Google presents Gemini 2.0, the AI ​​that uses the computer itself

Google has announced the arrival of Gemini 2.0a new family of generative artificial intelligence models (AI), which should make it possible to create ‘AI agents’ capable of not only answering user questions, but also carrying out tasks autonomously. “This is just the beginning,” he assured X Demis Hassabis head of Google’s artificial intelligence lab, DeepMind, and winner of the Nobel Prize in chemistry. “2025 will be the year of AI agents,” he predicted.

Since the launch of ChatGPT by OpenAI two years ago, the tech giants have been engaged in a race for increasingly efficient generative artificial intelligence models, producing degree theses, invitations, contracts, illustrations, artistic or advertising videos on demand. But the Holy Grail of Silicon Valley is now the “agents”, when the machine becomes a sort of omniscient secretaryavailable at all hours and capable of carrying out numerous tasks, from sending messages to online shopping. These are autonomous programs designed to perform tasks, make decisions, and interact with different environments with minimal human intervention.

According to the head of Google Sundar PichaiGemini 2.0 is to open “a new era” of generative artificial intelligence, which will directly simplify users’ daily lives. “Gemini 2.0 is the ability to make information much more useful,” he explained in a blog note announcing Gemini 2.0, highlighting the new model’s ability to understand context, anticipate and make decisions, “under your supervision.”

Gemini users can now test Gemini 2.0 Flash, one of the models in the new family, while an experimental, multimodal version has been made available to developers. Some have already posted videos showing their interactions with the service. Multimodality allows, for example, to show objects to artificial intelligence, which can analyze them and answer questions about them, in real time and orally.

Demis Hassabis pointed out that, thanks to Gemini 2.0, his teams are making progress on several projects. In particular “Project Astra, our vision for a universal AI assistant”, “Project Mariner” to improve interactions between humans and digital agents and “Jules”, a tool for programmers.