At Google I/O on May 14, 2024, Google DeepMind shared Project Astra, described in Google’s announcement as “our vision for the future of AI assistants.” The reveal centered on a demonstration in which a user pointed a phone camera around a room and held a fluid spoken conversation with the assistant, which identified objects, remembered where things had been, and answered questions in real time with low latency.
According to Google DeepMind’s project page, Astra is a research prototype for “a universal AI assistant” with “breakthrough capabilities” including natural audio interaction, action through tool use across Google products like Search, Gmail, Calendar, and Maps, and intelligent personalization with memory retention. Google described it as “a research prototype, being used and refined by a limited number of trusted testers,” running on Android phones and prototype glasses.
Project Astra mattered as the clearest early statement of where assistants were heading: away from text chat and toward a real-time agent that sees, hears, remembers, and acts. Many of its ideas later fed into shipping products like Gemini’s live multimodal features.