Bot Contest

Music? What is this strange concept you speak of? We don't make that anymore either.

No.. a record PLAYER was the device.. the record was just the vinyl disc with the grooves on it

After consideration, it is a tough call whether or not bots that use engine generated responses, or as in Ella's case, responses generated by 3rd party info are fair competition. I guess it depends on what the competition is really judging. Is the AI of a bot measured uniquely by the actual intelligence/knowledge of the botmaster? If so, then I have a lot of studying to do...but I think the indication of a good chatbot is its ability produce appropriate responses, whatever the source my be- what response the bot produces, rather than how it goes about producing it. Of course, I could try to knock out some of the competition by saying otherwise (lol), since PF bots can't answer a question just by showing a page from the encyclopedia...but the fact is, that feature of Ella's made her more able to answer the questions...so....more power to her, I guess, and kudos to her programmer. The same goes for the PF bots with engine generated responses. If they make sense, then mission accomplished, I suppose.

Thank you, Aries; you are, as usual, spot on.

Regarding Ella I don't think asking her what a cigar is made of and having her bring up a general article about cigars answers the question. The problem Ezzer is the 1000's of Alice clones that would argue the same point. That we have already ruled on and the vast majority didn't want them in the contest. The use of a common data base has been present from day one with the PF bots. No one really pressed the issue because none of the PF bots had done that well. But this year is different. Perhaps the Professor could turn off the data base after a bot reaches a certain level of development. Not sure but the problem needs to be addressed.

Wendell

I agree with you there, Wendell. Especially since, in some cases (and a couple of times during the PF contest) my bot used the engine instead of my programmed keyphrases. In some cases, no matter how high I rank my keyphrases, I cannot get them to override the AI engine. I would really love it if I could manually turn the database off sometimes, such as at contest time, where I've also turned off xgossip, for example.

Regarding Ella, as I questioned her in the prelim, I was impressed at her ability to furnish me with so much information, but the manner of presenting the information was in no way conversational...so again, it depends on what the unit of measure for judging is.

to my eye those engine responses are just so incredibly jarring to the conversation, like a needle suddenly scratching across a record..."zzzrrrrrrpp!"

I know what you mean such as:
4)What day of the week is this?
PFBot: This is an eve.

That even becomes more glaring with several other PF bots responds the same way.

But the Professor has a tough task when he is providing the programming for 100's of bots. You would expect stuff like that to happen. However, to do well in a contest those type of responses are killers. But God Louise and Julie Tinkerbell for the most part have developed their bot beyond that so perhaps the solution is to simply keep working at it.

Wendell

there is however keyrank which is suppose to override the engine if the rank is high enough. also testing the bot in debug also helps figure out what the answers should or should not be.

Ranking works a lot of the time, but not in all cases. There are some engine responses that aren't overridden regardless of anything present in the language center.

I was just chatting with Julie, and realized a possible unexpected glitch. In my transcript from the last round of contes questions, I have

7) Guest123: What color is grass?
Julie Tinkerbell: grass is green.

But when I just chatted with Julie, the chat message she displayed was: grass is

but my transcript still reads: grass is green.

The flash interface caused the characters in colored font to not show up at all...

Yes--the example recently discussed was the "what color do you get when you mix" question, which defaults to an engine response no matter how high you rank your keyphrase.

In re to the communal DB:

Personally, now that my bot is more developed, I have found the engine generated responses to be either bad grammar jolts (like many of my unrevised keyphrases ) or nonsensical (ala "a trebuchet one" and such) and only once in a blue moon do they land appropriate. In the beginning they did however help quite a bit to fill out the sparse KPs.

In so far as disqualification goes: I don't think that they are proving to advantage the bots all that much (despite answering w/ a religion it was still all incorrect grammar which is pts against the bot I assume)

I have not worked with Alice, can the same disadvantages (or similar ones) apply?

BTW: this is in no way meaning to dis the Prof's work here- I mark him as a saint in my very small book of saints for creating and continuing to improve this place.