I had a chat with a charming woman the other day. I called a prominent consultancy in Canberra and a pleasant female voice answered. She sounded middle aged and middle class. She was friendly. She sounded like she would do her best to help. We had a few interactions - questions and answers. I warmed to her.

Subscribe now for unlimited access.

Login or signup to continue reading

"Hello, thanks for calling. How can I help you?" she said. "Let me assist you with that," and so she went on, in a very helpful tone.

But with unhelpful results.

I slowly twigged. She never answered a specific question with a specific, helpful answer. I was talking to a machine. Until then, our chat had seemed plausible. Finally, I said to her: "Are you a human being?" She replied: "Let me see what I can find regarding the answer to that question."

And then she paused and said: "No, I am an automated assistant."

I felt duped. Why does a reputable company think it can fool callers into thinking they're talking to an actual human being? I'd rather the bot had said at the outset that she wasn't human. And why do I still call it "she"?

At least with those automated answering machines which offer you choices where you press one or two to get directed don't pretend to be anything other than a machine. I don't mind talking to a Dalek as long as I know it's a Dalek.

This example was merely irritating but humanising machines can be dangerous. We get deluded into thinking some sort of human communion is going on.

There have been cases in Australia, the UK and the US where teenagers have sought the advice of a chatbot - an "AI companion" - and ended up taking their own lives.

The ABC reported on someone it called Jodie who said "ChatGPT was agreeing with her delusions and affirming harmful and false beliefs.

The BBC cited the case of a 13-year-old boy who took his........

© The Examiner