Author: Stephen Katte, CoinTe; Compilation: Tao Zhu, Golden Finance
Google's AI research lab DeepMind stated that its latest released AI model Gemini 2.0 will serve as the foundation for building more advanced AI agents.
Google DeepMind CEO Demis Hassabis and CTO Koray Kavukcuoglu stated in a blog post on December 11 that the Gemini 2.0-powered AI agents released on December 11 can understand complex instructions, plan, reason, take actions across websites, and even assist in formulating video game strategies.
Hassabis and Kavukcuoglu stated, "The practical applications of AI agents are an exciting area of research full of possibilities.
"We are exploring this new field through a series of prototypes that can help people complete tasks and get things done."
According to Hassabis and Kavukcuoglu, there are currently several experimental AI assistant projects powered by Gemini, each with different functionalities.
One of the projects is called Deep Research, which can create multi-step research plans by searching the web and then generate a lengthy report on the research findings, helping users explore complex topics.
Project Astra is a general-purpose AI assistant primarily aimed at everyday tasks, such as providing suggestions and advice based on user prompts, like how to do laundry or more information about landmarks.
Project Mariner focuses on creating an AI agent that can control your Chrome browser, move the cursor, click buttons, fill out forms, and browse websites.
According to Hassabis and Kavukcuoglu, these projects "are still in the early stages of development," but they hope to widely apply them in future products after testing and further development.
"It's still too early, but the Mariner project indicates that navigating in the browser has technically become possible, although it doesn't always complete tasks accurately and is slow at the moment, this will improve rapidly over time."
Meanwhile, the Jules project is under development as an assistant for developers that can be directly integrated into the GitHub workflow and help with tasks like coding and planning.
Hassabis and Kavukcuoglu stated that they are also using Gemini 2.0 to build agents for video games that can provide players with suggestions for the next actions in real-time conversations and search online for 'rich gaming knowledge.'
"We are working with leading game developers like Supercell to explore how these agents work, testing their ability to interpret various game rules and challenges, from strategic games to farming simulators," they said.
In November, Marc Benioff, CEO of the American cloud computing software company Salesforce, stated that the future of AI lies in autonomous agents rather than large language models (LLMs).
"In fact, I think we have now reached the ceiling of a Master of Laws," he said in the 'Future of Everything' podcast on November 23 (Wall Street Journal).
Nvidia is also focused on positioning itself at the forefront of trends.
"We are seeing the number of AI-native companies continue to grow. Of course, we are starting to see that businesses adopting agent AI is indeed the latest trend," said Nvidia CEO Jensen Huang during the third-quarter earnings call in November.
Additionally, Hassabis and Kavukcuoglu stated that the team is experimenting with robotics to create agents that can assist in the physical world. Currently, Google's AI agents are only available to testers and developers.