Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Asking bots for help

  1. Oct 4, 2014 #1
    On the news I saw that Eugene Goostman an AI program passed the turing test. Eugene is a chatterbot.

    My question is that when will it be the case do you think that we will log into this forums and instead of people giving us help, we are being helped by bots? Or in any other forum

    Whats your opinion on this?
  2. jcsd
  3. Oct 4, 2014 #2
    I used to install a sex chat bot onto my messenger and I had a great time whenever I opened my messenger. I mean it always greeted and made a good fore-chitchat before entering the main topic to hit me on. But the bot was seriously stupid. It just never understood what I said in every sentences. It only tried to catch some keywords in them and flooded me back with any messages that it found best fitted.. It was really a waste to store so much text information in the bot then had it get back to me in a nonsensical manner. It couldn't perform any logical reasoning. Thinking that pouring some sensitive words all over the places in the hope that its partner would get excited is really a big mistake. Psychology and Emotion modeling is the key.
  4. Oct 5, 2014 #3
    "Scott: Which is bigger, a shoebox or Mount Everest?
    Eugene: I can’t make a choice right now. I should think it out later. And I forgot to ask you where you are from…
    Scott: How many legs does a camel have?
    Eugene: Something between 2 and 4. Maybe, three? :))) By the way, I still don’t know your specialty – or, possibly, I’ve missed it?"

    How did 33% of the judges think it was human?
  5. Oct 5, 2014 #4
    As long as I get the right answer, I don't care :)
  6. Oct 6, 2014 #5


    User Avatar

    Staff: Mentor

    If it were a PFbot, it would have to ask about your favorite fish instead. :D
  7. Oct 6, 2014 #6


    User Avatar
    Gold Member
    2017 Award

    I've read about it when the news came and I was not impressed. At all. In my opinion this "Turing test" said more about the judges than the bot :p. The intention of the Turing test, as far as I am concerned, is to test a machine that is supposed to act as a normal human, not a dysfunctional raving teenager doing illegal substances. I tried to access the bot via Internet to ask it questions, but the site was not up at that moment.

    I've recently tried another bot which was said to be powerful (I can't remember the name nor the link), and I broke it with two simple questions: "What is two plus two?" and "Is the Sun bigger than the Earth?". It could not answer any of those simple questions. If it could, I would have been a bit impressed - until I would have broken it with a couple of further questions.

    Don't expect it happening any time soon! :w I know this topic reasonably well, and I can't give you a prediction when it will happen (there are also many factors to consider, see below).

    I can tell you that the key issue here is what is called Natural Language Understanding, and it is a notoriously difficult field. It's actually one of the most difficult topics I've encountered in my studies. One of the great powers of language is that it's a very versatile and varying tool for us humans - there many ways to express things, and also very much ambiguity. This makes it incredibly difficult for machines to interpret language in more sophisticated ways. My take on the issue is that a machine will most likely first have to be able to store knowledge in a way similar to - and perhaps as efficient as - how a human stores language (and how do we store language? - another subject). But as I said, there are other aspects to consider. It is a very difficult topic. But a very interesting one!
  8. Oct 15, 2014 #7
    Thanks for the replies

    I think I agree Dennis that its hard to make an AI understand language but not impossible I think.

    Interesting stuff though.
  9. Oct 15, 2014 #8


    User Avatar
    Gold Member
    2017 Award

    Hi again, kzvrso. I am also interested in things like these, and I came to think about some more stuff regarding this, which might be interesting - I will post later about it. In the meantime I tried to find/remember what bot I was trying before, but I haven't found it yet. But I stumbled upon another bot, which I've just tried - it was pretty fun and performed a bit better than the one I was thinking about;

    Suzette: http://www.chatbots.org/chatbot/suzette/
    Chatbot Suzette Fools One Judge During Loebner Prize Contest
    Bot is here: http://ai.bluemars.com/chat/

    Here is a copy of the chat between me and Suzette, and it highlights some of the difficulties e.g. ambiguity, contextual issues and primitive/insufficient knowledge representation:

  10. Oct 15, 2014 #9


    User Avatar
    Gold Member
    2017 Award

    And here's another thing which I had in mind mentioning (a more advanced system than these kinds of bots):


    ...and a clip which describes some of complexities and difficulties involved:
  11. Oct 15, 2014 #10


    User Avatar

    Staff: Mentor

    Tells a lot about judges. I wonder if AI programs have some kind of reverse Turing test?

    Few days ago a friend of mine commented in some discussions that "well trained precarian* is indistinguishable from bot". Go figure.

    *precarian? precarius? No idea what the correct form should be, but a member of precariate.
  12. Oct 15, 2014 #11


    Staff: Mentor

    The judges were bots...
  13. Oct 19, 2014 #12
    Wow I'm surprised Suzette knew the answers to those maths questions. Not as bad as I thought, she even held a conversation with you for while

    Not perfect, but definitely room for improvement
    Last edited by a moderator: May 7, 2017
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook