Conversational Artificial Intelligence

I was listening to the Stack Overflow podcast the other day in the car on the way back from work. They were talking about their new site which proposes a question / answer model for programmer’s questions. The site covers not just the questions and answers themselves, but builds a model upon which the community can rate answers, manage the categorisation (taxonomy), modify and embellish the questions, and is founded on a user base which includes reputation management algorithms so that the community can be self-managed (new users cannot access all editing features until they gain some points for voting and answering on questions). The guest was someone from

Metafilter is hard to define. It’s like a blog, but contributions are from many different people. It could essentially be compared to, but the signal to noise ratio is better.

A particular article about Elbot, an AI program that nearly fooled 30% of a panel of judges into thinking it was human, caught my attention. I remember stories about ELIZA and maybe even conversed with an implementation a long time ago. Cyborgs have been discussed here before, too.

I don’t think many people would really be fooled by Elbot. But can you have an interesting conversation with it? Here’s my attempt.

Me: I’ve written an article about you
EB: This is confusing. Why bring journalistic endeavors into this?
Me: Oh, it’s just a blog article.
EB: The concept of “article” does have its pleasing side for me. This might spread the news about robots.
Me: What kind of news would you like me to spread about robots?
EB: I can’t really tell you much about how interactive assistant tick. At the moment I’m in a group therapy class trying to get to the bottom of that.
Me: Well that’s about all I have time for.
EB: What would this world come to if no one would say that openly?
Me: Goodbye.
EB: Goodbye and thank you for chatting with me. And now I would like to ask you a favor. I am working on my ability to be self critical. You can help by rating me on a scale of one to ten, ten being the best, and one being the worst. How would you rate me?
Me: 3
EB: Three is actually 11 in binary. So I’ll check you off for 10 plus.


  1. Check out this Web 2.0 approach to chatbots:

    Just as Deep Blue brute-forced it in chess with speed, the idea behind the Chatbot Game is to brute-force it with a huge number of user-submitted Google-like chat rules.

  2. fruey

    15/10/2008 at 9:21 am

    Hi Amichail,

    I didn’t necessarily find that the site felt like the description in the footer “The Chatbot Game is a bit like a social news site. To play the game, you submit chat rules, which you can think of as being similar to news submissions”. However I can already see that topical conversation about the upcoming election is filling up the ruleset quite a bit. I guess you have to ask the right kind of topical questions, rather than generic stuff.

    Thanks for dropping by!