Google introduced a new version of its flagship model of artificial intelligence — Gemini 2.0. According to the company, it is twice as fast as the previous version, can generate images and audio in different languages, help with Google searches and more. Experimental agents for programming and video games were also shown.
Google introduced a new version of its flagship model of artificial intelligence — Gemini 2.0. According to the company, it is twice as fast as the previous version, can generate images and audio in different languages, help with Google searches and more. Experimental agents for programming and video games were also shown.
Gemini's new capabilities "allow you to create agents that can think, remember, plan and even act on your behalf," said Tulsi Doshi, director of product management at the company, during a conference call on Wednesday, according to Bloomberg.
The company said it will start testing Gemini 2.0 this week in Search and in AI Overviews, which appear at the top of Google searches. This will improve the speed and quality of search results for more complex questions, such as complex math equations. On Wednesday, the company also gave developers access to an experimental version of Gemini 2.0 Flash , a fast and efficient artificial intelligence model that Google says can better process images and come closer to human reasoning.
The company also introduced a new web feature called "deep research" that will allow Gemini users to use AI to dive into topics with detailed reports. The feature, called an AI-powered research assistant, will be available Wednesday to users of Gemini Advanced, Google's paid AI subscription product. Meanwhile , Gemini users worldwide will be able to access a chat-optimized version of the experimental Gemini 2.0 Flash . Next year, this model will appear in other Google products.
Google's DeepMind lab is expanding trials of Project Astra, an AI agent that uses a smartphone's camera to process visual input. Google employees demonstrated how Astra can summarize information on a page.
DeepMind researcher Greg Wayne said the agent has improved since it was first introduced at Google's developer conference earlier this year, and it can now respond by speaking at the same speed as a human.
The company is also testing Mariner, an experimental web assistant designed to help users fill their online shopping carts and organize their digital lives.
Two more artificial intelligence agents were also shown, with which researchers experimented within the company and with groups of trusted testers. The first, called Jules, is an AI-powered software agent for engineers that focuses on fixing bugs in software code and performing routine programming tasks. The second is an as-yet-untitled AI agent for video games that aims to help players by reasoning about the game based on the screen and offering suggestions in real-time. The company called it an "early experimental stage" that aims to demonstrate some of the capabilities of the AI agents that will appear in Gemini 2.0.
