In a recent experiment, I navigated an office building from the lobby to a 5th floor office, guided by ChatGPT's computer vision capabilities, and the direction instructions it gave me. This journey shed light on both the strengths and limitations of AI in navigation. In open areas, the AI was pretty accurate in its estimate of how many steps it would take to cross a room, it could read signage, instruct me where buttons were, and it could often estimate the right direction to get to where I wanted to go. However, the AI faced challenges in tighter spaces, some inaccuracies in direction, and occasional confusion between right and left turns. Below are a series of photos I took and uploaded to ChatGPT as I walked through the building. I followed its instructions on how many steps to take and the direction to go in. It was not perfect but shows some potential.
This experiment shows that this could be a useful navigation tool, but would need further refining for real-world applications. By enhancing AI's capabilities and integrating it with technologies like Meta's glasses or Google Maps, we could facilitate seamless navigation within interior spaces. The potential applications of such technology are vast, ranging from personal navigation aids to advanced uses in robotics, drone delivery, and even ethically questionable military applications.
The design and layout of the building played a crucial role in the effectiveness of the AI navigation. Clear signage, particularly those well-lit and prominently placed, greatly aided the AI's ability to guide accurately. The simple lobby layout and consistently designed elevator buttons also contributed to the smooth navigation experience. These observations provide valuable lessons for architects and user experience designers: designing with predictability and clarity not only benefits those with visual impairments but also enhances the overall user experience. It's a reminder that inclusive design can lead to improvements for all users.