Navigating Number Sequencing in Phone Calls
Published on: May 29, 2025
On November 17, 2023, an interesting event unfolded when RHOTO assisted a caller on a personal phone line. The caller, in an attempt to leave a message with their phone number, naturally paused between the sequences, articulating it as '555 (pause) 666 (pause) 1212'. This pattern of speech presented a unique challenge for RHOTO's existing programming.
Originally, RHOTO's SIP interface was configured to 'stop listening' while processing commands. This feature was implemented to avoid the accumulation of queued commands and the resulting overlap of speech. Additionally, it prevented RHOTO from erroneously picking up its own speech, a common issue in scenarios involving speakerphones. However, this programming meant that RHOTO missed the second part of the number during its processing phase, as it was not listening but processing the initial segment.
The caller, treating RHOTO akin to a human with limited memory, eventually provided the full phone number in a single sequence, which RHOTO successfully processed and recognized as a complete phone number. Despite this workaround, the incident has been noted as a bug in RHOTO's system. No immediate update was made to RHOTO's behavior, but the issue highlights a potential area for future refinement in the AI's understanding and processing of spoken information.
Read more research. See the list.