The AI Engine

New subject/question: I'm a new designer and I'm looking to do some tricky sorts of things. I'm approaching this from an AI theory perspective, wanting to customize the AI engine to spec. Mainly, I would like for the bot to control what keyphrase is chosen, instead of the AI Engine; and I would like to accomplish this without using hard-coding like ranks. Put another way ...

The way I understand the AI Engine, it assigns each word or phrase a value, adjusts for ranks and emotion, and then picks the highest value as the keyword. Then the bot can use AI script, seeks, memories, etc. to arrive at a hard-coded response. Is this correct? If so, I would like the bot (not the AI Engine) to do that job. My idea is simplified then as this: bot picks keyword(s); bot picks response to keyword. Is that possible in this setup?

For example, let's assume the input is: I was sad yesterday, but I am happy today. I would like the bot to sometimes respond to happy, sometimes to sad, sometimes to happy today, sometimes to sad yesterday (and maybe even to I am or I was). Any ideas/suggestions/comments? If you want, feel free to email me at jdearles@gmail.com (but please not if you are going to spam bomb me!). Thanks!

i spoke to jabby wacky lastnight do u know hal no i dont then she says i should care but i dont

Clerk, sorry I pasted in the wrong link. I have Firefox so it took me to the annoying Google toolbar that installs itself with everything. Did you find Firefox?

Unoriginal, you can have the bot respond according to emotion or time of day or other things in the PF script you add to your bot, but you would be using the PF engine. Take a look at the advanced section of the book of AI.

Bev, Yeah, Firefox did solved the problem, so I'll use it at least for the PF. Thanks a lot -- I never would've used a different browser. There are a few ways in which IE is better, at least since I know a little about it, but they're not too important -- like I can't click on the emoticons below. Well, I always hated those things to begin with.

New question. Re-read the book of AI again and again. Not that it's not helpful. But Scrivener's transcript showed the question "Who is Boswell?" (Boswell is his partner, okay, and he has this information in his memory.) But I hadn't gotten to "Who is" questions and tried a Keyphrase of:

^who (re) (verb_be) (person51541)

where person51541 is someone Scrivener is familiar with and willing to admit it. That works all right, but when I try a conditional AI expression to root out the person he's talking about, like

<?PF (re) if (person54541) is "Boswell"; ?>

it won't take it. Is there a way to use conditionals for this so I don't have nine thousand keyphrases for different people?

I considered making "Boswell" a keyphrase, but then I would run into the same problem with the conditional, trying to figure out what context Boswell was being mentioned. So. Shouldn't I be able to do a conditional one way or the other?

Don't anybody go off shooting themselves over this. Or me.

Bev,

thanks for your response. i have read through the whole AI book (twice or three times). my goal is not to have the bot respond based on conditions (emotions, time of day, etc.); it's to have the bot control which keyword is chosen. do you (or anyone else) know if that is possible?

thanks again for your help

jared

... or another idea. is it possible to use AI script to affect ranks? Can I let the bot change different keyword ranks?

Unoriginal, the only way to have the PF bot "choose" is to use the custom PF scripts and rankings. If you are on the PF you are using the PF engine. So far, the Prof does not seem inclined to make that open source or share control of that.

How much did you change the ranking of the phrase you want the bot to use? You can get the same effect as what I think you are describing by playing with the scripts and other features described in the book of AI (with a few things that are hard to override). Otherwise, you have to just build your own bot.

Bev,

I haven't started building the bot. I've just been testing. I am developing an architecture for the bot, primarily along theory lines. Unfortunately, I don't have enough programming knowledge to build from scratch. Scripting, however, I can do.

My goal goes something like this: for the bot to test user input versus previous bot output, dynamically choose an appropriate keyword(s) based on configuration (the two-step process I have dubbed the "understanding" function), "think" about it (run a set of subprocesses as the "experience" function), and decide how to respond (another set of processes I call the "deciding" function).

So, now I have a couple of options: a. Use the keyword the AI gives me as the input for the "understanding" function; b. make several catch-all keywords with uber-high ranks, and use scripting to analyze all the pre and postkeys and use that as my input.

Which then leads me to another question: if the engine goes to xnone, is there a way for the bot to find the input?

Bev, thanks again for your responses and help. I really appreciate it.

Jared

Unoriginal, you will always be using the PF engine for a PF bot, so the "bot" will not be doing the things you want it to do, at least not in the way you describe it. However, you can script things into the bot to make the engine respond to input differently depending on rankings of key phrases, memories or other conditions. This will give you tyhe same effect if you work at it.

I would not make your general phrases uber high. You don't want them to override more specific responses. I suggest keeping your rankings low to start and raising phrases that you need to have override the system as you go along. That is up to you though.

If something goes to xnone, it means you don't have a key phrase that matches the input, and the only way to make the bot find the "right" KP is to re-write your KP or add it. The debug mode may help you on this.

Although thinking is good, try tinkering with a simple bot and then re-read the book of AI. it makes more sense after you play with it and see what actually happens. I could be misunderstanding your questions, but I get the impression that you are misunderstanding the relationship between the bots here and the AI engine. In essence, if you use memories, the AI engine tests against previous chats for memories you put into your bot. Of course, now that all guest chatters are the same entity, the memories don't work half as well as they used to do, but that is as close as you can get to the type of thing you are describing, at least as far as I know.

I am going to ask a question that may totally destroy my bot builder credibility: what is the difference between scripting and programming? Aren't they the same thing?

I think I understand not being a programmer but saying you can work with some scripts, but it still seems confusing to say "I can't program but I can write script." (I am not picking on anyone, I just want to use words correctly). I empathize with being able to copy and paste and adjust small parts of other people's script because that's how I do the bits of PF script, HTML or other tiny scripting projects I use but don't really comprehend. (Lately I've been cutting and pasting Linden script--in a virtual world you can literally make thing go 'splodey when you mess up!).

Does that make me a "script kiddie? What about Unoriginal, is he a PF script kiddie? Can an old person be a kid simply by lacking a specific skill? Inquiring minds want to know.