At the annual I/O developer conference, Google unveiled Project Astra, an ambitious endeavor to build a universal AI assistant of the future. The goal is to create a multimodal AI agent that can understand and respond to the dynamics of the world in real-time, assisting with routine tasks and answering questions. This concept is similar to OpenAI’s GPT-4o-powered ChatGPT.
The early version of Project Astra was demonstrated at the conference, showcasing its potential. The assistant displayed the ability to identify objects, describe their components, understand code written on a whiteboard, and even remember where the user kept their glasses. The videos demonstrated the assistant’s advanced capabilities, including suggesting improvements to system architecture.
According to Demis Hassabis, the CEO of Google DeepMind, for an AI agent to be truly useful, it needs to understand and respond to the complex and dynamic world like humans do. It should also have the ability to learn, be proactive, and respond naturally and without delay. To achieve this, Google’s agents continuously encode video frames, combine video and speech input into a timeline of events, and cache this information for efficient recall.
While Google has not shared a specific response time for Project Astra, Hassabis acknowledged that achieving human-level response time is a difficult engineering challenge. OpenAI’s GPT-4o, on the other hand, processes all inputs and outputs across text, vision, and audio, delivering responses in an average of 320 milliseconds. It remains to be seen if Project Astra agents will have the same emotional range as OpenAI’s GPT-4o.
As for availability, Project Astra is still in its early stages and there is no confirmed timeline for its launch. However, Google has confirmed that some features from Project Astra will be integrated into the Gemini app later this year. Gemini Live will allow users to engage in two-way conversations with the AI assistant, and eventually, it will include vision capabilities, enabling users to discuss their surroundings using their cameras. Users will also be able to interrupt the assistant during these conversations, similar to OpenAI’s ChatGPT.
Google envisions a future where people can have an expert AI assistant by their side through a phone or glasses. With technology like Project Astra, everyday tasks could become easier and more efficient. Google’s commitment to building a comprehensive AI agent reflects its dedication to advancing AI technologies and making them more accessible to users.
