Google readying Gemini 2.0 and AI agent for automating web tasks: Report

Other companies, including OpenAI, Meta, and Anthropic, are also working on new AI models for potential late-year releases

Google is reportedly planning to launch its next-generation Gemini AI model in December this year. According to a report by The Verge, the company aims to release the Gemini 2.0 model to compete with OpenAI, which is also expected to launch its next flagship AI model before the end of the year. Additionally, Google is said to be developing an AI-powered agent called “Project Jarvis” for automating web-based tasks.

 

Gemini 2.0: What to expect

 

The report indicates that Google is working on the next-generation Gemini AI model, which the American technology giant plans to release “widely” in December. However, the model is reportedly not demonstrating the expected level of performance. “The model isn’t showing the performance gains the Demis Hassabis-led team had hoped for,” the report states.

 

Google is not the only company eyeing December for the release of its next model. Microsoft-backed OpenAI is also reportedly planning to launch its next major AI model by December. Last week, The Verge reported that the new AI model, codenamed Orion internally, is considered the successor to the GPT-4 model. Unlike Google, however, OpenAI is reportedly planning to release the new model to its partner companies, such as Microsoft, allowing them to build their own products and features.

Elon Musk’s xAI, Meta, and Amazon-backed Anthropic are also reportedly developing their next-generation AI models, some of which could be released before the year ends.

 

Google’s Project Jarvis: Details

 

Citing The Information, The Verge reported that Google is working on a large action model codenamed “Project Jarvis.” This concept is said to be powered by a future version of the Google Gemini model and aims to automate online tasks such as research, product purchasing, ticket booking, and more.

 

According to the report, the AI-powered agent is specifically tuned to work with web browsers, particularly Google Chrome. In its current state, the AI agent takes and interprets screenshots, then clicks buttons or enters text. The feature is still under development, and the report noted that there is currently a delay of a few seconds before it takes action.