My apology; And why I won’t believe anything SpinVox says in future

Spin is inherent in the mobile industry.

I read it all the time and most of the time it’s gently benign.

“Our application is available on 110 handsets,” is a good example. Peer into the list and you’ll find the majority of them are Nokia-only. The company hasn’t bothered developing for LG. Or Sony Ericsson or anyone else. But the 110 figure sounds better than saying ‘it only works on Nokia’.

This is fine. It’s acceptable.

Doing a SpinVox is not. Let me elaborate.

For years I’ve been telling everyone who’ll listen not to worry about SpinVox. No human actually sees / hears your entire message, I’ve been telling them. I’ve been quoting near-verbatim what I’ve been told by the SpinVox team. I accepted this as truth, not spin — truth about precisely how their system functions.

I found out yesterday that this wasn’t entirely accurate.

In fact, based on the media demonstration yesterday, I’ve been telling outright falsehoods.

I can’t quite believe that I’ve been bullshitted so much.

Here’s a key paragraph from Milo’s TechCrunch post (discussing the demonstration yesterday):

Here’s where it got ugly. From observing the “tenzing” process in action, it was clear to us that the system had failed to pick up a single word in the message correctly. The agent in the room had to listen to and manually type the entire message, from beginning to end. SpinVox has previously claimed that agents do not get to hear entire voicemail messages; only enough to give context and enable transcription. That’s not what I saw this morning.

I refer you next to Andrew Orlowski’s post about the SpinVox visit on The Register:

It may be fair to say that we saw the difficulty of machine translation: all but one of the messages – a simple one placed by the SpinVox chief technology officer in a silent room – tripped through to the Tenzing console for manual interpretation. Pretty much in their entirety. So much for call center staff, sorry, agents, only seeing occasional word fragments.

And finally let’s have a look at the official SpinVox clarification post from last week:

No. As per fact one, SpinVox only employs agents to step in when messages need analysis and the machine gets to decide. However, the agents in question will only ever hear/see the specific parts of the messages that need work on. They never see fully automated message conversions because we don’t send them on once they’re complete.

Right then. That doesn’t quite tell the whole story, does it?

Who’s the chump then? Me. I’m the one who believed it. I’m the one who went round telling anyone who’d listen that this SpinVox stuff was amazing and, hey, no don’t worry about your privacy — it’s only the *specific parts* of your message that are ever going to be listened to, you know, the weird word or phrase that the system can’t understand. THAT bit is sent to the humans for assistance.

I apologise profusely to everyone that I’ve misled. It most certainly wasn’t at all deliberate.