Such artificial 'pause fillers' as machine-generated supplements for silence may help to produce a more natural-sounding cadence in machine-generated speech patterns and so help to convince the AI's human interlocutors that they are engaging with another human. AI's 'humanity' of silence emerges in these contexts not least because a short period of silence is likely to be interpreted as a 'pause for thought', therefore reinforcing the illusion that the machine is 'thinking' before responding, just as a human might. However, the technical issues flagged in AI's poor handling of higher duration and frequency silences, no less than the machine inference that silence can be filled with relevant sounds in some conversational situations, are both significant. In this context at least, the AI may be starting to recognise silence not as an absence but as meaningful data. Yet the AI's confusion when presented with substantial silences, and its understanding that short break tags are equivalent to vocal markers such as 'uh's' and 'ah's' indicate the AI's continuing preferences for sound over silence, for presence over absence.