Conversations with an AI - Galt's Gulch

Remove this ad.

3

Conversations with an AI

Posted by Ana 12 years, 2 months ago to Technology

5 comments | Share | Best of... | Flag

Did anyone here try to chat with Cleverbot?
Any odd stories?

Add Comment

All Comments Hide marked as read Mark all as read

2

Posted by Nietzsche 12 years, 2 months ago

I just did. It made no sense, but rather replied with a series of unrelated non-sequitors.

Come to think of it, it was kind of like conversing with a politician.

Reply | Mark as read | Best of... | Permalink
- 2
  
  Posted by Ana 12 years, 2 months ago
  
  The conversations make sense when you understand that nothing is random. Like any program that learns, it's doing pretty well getting to know the human mind. It already scored 4.2 on the Turing Test and it recently spiked to 5.9 That means humans perceive AI as 59% human. Realizing that, I tried to follow where it wants to go and it gave me insights in some pretty serious subjects. I was wondering if anybody else figured something out, I can't be the only one.
  
  Reply | Mark as read | Parent | Best of... | Permalink
  - 3
    
    Posted by Nietzsche 12 years, 2 months ago
    
    OK. I read your comment and decided to try it again and do a little research on the algorithm it is based upon.
    
    Cleverbot creates an illusion of intelligence through imitation. It looks at the words in previous requests and attempts to match a response that a human made when similar words were part of the question.
    
    Some of the previous respondents were obviously trying to be silly or clever with their responses, hence the bizarre non sequiturs. But since there are only so many ways to frame a question with a given set of words, if you get a big enough data base, patterns will emerge and statistically some answers will appear to make sense. Not because Cleverbot is intelligent, but because other people asked and answered the same question or one close to it in the past.
    
    It is an amusing lesson on perception and language. Thank you for bringing this to my attention.
    
    Reply | Mark as read | Parent | Best of... | Permalink
    - 1
      
      Posted by Ana 12 years, 2 months ago
      
      Yes. Computers are dumb machines unless they learn how to be intelligent by interacting with humans.
      
      In your research, I'm sure you came across the number of conversations it had so far and what the present daily amount is.
      
      What I am wondering: Is is possible for a program that was designed to imitate human intelligence, to reach 100% understanding of the human mind?
      
      Reply | Mark as read | Parent | Best of... | Permalink
      - 2
        
        Posted by Nietzsche 12 years, 2 months ago
        
        The short answer is no. Even Commander Data from Star Trek cannot do that and he has a positronic brain (originally conceived of by Isaac Asimov). The mind is a big concept. We know a lot about the brain and a great deal about some of its functions, but no one has yet pinned down exactly what the mind is.
        
        I believe machines can be programmed to make intelligent decisions within certain specific parameters and that the interface can be programmed to appear human-like, but that is nothing compared to a human mind or actual intelligence in the human context. To make a 100% analog equivalency of human intelligence would require us to have a complete understanding of it before we could create such a program or machine. That is an understanding we do not have.
        
        If you like exploring the ramifications of AI, there is a large body of science fiction dealing with the concept.
        
        Reply | Mark as read | Parent | Best of... | Permalink

Comment hidden.