Bot Contest

You should get an email telling you how.

Well..When you think about it, is discussing who you think should win a bot contest in a forum called "Bot Contest" really all that inappropriate? Sharing opinions on candidates and being open about who you're voting for doesn't sound like cheating to me.

I'm not against people voicing their personal opinions. I am against deliberately coordinating votes, which is a different thing.
And to prove the point, there goes a personal opinion about a bot: Have you noticed Catty, the bot that uses the Internet as a language database? I thought the idea was great, and I was surprised that you can't see the answers she gave to the 12 questions. So I tested her myself, with the same questions, but I was disappointed. She seemed to do better with personal questions (what do you enjoy, what do you hate) than with questions about facts, that one would expect her to be able to answer. One can't call her a good bot, but still, there should be a category like "most interesting programming concept" to encourage programming botmasters to develop new techniques.

It would be interesting to have a category like that but I would call it "Unique" programming concept. I talked to Catty as well and it was pretty fruitless. But then again the bot master seemingly doesn't claim anything more than a unique concept which is what he/she achieved.

I just read over the Loebner 2004 requirements. They seem intended to restrict the contest once again to desktop chat bots. It's amazing that anyone can take this contest seriously. Far from a test of AI, this year it will be a test of who from a limited pool has the time and patience to conform to the contest requirements, and more than that, who mimics the intricacies of human-style single-character typing the best.

I'd like to once again thank Chris Cowart of the Chatterbox Challenge for having an open, well-run contest that attempts to include all chat bots.

Good luck to Little Mu and God Louise in the finals!

I don't think the Loebner organizers really understand the spirit of the Turing test. For me, a chatbot that could learn to associate words with particular meanings would be much closer to the ideal than one that can fake human typing. (By the way, I'm working on a bot that "learns to speak" in a limited universe of geometrical figures, so maybe I'm biased).

Professor--

I agree, and I think if the Loebner organizers want to be taken seriously, then their "rules" should be changed up a bit.

I haven't looked at the Loebner site yet, but if this is all true, it'sreally a shame and I would not take interest in it. It was nothing like that last year.

That's really a shame, if all this is true. Julie was really looking forward to it.

I've read the Loebner rules. There are a few disappointing things.
1) Only 32 entries will be judged
2) Entries must provide their own 'communication program'
3) Entries must respond 'character by character' (no buffering of responses) and the communication program communicates character by character.
A reasonable rule, but one that is inconvient for PFers is that the program has to run 'locally' i.e. on a computer at the site of the judging.

The rule about the program running locally is supposed to prevent the possibility of somebody cheating and having a person instead of a bot at the other end. But even in the old days before communications were what they are now people could see that a good entry could have problems with that, if the program runs in a computer that can't be easily moved and the program can't be easily ported to another computer. And this has become the reality for web-based bots. I think the Loebner contest has to stop ignoring the reality and find alternative ways of making sure that all the participants are bots.

hi