chatbots

[Paul Bowden]

Discussion of chatbot techniques for AGI (artificial general intelligence)

[Paul Bowden]

Anyone out there want to talk about chatbots and artificial general intelligence? I’m building such a system myself…

[Andrew Martin]

There’s plenty to talk about, I guess you need to clarify if you’re more interested in various implementation practicalities, more general linguistic principles, or very general philosophical principles. 🙂

[Paul Bowden]

I want to talk about practicalities of making an AGI: a chatbot that can reason like a human does. For example, what about Winograd schemas? These are rather difficult it-anaphor or they-anaphor resolution problems. (“The soldiers fired at the protesters and we saw several of them fall.” What does “them” point to? The soldiers, or the protesers?) Are there any tricks to solving them, or do we really have to put World Knowledge into our chatbots? (OK, that’s philosophy too, I grant you!)
Or how about “Could you use a lobster as a weapon?” – how could you solve ANY “Could you use an X as a Y?” question? Surely, WK (World Knowledge) is needed here… but is this an “AI complete” problem (meaning that you really do need an AGI, with full WK, and full human-strength reasoning abilities) or not? (I suspect that it IS AI-complete.) I’d like this particular chatbot talking-shop to be all about AGI aspects of chatbots/conversational agents.

[William Rayer]

I have some experience with chatbots. IMO you can give them a limited amount of real-world knowledge and in a sense they can reason in the same way as a human. Most chatbots use rule-based pattern matching approaches and have evolved from a pragmatic approach (“make something that can hold a conversation”).

[Matteo Trevisan]

Yes chatbot cannot seems intelligent, they are made of algoritms but who can really understand how is made human brain? We speak and somethimes we dont know what we speak about are we in a random mode? recognizing what we read?
mah!

[Author]

Posts