I don't have Siri. I have her younger and less glamorous sister, Prototype Gretchen, who came with my iPhone 4. I like the little wiki-doodle clinging to the cord of my Apple earbuds that allows me to talk to Gretchen. I like to command her to do tasks like “next (song)” and “call Steven Crabgrinder on mobile.” But there's one particular job that Gretchen can't – or won't – do, because there's no way to tell her how to do it. Interested as I am in the way the human voice wraps around words, I often find myself wanting to scan back several seconds, several times in order to hear a particular inflection or phrasing that just happened. I don't need the whole track over. Just the last few moments. There is no way I have figured out to express this desire to Gretchen. She's seems to be an all or nothing type of gal.
When the magicians at Apple address this (they may have already done so, but I don't know because, like I said, I'm no acquaintance of Siri's), I hope that they choose some method other than crass speech. Calling out fixed durations, like “repeat five seconds,” may be direct and efficient but it's also uninspired. More elegant, and publicly pleasing, would be for the listener to add her/his own voice as instruction to the world. So, for example, if one would like to plunge backwards into the present track twelve seconds: press the dangly-widget and sing out within a particular tonal range. Perhaps the lower the pitch, the faster in reverse you go; the higher the faster forward. Imagine the street scene, as our musical devices encourage a musical world. Because out of that insular feedback loop between human and machine, some song should spring.