Creating Turing-test passing chatbots is getting easier
“The Turing test, developed by Alan Turing in 1950, is a test of a machine’s ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human.”
I’d heard this neat story recently, on the More or Less Human episode of Radiolab, on how it has become easier to write a chatbot that passes to Turing test, because how we communicate has changed.
Over the last 5-10 years, most of our chat clients (eg: Messages, WhatsApp, Android Messages) have auto-complete, canned responses. Often, when you try to type something unique, the chat client suggests something else.
AND, data entry, whether by keyboard or voice on mobile devices, make it challenging to write human sentences.
Even when a human is writing, because of auto-correct, suggestions, poor data entry on mobile devices, and canned responses, our writing has become a lot more bot-like. Our expectations are for bot-like communication.
So, even without AI advances, it has become easier to write a Turing-test passing bot. Alternately, a human from 30 years ago would probably identify a human using a chat client on a mobile phone as a machine.