments in AI highlight the race underway in Silicon Valley to commercialize AI, in the wake of OpenAI's stunning success with its chat engine ChatGPT.

In an interview with Reuters, Sundar Pichai, CEO of Alphabet Group, which owns Google, said that the new technology provides a new way to search for information on the Internet.

One example discussed internally at Google is the case of a director who asked the AI ​​to rate his film like a professional critic.

“This is just one case we discussed as a team, but the sky is the limit,” Pichai said.

In another example he provided, a user could perform complex searches that included the financial results of several companies at once.

In a video provided by Google, it takes Gemini 59 seconds to analyze a 44-minute video and extract information. In other videos, the AI ​​is asked to search for answers in text and images simultaneously.

In addition to ordinary users, the new model also targets enterprise customers in the cloud.

Starting Thursday, Gemini Pro will be available to a small number of companies, and companies developing software based on the 1.0 model will be able to replace it with the newer generation when it becomes available.

Pichai also said that the increased costs resulting from the high computational requirements of AI are not a concern for the company.

“These are profitable activities for us,” he said. “Over time, moreover, we have become more efficient at operating these models.”