At the I/O 2024 event, Google unveiled a teaser for Project Astra that showcased the future direction of AI assistants. This new feature merges the intelligence of Gemini with Google Lens’ image recognition abilities and sophisticated natural language responses. Despite the impressive promotional video, hands-on experience revealed that Astra’s practical application is still under development. Here are three key insights from our initial interaction with Google’s cutting-edge AI.
Sam’s Perspective:
Digital assistants are commonly interacted with via voice, making Astra’s multi-modal approach (incorporating sight, sound, and text/speech) a fresh and interesting concept. This approach aims to make computer-based assistants act more like real-life assistants or agents, moving beyond the robotic execution of voice commands. During our demo, we engaged Astra in an imaginative exercise where it spun a narrative involving a dinosaur and its trusty baguette, a demonstration of its fun, albeit limited, capabilities. Although the AI delivered as might be expected, it didn’t live up to the all-encompassing assistant portrayed in Google’s teaser. In one experiment, the AI successfully identified objects drawn on a touchscreen and, in another instance with the Pixel 8 Pro, could track and remember the location of objects and even clothing items. These interactions showcased the potential and current limitations of AI, from helping locate misplaced keys to acknowledging the challenges in memory and processing.
Karissa’s Viewpoint:
Out of all the advancements in generative AI, multimodal AI intrigues me the most. The concept of an AI that can perceive and interact with its environment in real-time is thrilling, promising applications far beyond text-based interactions. Project Astra hints at such futuristic applications, including potential integration into smart glasses, despite currently being experimental. In comparison to similar technologies from other companies like Meta, Astra stands out with its “memory” feature. However, its memory is currently short-lived, though theoretically capable of expansion. This capability suggests a future where Astra could provide assistance based on a more extended memory, approaching the type of assistance seen in science fiction. Despite the potential, it’s evident Google has significant work ahead to realize these exciting possibilities.
Stay updated with all the news from Google I/O 2024 here!
