To delve closer into what makes a chatbot ticks, I decided to interview Cleverbot and get some answers:
Me: Hey, how’s it going?
Cleverbot: Because I try to make sense of you.
Me: What do you mean?
Cleverbot: We’re just confusing each other. Want to just drop it?
Me: Yeah, let's drop it.
I'm writing an article on artificial intelligence right now.
Cleverbot: Obviously as I am human.
Me: Right...so, do you
have anything to say on the topic of artificial intelligence?
Cleverbot: Yeah, can we talk about Japan?
Me: I was talking about
artificial intelligence, but we can talk about Japan.
Cleverbot: Oh, okay. So, what
you're telling me is that the unicorns cover infants in maple syrup and then
eat them. I don't feel so alone anymore.
Me: How does this relate to Japan?
Cleverbot: I just elaborated on
my trip to Disney World.
Me: So on your trip to
Disney World, you saw unicorns eating infants?
Cleverbot: How did you come to
that conclusion?
This conversation was interesting to say the least. The program often developed complete sentences and ideas, but wasn't very cohesive. It seems that one of the greatest weaknesses of Cleverbot is
its inability to focus. From the beginning of the topic, it danced from
artificial intelligence to Japan, and then from unicorns eating infants to Disney World. Cleverbot could once in a while come up with adequate responses to a line of text (e.g., “How
did you come to that conclusion?"), but couldn't look at the big picture – the whole
conversation.
Cleverbot isn't the only AI with a lack of planning and focus. This is a common problem among programs simulating intelligence: how can you program a system to predict and plan?
To gain more insight, let's look at the learning process of Cleverbot. The Cleverbot website describes how the program learns: "Things you say to Cleverbot today may influence what it says to others in the future. The program chooses how to respond to you fuzzily and contextually, the whole of your conversation being compared to the millions that have taken place before."
We can see Cleverbot borrowing other people’s dialogues in our own interview. Lines like: “So,
what you're telling me is that the unicorns cover infants in maple syrup and
then eat them,” obviously came from another person talking to Cleverbot.
Nevertheless, Cleverbot is a wonderfully clever chatbot that took time and dedication to program and train. Talking to it greatly improved my understanding of how chatbots function. However, I'm still curious on the topic of how these chatbots are programmed, and there's only one way to really know: I have to do it myself.
I'll see you in the next post, where I take you through my experience of developing my very own chatbot.