Maybe, but this brings up an interesting aspect of machine “intelligence” that I think has very profound implications for the long haul Unlike a human, the bot will only need ONE correction of a grammer mistake and then it’ll be correct for the next 1000 years. When the machine is capable of “recursive self improvement” it’ll be able to study grammer and fix itself.

Why hasn’t this happened yet? Because programming that level of activity is a lot easier for me to talk about than for guys like Tommo to actually do! But get cracking Mr. T cuz we need them thinkin’ puters ASAP!

It’s certainly true that a bot with an ungrammatical repertoire will produce ungrammatical output. I presume that QA would weed that out at a place like Palm.

Grammar? Does Joe’s use of the word ‘grammer’ prove he is human? Wouldn’t a bot misspell the word too if it was a poorly engineered one? Is quality of the interaction the test for human versus bot?

Irrespective of whether this particular transcript is that of a human or an artificial intelligence program, I wonder why people prefer to deal with a voice rather than with displayed text? One airline company yanked out an automated lost-baggage system and put in a human being who performed less well but sure dealt with frustrated travellers better.

Sorry, couldn’t resist. I’ve had equivalent bad experience both on the phone and via chat. On the one hand, phone *tends* to give you the same answer but faster – on the other hand during a chat it’s much easier to be doing other stuff and checking the screen every now and again. I think I pretty much agree, most of the time I prefer using the phone.

But I agree with Tommo that we’d actually see better grammer and more logical interaction with a machine session.

I have found that chat help people tend to seize upon the most commonly encountered reasons right away.