I probably will. Most likely at the start of March. The lack of entry fee and sponsorship makes it easier.

I would like to get to version .004 before I do, but it is hard to say how much of what I have been working on will be ready for prime time. The hardest part is integrating it into the conversation flow. Last year Skynet-AI took 4th place. If I enter, I need to revamp a few things (like text speak) to perform better.

I have enjoyed the competition in the past and have entered a number of bots over the years. It has always provided good feedback. Some years, especially last year, judging has been a bit spotty. The quality of the competition is influenced heavily by the question selection and who the judges are. The original creator of the contest is back, so it should flow a bit easier in 2011.

There are a few new wrinkles this year; all AIML bots are eligible so you are competing with the “Alice” knowledge base, text speak looks to be more prominent, $1000 grand prize.

I would like to get to version .004 before I do, but it is hard to say how much of what I have been working on will be ready for prime time.

I know what you mean. I also wont get everything working that’s currently in dev stage. Wendell told me that they wont be downloading/trying anything until the start of the contest, so I figure I’ll upload the latest version the day before the thing starts. I’ve already entered, just to make certain I can’t chicken out.

The quality of the competition is influenced heavily by the question selection and who the judges are.

Very true, I can easily come up with 10 questions for which my bot can get 3 points (4 is somewhat out of my reach at the moment, I think). On the other, it’s a lot easier to come up with 10 for which it gets 0 each
That’s why I think a more ‘extended/unified’ test might be interesting, like discussed in the other thread.

Jan Bogaerts - Feb 8, 2011:

I can easily come up with 10 questions for which my bot can get 3 points (4 is somewhat out of my reach at the moment, I think). On the other, it’s a lot easier to come up with 10 for which it gets 0 each
That’s why I think a more ‘extended/unified’ test might be interesting, like discussed in the other thread.

One of the reasons I originally entered was because I needed to broaden the types of input and I wanted to focus on the question answering task. In years where the question topics were to narrow, I have had bots do poorly. One bot I entered had no personality built in. That year many of the questions related to the bot persona. Who wins and how well you do is a bit of luck depending on if the questions map well into your bot’s capabilities.

Even with all its flaws over the years, I consider this contest the best in the industry. It is the most open and easiest to enter. Unlike the Loebner Turing test you don’t need to work on the mechanics of how the bot communicates. Make sure though that your bot can give the judges a log. It helps make the contest easier.

We could use more bot contests, but it takes a great deal of commitment and effort. Wendell has done a great job of keeping this contest alive every year over the last decade.

I finally (in large part, due to Jan’s prompting) sent in my entry request for Morti. As there are still several technical issues that I need to iron out, I’m not certain how well he’s going to do, but I’ll keep at them right up to the contest. Last year at this time, I was very hopeful to have Morti on not only a different platform (AIML is great to start off with, but somewhat limited, IMHO), but also well on his way toward an entirely different level of comprehension, as well. Perhaps I’ll need to re-evaluate my goals, and work on a longer list of smaller achievements. Oh well. Morti’ll get there, I have no doubt. It’s just going to take longer to do so, is all.

As there are still several technical issues that I need to iron out, I’m not certain how well he’s going to do, but I’ll keep at them right up to the contest.

Same here.

Dave, I have the same problem. I have many big plans, but find myself going back to fundamentals again and again and refining them. Although the improvements I’m making are important, I think, they certainly aren’t as exciting as my grandiose ideas.

As a contest veteran, I do recommend entering and letting your bots out in the light of day. It is good to have a deadline, and it can shine a light on weaknesses you never knew existed. One of the first things I stumbled on was the fact that other people’s language patterns were different from my own enough that things that the bot should have understood were not recognized.

Dave, I agree with your comments about AIML. It provides an excellent foundation (especially for casual botmasters), but some of its limitations make it hard to reach “understanding”.

That’s what I’m hoping to overcome at some point, with the insights and ideas that I’m garnering from everyone here, Merlin. The concepts and philosophies that I’m reading about here in these forums are allowing me to map out an overall scheme or outline of what I want Morti to become, and, in a general sort of way, how he should get there. It’s not my goal to provide the best Synthetic Intelligence example in the world, but to have one that can just carry out a conversation that can hold a person’s interest for hours, if need be, without ever feeling “artificial”, or “contrived”. That’s not “shooting too high”, is it?

Dave Morton - Feb 8, 2011:

It’s not my goal to provide the best Synthetic Intelligence example in the world, but to have one that can just carry out a conversation that can hold a person’s interest for hours, if need be, without ever feeling “artificial”, or “contrived”. That’s not “shooting too high”, is it?

I think it is a great goal! I can see glimpses of it myself in some of Skynet’s conversations. With hard work and good input we may all be able to push this field in a direction that makes our bots much more exciting.

One of the elements of any good AI is self awareness.

I am testing Skynet-AI’s sense of self. You can ask it questions here. What types of questions should a self aware AI be able to answer?

I have entered Skynet-AI into the 2011 Chatterbox Challenge.

It is hard to say right now if I will be able to code and test version .004 before the contest begins. In the mean time, I am trying to also expand the coverage and capabilities of the current on-line version.

Merlin - Feb 9, 2011:

One of the elements of any good AI is self awareness.

I am testing Skynet-AI’s sense of self. You can ask it questions here. What types of questions should a self aware AI be able to answer?

One thing I’ve noticed with ALICE and friends is that they can’t delve very deep into their own “likes”. If a bot has a favorite movie or book, I’d like it to be able to discuss (or at least name!) the main characters, why it liked said movie/book, what its favorite part was. All standard questions that can expose the personality of the bot and steer a conversation to other aspects the bot’s “self”.

Best of luck, Merlin. May the best bot (read: Morti) win!

C R Hunt - Feb 16, 2011:

One thing I’ve noticed with ALICE and friends is that they can’t delve very deep into their own “likes”. If a bot has a favorite movie or book, I’d like it to be able to discuss (or at least name!) the main characters, why it liked said movie/book, what its favorite part was. All standard questions that can expose the personality of the bot and steer a conversation to other aspects the bot’s “self”.

Actually, CR, that’s entirely doable, if you’re willing to put in the literally hundreds (or even thousands) of categories to support it. Through the use of <topic> and <that> tags, a rich, full dialog can be constructed to discuss characters, plot twists, scenes, or any number of aspects of a movie or book. The problem is, during the process of creating the dialog, and all of the possible directions the discussion is likely to take, you may start talking to yourself, and run the very real risk of being tossed into the Loony Bin.

Ha ha, fair enough. But it’s a real show-stopper when a bot introduces any topic and then can’t seem to follow through. It’s one thing if I pick a subject outside the bot’s expertise, but another thing if the bot itself leads the conversation astray! Not sure what the solution is. (I’ve never even directly worked with AIML.) Just something I’ve noticed talking to Alice.