More than ever, interacting with AI has become part of our daily lives. AI helps predict what movies, TV shows and music we might like — taking the role of a trusted critic for many Netflix enthusiasts. We ask Alexa and Siri to manage our schedules, remember our friends' addresses and pick the right music for our mood.
To help these AI perform as well as possible we ask them to learn to predict our behavior and our desires. And now, many AI have become more than useful knowledge repositories — they're trying to predict our taste, and anticipate our needs before we are even aware of them. As conversational AIs continue to get smarter and learn, how long will it be until they know us better than we know ourselves?
How does conversational AI learn?
Conversational AI uses natural language processing to be able to respond to human speech with sentences that feel, well, natural. If you ask Siri for the weather, she's likely to respond in a full, grammatically correct sentence — no matter what language you ask in.
Traditional, scripted chatbots — one-time AIM users may remember their old conversation partner SmarterChild — can only respond to designated keywords with pre-programmed responses. Modern conversational AI is much more sophisticated, and is designed to learn how to understand questions and requests, rather than just spitting out an automated reply.
A conversational AI needs to "learn" how to speak, and typically that doesn't look entirely different from how a baby would learn. The AI will be exposed to a large amount of "human-to-human" interactions to learn from, and will have its own "human-to-machine" interactions which will help it continue to improve. Modern conversational AI using machine learning will continue to learn from its "human-to-machine" interactions even after it's been deployed.
Meeting needs vs. predicting needs
Much of conversational AI is currently in the business of chatbot interfaces. That means when users are interacting with a website, they may encounter a conversational AI that offers to help them, similarly to a sales associate in a brick-and-mortar store.
However, where traditional chatbots fail is that they can only respond to direct inquiries. They have no capability at all to predict what people may need, or to draw conclusions from multiple parts of a conversation.
A great sales associate can help a customer find something that even they didn't know they needed — and help them articulate the needs they do have. That's what makes conversational AI a powerful and innovative tool; it's capable of both responding to direct requests and questions, aka reactive engagement, as well as predicting something that hasn't been directly stated, known as proactive engagement.
Conversational AI is capable of drawing connections between different parts of a conversation, learning about one user and their behavior patterns, and then further connecting those learnings to its wider database of knowledge regarding interaction. That may sound complicated, but it's something that humans do intuitively — if a friend asks to go out for the night, our previous knowledge of them will inform us as to whether they're more likely to want a quiet night in or a big night out, and our informed knowledge of the world will tell us what kind of activities would suit either direction.
Making these connections is an intrinsic part of what makes conversation feel natural as well as engaging — and it's what makes conversational AI both a better sales associate and a better chat partner than the pre-programmed chatbots of the past.
Press Release Service by Newswire.com
Original Source: Does Conversational AI Know You Better Than You Know Yourself?
