Tested ChatGPT video in the train station! Just like a sighted stranger: 1/3 actionable info,2/3 "it's crowded", "there's a handrail", "watch your step". Congrats overhelpers of the world — you worried about Blind people so loudly and for so long you've got a supercomputer giving worry-as-a-service

From a Blind person motivated to check: ChatGPT voice mode with video grabs a frame and responds. It's not parsing video really. Count to five while showing five gestures to describe, or ask it to decide if you blink a light. It can't. It gets by on freeze frames and vibes