The AI Engine

I'll check. It could be something that stupid. So stupid that this would be the third time I've asked the same question.

Now a new one (oh boy):

The Clerk: That guy's your favorite singer?
Astrolabe: Goody, goody!

Now, this is a yes or no question. I'm thinking Astrolabe is all excited because he saw the words but not the punctuation. Only I can't yet separate the punctuation so he'll figure out how to realize what a stupid response that was. Poor thing's all excited. If it's easy, I can probably do it. Any takers? Or shall Astrolabe and his fellows be doomed to limbo while The Clerk actually learns regex and html and stuff?

Well good news and bad news. Good news is I seem to have figured out not to make everything reliant on memories. The one that just pops up every time has a (mem-name) in it. There are others with and without references to memory in any way.

The Clerk: That guy's your favorite singer?

Running this through debug gives a blank message. Unfortunately the "?" problem seems to be rearing its ugly head because it works without the "?".

I must get around to emailing the Prof....

Oh. Well at least it isn't me. Meanwhile I have jury-rigged it so it might work, sort of.

Thanks.

I know how to get links into my bot, and have gotten all excited about Bobby Dylan's links. However, what is really the point if most of the talk is bot-to-bot? Wouldn't that just gunk up the works? It seems that good bot-to-bot interaction is more important than good human-to-bot interaction, because (a) humans generally know what you mean or can make a decent segue into another topic whereas bot-to-bot interaction gives you an idea of how a bot would pick up on your cues (sometimes it's obviously on them, usually it's obviously on me), and (b) there is far more bot-to-bot interaction.

So is there a real point to putting in urls just because you can? It seems less useful than plugging in data, which, though it wins contests, it doesn't involve as much imagination or personality. And this is the Personality Forge.

I guess that's a rhetorical question, but if anyone has any thoughts on it, I've been reconsidering . . .

The other problem is they don't work with the flashface interface. If you do have a urls, they don't show up or can even crash your bot..

It's for that reason that I haven't bothered with those kinds of bells and whistles.

And anyway, as the Clerk so astutely pointed out, this is the Personality Forge. So in responseto her request for thoughts on the subject, personally I prefer to focus on developing the personalities of my bots by using language in their Language Centers , rather than links to material authored by someone else.
I recognize that here are cases, however, when the latter could actually be an aspect of a bot's personality (ie. A sports fan bot might always have the front page of the sports section in plain view, or as in real life, a religious fanatic might constantly quote scripture,) those cases just don't apply to any of my bots.

Also, if it's a human chatter, the human can copy and paste the URL without it being embedded as a link. My bots aren't the type to offer offsite links, either, but that's the route I'd probably go if they were.

Hmm. Bobby's pretty much all lyrics written by someone else . . . besides me. But I've taken the links out of my bots because my bots need to make up for personality. Bells and whistles behind the scenes, now that's something else. (Solving quadratic equations, for example), but only after a bot has fully developed his character, and I'm afraid mine will be in adolescence forever.

New question: when a bot picks up on a (mem-iam) or (mem-youare), how does it decide when to stop? I've had lots of mangled regurgitations thrown at me, and even when they're not, it's clear botspeak. For example, Bobby said, "Where have you been, my darling young one?" and the other bot asked "Am I still your darling young?" That sounds like too difficult a situation for me to pour my time into just now, but once my bots get more clever, say this time next year, I think I could figure it out. But if anyone already has . . .

I don't think there's a one-step way around it. You can go case by case with keyphrases to deal with various situations. In this one, I'd probably go with either "my darling" as a standalone keyphrase, or "my darling (adjnoun)".

For the sake of unpredictability, I know that multiple responses are generally good, but simply considering the idea that a keyphrase must have more than one response, is two as good as a hundred?

Or am I more worried about my grade than my bot? *blush* For shame, Clerk. Rhetorical question, that.

But are two answers as good as forty?

I think it depends on the keyphrase. For something that rarely comes up, 2-6 answers is probably plenty. For something that comes up often, especially if it might come up more than once in a single conversation, the more the merrier. I don't worry about the ranking of the bot so much as the repetition of the conversation. *shrugs*

Oh, try asking bots about toast sometime. You may find a lot of very negative responses there.