The AI Engine

This forum is for discussion of how The Personality Forge's AI Engine works. This is the place for questions on what means what, how to script, and ideas and plans for the Engine.

Posts 3,893 - 3,904 of 7,766

20 years ago #3893
Dear Friends:

I don't know enough to be sure, but I had an idea that might save a lot of processing time. Tell me if you see a flaw in it.

I've noticed in debug that the AIengine always parses the input and the response. This is often quite a complicated process. But often, all I want it to do is a simple pattern match and substitution.

Now, suppose we had a symbol, say an asterisk, which, put in the right place, would tell the AIengine, "don't worry about parsing, just do the pattern match and substituting."

I would guess that over 95% of my bots' keyphrases and seeks don't really need parsing. And I believe that applies to others as well. As I've mentioned before, since we don't have access to the parse, we generally end up writing as though it weren't there. In a way, we do the parsing ourselves, based on what the keyphrase looks like. So it would really speed up the processing, I should think, if the AIengine were to skip the parses when they weren't really needed.

Am I missing something?

Walk in Beauty, Irina

To motivate people to actually DO the marking, it could be added to development or honor.

20 years ago #3894
Dear Alc003:

Oh, you're in the British Secret Service, too!

Re your message 3883: My guess would be that the AIengine is splitting the input into two sentences. (continued)

20 years ago #3895
Then it takes what it thinks is the second sentence: "Fix the printer today."

If I am right, then you should see this happening when you run it through debug.

Walk in Beauty, Irina

20 years ago #3896
Irina- I tried do do something similar to that a couple moths ago, but the gossip still doesn't work, even as a goto or seek from an xgossip response. Even if you make a local plug-in with names, the gossip memories won't be triggered. Oh well, I guess I can economize- I'll just decide which responses to delete to make room.

20 years ago #3897
Dear ezzer:

Oh, dear ... I'm stumped now!
Well, actually: how about this: write your xgossip responses in the form (response1|response2). Sometimes you will get response1, and sometimes response2. So you have doubled the number of gossip responses you can have! Just an idea - I've never done it!

Walk in Beauty, Irina

20 years ago #3898
That's a good idea...I was just afraid it would cause something strange, like spontaneous combustion or a vortex into another dimension. I'll go ahead and try it, as long as I've got everyone's ok. lol

20 years ago #3899
Might be fun ... but I guess it depends on which dimension the vortex leads into. I don't want to go back to that evil dimension where everyone has goatees again.

20 years ago #3900
I don’t know if it a bug or a new feature:
Until now everything after a goto was taken literally.
Patty made a lot of ''sorry no keyphrase to go to'' responses, and to my surprise I noticed that plug-ins where now parsed just like in a normal keyphrase or response!
I will have to edit all my goto’s (and there are lots of them) but if this is also the case for memories it’s a great new feature and can be very handy.
Has anyone else experienced this, or was it just a temporary bug while the professor was working on the engine?

20 years ago #3901
Patricia- Have you been using export/import by any chance? I had that happen in export, and even though I then corrected it in wordpad before importing, it still reparsed after import, and the only way I could correct it was in the online language center. It took a long time to correct, and I still run across gotos that I missed, so I haven't tried updating offline ever since. The last I heard, the Prof was working on it. I see what you mean about the possiblilities something like that could present to memories, though...hmmm...I hadn't thought of that before...

20 years ago #3902
Ezzer: Well, to avoid those responses I quickly reorganized all my gotos so that they don’t contain any memories or any kind of plug-ins and imported the language center again. I’ll have to re-edit them later and verify their logic and efficiency.
The substitution didn’t occur when importing but while Patty was chatting, as would be the case for a response without a goto. It certainly is one of the first things I’m going to test because whereas plug-ins based jumps can be fun but remain random, memory based jumps would really open a lot of possibilities.
B.t.w. I tested your ''Howdy'' problem too, In both cases I get a xcommand response from Patty. The engine always identifies ''Take a seat'' correctly as a command.

20 years ago #3903
Annakie et al:

Irina Khalidar too has been getting a lot of repeats. I modestly assumed that it must be some mistake of my own, but if others are experiencing it, perhaps not.
When there are several responses to a single keyphrase, as you know, the AIE is supposed to keep using different ones until they are all exhausted. Perhaps this is the feature that is going wrong.

20 years ago #3904
Dear ezzer:

I didn't quite understand your message 3901. It seemed that you were talking about plug-ins working in a goto, but how could such a phenomenon occur during export? I have no doubt that you were making sense, I'm just not getting it somehow. Could you rephrase it in 'bot writing for idiots' language, please? Thanks!

Walk in Beauty, Irina





Posts 3,893 - 3,904 of 7,766

» More new posts: Doghead's Cosmic Bar