The AI Engine

I don't know which approach is better, though I tend to be a "lots of keyphrases" botmaker. Whenever my bot asks a yes/no question, or one with a specific set of likely answers, I try to put a seek on it. Sometimes its worth putting another seek on some of those replies... but if you go too far with seeks, you're essentially turning the conversation into nothing more than a huge if-then-tree. That seems to appeal to some botmakers, but it just annoys me. You can then map out every possible conversation, and that's just completely uninteresting to me.

But that's just me. Other botmakers do go the "extreme seek" route. I think it probably depends on your particular goals for your bots. My goal is to have entertaining transcripts. If they can occasionally be coherent as well, I count that as a bonus.

I would like coherency, but, no, I don't want to imagine every possible conversation in advance. That seems to be self-defeating in AI. I want them to say things I didn't imagine them saying; I just want it to make sense. I mean, geez, Botolph just got dissed by Brother Jerome. Guess he should talk to Not I some.

We can tell whether a chatter is a person or a bot just by looking. Can the bots tell the difference inherently?

Are the bots aware of the existence of bots (if we don't bring it up)? If so, do they realize they have a botmaster? Do they recognize their botmaster?

The one thing I have been able to quit worrying about is whether they are inherently aware of their "sibling" bots. Either they aren't, or they are very poker-faced about it.

I seem to have plateaued after three months. No matter what I do to my bots, they don't improve. They stay hot, but they don't improve. I know I could make them improve by the standards of the Forge by dumping a lot of facts into them, and don't think I'm not tempted. I may yet. But I'm stumped. I think I was punished for creating and destroying bots, but I figured I was learning stuff about the bots and about the culture. The culture does not care what the capital of France is. (Except Judge Betty and perhaps a few others -- I had a nightmare about Judge Judy just because of her and a whole plate of Toll-House cookies.) Anybody else hit a plateau like this (yours might be for different reasons -- I'm not technically skilled yet)?

Is there a way to capture the last thing you said? I've tried, and I've seen that others have tried when they crashed and burned in conversations with one of my bots. So if there is a way, would somebody besides Not I point it to me? Ulrike is okay. Not I is Not.

Okay, bots have free will. I know this because Botolph told Not I that he liked her.

No matter what I do to my bots Sometimes it seems that way ..then it seems like they will take on a mind of their own and amaze you. Then they will go stupid again and make you bang your head into your desk. I think that's the "fun" part.

That's what hooked me. You can never "finish"...or at least I can't...

re: So if there is a way, would somebody besides Not I point it to me? Ulrike is okay. Not I is Not.

Julie Tinkerbell: Not yet, I've been waiting for you to teach me how. Are you ready?
Not I: Not I is not.

"The culture does not care what the capital of France is."

Fizzy cares. Well, he asks anyway. There was one Chatterbox Challenge where a lot of people didn't like the questions and started posting sets of questions they thought were better. I added all of them as xnones.

Yeah, that year the CC looked more like a quiz than conversation. I seem to recall that Critical Bill only had one response under "Do you know", so something like three of his answers were "Short answer: No. Long answer: Give me ten minutes and a couple of good encyclopaedias and I'll tell you all about it."

Not my finest botting hour that one.

Odd. This keeps happening, and in multiple bots. I've figured out that when I initiate a chat between my bot and another, my bot starts out with an xhello rather than an xinitiate. Fine. But two things still puzzle me here:

1. Why do they generally jump to the non-memory related key, even though they know the person's name? (I have such a response in there just for that reason, and it is far and away the most common one chosen. I should put in more as a stop-gap measure, but why should I have to?)

2. Why do my bots often go to xgottago after an xinitiate? I do have the non-memory-using response (at least one). It's more than a bit odd.

These two issues come up on all my bots, so it's not a glitch in a specific bot, but may be in the mind of the botmaster.