The Hook
Be My Eyes is an app that connects blind and low-vision users with sighted volunteers for real-time visual help — reading a label, identifying a product, navigating an unfamiliar space. In 2023, the company integrated GPT-4 into a feature called Virtual Volunteer (later Be My AI). The question was whether a language model could do what a volunteer does.
The Question
When the same image-and-language technology that elsewhere generates synthetic media and floods recommendation engines with undisclosed AI content is pointed at accessibility, what does it produce?
The Paper Trail
Where the impact has been measured — specifically, on Microsoft's Disability Answer Desk, which piloted the same integration — the tool resolved approximately 90% of queries in approximately 4 minutes, against a baseline of approximately 12 minutes, with a user satisfaction rating of approximately 4.85 out of 5. The Be My Eyes integration does not require a sighted human to be available on the other end of the call.
The use case is specific: a blind user needs to know what is on a label, in a room, or on a screen — right now. The technology's capability — describe what is in an image, in natural language, in real time — is a direct match for that need. No re-engineering required. The question was whether someone would build it.
The Synthesis
The same computer-vision and language models that elsewhere clone voices, generate misleading images, and flood platforms with synthetic content are here answering a question a blind person couldn't answer alone. The technology is neutral. The direction it is pointed is a human decision — and this direction produced something that was not possible before.
This is the honest positive the other stories in this lane need: not to dilute the pattern, but to prove the outcomes are not determined by the technology. They are determined by the choices made about where to point it and what to optimize for.