99 Percent Invisible covers the origins of conversational human-machine interaction.
ELIZA was a simple computer program. It would look for the keyword in a user’s statement and then reflect it back in the form of a simple phrase or question.
Is that what real conversation is? Conversational AI makes a lot of promises. Reflective listening is indeed one form of human interaction, but Eliza falls short of even that, and at some point inference, reasoning, and other contextual concepts are necessary to truly engage. When will systems move beyond the Q&A of reflective listening and fact-finding?
Though the "the speech patterns of a therapist might be easy to automate" and people report feeling good after the interaction, the role of conversation in our lives is much richer and more of a tool that requires wielding skill on both sides. Many of us might feel better after a venting session in which we can release certain emotional burdens, however when it comes to performing a job or organizing a family, a system that doesn't "know anything about the world" isn't going to be much help.
Of course, now conversational systems are being used for much more than therapeutic interactions. But the growth of subject coverage has not been kept up with by increases in the sophistication of the interactions. Perhaps "we’re just fascinated to see ourselves reflected back in these intelligent machines" for now, however, today's capabilities are unlikely to meet tomorrow's demand.
That's where Points Made comes in. Let's make this better. Sign up to stay up to date today!