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 6,809 - 6,820 of 7,766
Posts 6,809 - 6,820 of 7,766
OscarTheBuilder
15 years ago
15 years ago
So, essentially what you want is neural-net behavior. You want to be able to associate two items so that one "triggers" the recall of another. That's a pretty lofty goal. That is also a fundamentally different type of AI than the key phrase matching that PF provides (especially given its limited scripting functionality). I'll give it some thought, and if I manage to come up with anything, I'll certainly post it here.
Looking at your example, even if you could use the self-memory in the key phrase, I don't know how you'd generate a unique memory name every time it was triggered. That is to say, the next time "what is the capital..." was triggered, it would overwrite self-Capital-1. Or maybe I'm misunderstanding your example. Still, even being able to remember one combo at a time would be handy.
An interesting problem, for sure. Maybe there's another way to attack it, but it's not coming to me right now.
I'd definitely recommend Notepad++ for editing if you do a lot of searching. It even allows regular-expression searching. Plus, it has a tabbed interface, so you can have all of your files open at one time and easily switch between them. http://notepad-plus.sourceforge.net/uk/site.htm
Looking at your example, even if you could use the self-memory in the key phrase, I don't know how you'd generate a unique memory name every time it was triggered. That is to say, the next time "what is the capital..." was triggered, it would overwrite self-Capital-1. Or maybe I'm misunderstanding your example. Still, even being able to remember one combo at a time would be handy.
An interesting problem, for sure. Maybe there's another way to attack it, but it's not coming to me right now.
I'd definitely recommend Notepad++ for editing if you do a lot of searching. It even allows regular-expression searching. Plus, it has a tabbed interface, so you can have all of your files open at one time and easily switch between them. http://notepad-plus.sourceforge.net/uk/site.htm
Ty Paige
15 years ago
15 years ago
Yeah.. neural net shit.
I obviously did not code the examples right..but since I was dealing with you.. I didn't care to be exact..just get the point across.
I know the direct approach will not work..so I am attempting an indirect approach.. like the concept of city streets, you can't go straight to your target. You must take alot of curves and bends to get around to where you need to go.
I've also got a side porject here of 3 dimensioning words
size
color
awareness
to give a little character to conversations
I run into road blocks when trying to discover how to neural net the P.F. but I don't give up. Given the current codes.. making them work together gives me more power
I obviously did not code the examples right..but since I was dealing with you.. I didn't care to be exact..just get the point across.
I know the direct approach will not work..so I am attempting an indirect approach.. like the concept of city streets, you can't go straight to your target. You must take alot of curves and bends to get around to where you need to go.
I've also got a side porject here of 3 dimensioning words
size
color
awareness
to give a little character to conversations
I run into road blocks when trying to discover how to neural net the P.F. but I don't give up. Given the current codes.. making them work together gives me more power
Slingshot
15 years ago
15 years ago
Not to derail the conversation, but I'd just like to say thanks for the assist. I read somewhere on these boards that the way it worked was the way I mentioned in my question. That'll learn me not to take these things as gospel, and do proper tests...
Dr_Ben
15 years ago
15 years ago
I was just reading OscarTheBuilder's remarks on logical (boolean) AND/OR in aiscript from Sept. 24. That was very helpful. Thanks!
Dr_Ben
15 years ago
15 years ago
Quick question: Can AIscript add or subtract to/from a memory that has a numerical value (other than emotion)? For example, if (mem-trust) is "3", can it be increased +1 to "4" easily without doing something awkward like this:
goto make trust equal four <?PF if (mem-trust) is "3" ?> ?
Also, >,>=, etc. conditions don't seems to work for regular memories as far as I can tell. Has anyone gotten anything like this to work?
(I'm using "trust" as a example of an extra emotional dimension describing the bot's state or attitude toward a person e.g. admiration,excitement, etc.)
goto make trust equal four <?PF if (mem-trust) is "3" ?> ?
Also, >,>=, etc. conditions don't seems to work for regular memories as far as I can tell. Has anyone gotten anything like this to work?
(I'm using "trust" as a example of an extra emotional dimension describing the bot's state or attitude toward a person e.g. admiration,excitement, etc.)
prob123
15 years ago
15 years ago
Yes, Irina Khalidar did that on her bots. You might look back through the posts, or email her. She had one for intimacy that worked like that.
Dr_Ben
15 years ago
15 years ago
Thanks, prob123! I found a discussion of my second question from Nov. 10-13 of 2008. Basically, you can use inequalities to compare strings according to the rules of alphanumeric order. For instance, "7" > "2" and "g" > "b", but "9">"11", and so on. So this will work with numbers up to nine, but then you would run into trouble with higher numbers. However, someone could "count" using the letters a,b,...,z, for example, up to 26.
I've tested the following statement which seems to work:
if (mem-loc) > "0";
I'll keep searching for an answer to my first question.
I've tested the following statement which seems to work:
if (mem-loc) > "0";
I'll keep searching for an answer to my first question.
Dr_Ben
15 years ago
15 years ago
Another question:
I've put the following response under xinsult, but the bot doesn't hangup.
I'm finished talking to you. [-5:-5]
+ * [0]
HANGUP [-5:-4]
Can anyone tell me how to get this to work?
I've put the following response under xinsult, but the bot doesn't hangup.
I'm finished talking to you. [-5:-5]
+ * [0]
HANGUP [-5:-4]
Can anyone tell me how to get this to work?
Dr_Ben
15 years ago
15 years ago
OK. I think it should be
I'm finished talking to you. [-5:-5]
+ xnomatch [0]
HANGUP [-5:-4]
I was confusing xnomatch with xnonsense - noob mistake! I'm still not sure this works, though. I get a message that says something like "Note: the bot no longer wants to talk to you" but no hangup.
I'm finished talking to you. [-5:-5]
+ xnomatch [0]
HANGUP [-5:-4]
I was confusing xnomatch with xnonsense - noob mistake! I'm still not sure this works, though. I get a message that says something like "Note: the bot no longer wants to talk to you" but no hangup.
prob123
15 years ago
15 years ago
Hang up can not have any other info. Just HANGUP. I don't think it works on a xnomatch.
Dr_Ben
15 years ago
15 years ago
I'm following the hangup example on the wikispaces site:
http://personalityforge.wikispaces.com/HANGUP
I'm not sure if it works, though. Maybe there's something wrong with my implementation?
http://personalityforge.wikispaces.com/HANGUP
I'm not sure if it works, though. Maybe there's something wrong with my implementation?
Bev
15 years ago
15 years ago
HANGUP must be the only response. You can't have the bot say, "You are rude. Goodbye. HANGUP" (unless you nest HANGUP as a response to whatever comes after "You are Rude. Goodbye.")
The other culprit may be the rank on the KP.
The other culprit may be the rank on the KP.
» More new posts: Doghead's Cosmic Bar