Google’s long-term ambitions for Gemini are becoming clearer, as the company moves toward transforming it into an AI system capable of deeply understanding and interacting with the world in a human-like manner.
In a recent update shared by DeepMind CEO Demis Hassabis, Google outlined its strategic direction for Gemini, particularly its Pro 2.5 version. The ultimate goal? To create a “universal AI” system that serves as a dynamic, adaptable assistant in everyday life. Central to this vision is the development of a “world model” — a concept involving the AI’s ability to interpret, predict, and even imagine aspects of the real world.
According to Hassabis, Gemini’s evolution hinges on its potential to mimic cognitive processes similar to those found in the human brain. This includes the ability to understand context, anticipate needs, and perform actions across multiple devices in real time.
Google believes that early glimpses of this capability are already emerging in Gemini’s interactions with natural settings, showing promise in simulating real-world scenarios. This foundational groundwork, they argue, positions Gemini well for further enhancement.
Enhancing Capabilities Through Project Mariner
One of the major components supporting this shift is Project Mariner, which Google began testing with users in December. Since then, the multitasking agent has reportedly matured to the point of handling up to ten tasks simultaneously. Mariner enables Gemini to perform complex operations like retrieving information, booking tickets, and managing research activities — all in parallel.
Hassabis emphasizes that such multitasking is crucial for turning Gemini into a comprehensive AI system that not only understands but also acts with purpose and precision.
The Role of Project Astra
To complement Mariner, Project Astra is also being integrated more deeply into Gemini’s framework. Known for its features such as real-time video comprehension, screen sharing, and memory capabilities, Astra has already found a home in services like Gemini Live.
Google revealed that the feedback gathered from Astra’s usage is informing improvements across several offerings, including Search, Gemini Live, and the Live API. The company formally announced Astra’s transition into Gemini back in March, signaling the growing importance of this project in shaping the AI’s future.
Together, these initiatives underscore Google’s broader aim: to move beyond simple chatbot functionality and build a truly intelligent, world-aware digital assistant.
