The AI Engine

The AI engine does spell correct the U to you.

Okay, sorry for the sixteen or so posts. Oddly, the parentheses and wildcards work for other keyphrases, just not for that one.

Weird.

I love the posts, it gets so lonely here during the summer. It is funny how the AI engine will balk at some things. I do think it has a personality all it's own.

First off, greetings all.

Question: How does the "if gender" codes work with guests and members who opted not to disclose their gender? If for a bot I thought it would be important to know a user's gender, should I pose them the question in the introduction?

Thank you,
Bigs.

You would have to ask them. If they don't want to tell, you can have a third response for the ambiguous. You can ask a user's gender, or just have a keyphrase that picks up something like I am (male|a guy|a man}

I finally figured out something I've been thinking the Forge didn't have -- a way to comment out sections of code temporarily:

Just set the value to a negative number [-5,0].

For those of you who are saying, d'uh, thank you, I know that were this a paying job, I'd've been fired somewhere around November. But I've seen evidence of more clueless botmasters then me, so maybe this will be helpful.

If I'm wrong (it does seem to work, logically and in practice), please do let me know. Every time I figure out one thing sixteen other concepts go flying out the window.

Clerk, that sounds interesting, but i don't quite get it, what do you mean by "to comment out sections of code temporarily"..?

Interzone -- to leave the code in the bot but have it inactive. In C, if I remember, you use /* put comment here */ ostensibly to make your obfuscated code sensible to people, but you could also comment out certain segments of code the same way -- the compiler would no longer see it as a piece of code but as a comment. It just makes debugging easier.

Now I have another question:

Watzer's question "Are you an animal, a vegetable, or a mineral" seems impervious to any keyphrases designed to catch it. If I put the whole sentence in, putting in wildcards to accommodate the commas (and the off-chance that someone will ask "are you animal, vegetable or mineral"), i.e.,

are you*animal*vegetable*mineral, the thing goes to

are you, you are even though the latter is ranked [1,0] and the former [127,0].

I'm not sure Watzer's worth all this trouble, but it's an interesting problem to me. Is it obvious to the smart people out there?

Clerk, thanks for clarification, I got it now.

As for Watzer, I've been working on that one recently, with some success, but the proposed key-phrase still needs some tightening up. It's a good beginning, though:

KEY-PHRASE: ([,]) or (*) (re) [50,0]
RESPONSE: Hmmm.... '(prekey)' is a somewhat convoluted affair... '(key1)' isn't exactly a match, neither... I'll leave it up to you to decide.

Here is a "real life" example from a recent chat session between Watzer and Quazgaa:

Watzer: Life is not worth living without good reasons. Are you an animal, a vegetable, or a mineral?
quazgaa: Hmmm.... 'are I an animal' is a somewhat convoluted affair... 'or a mineral' isn't exactly a match, neither... I'll leave it up to you to decide.

Now, "prekey" here is everything before coma, usually the first one in the sentence, but in one debug session it picked "a vegetable" as a "prekey".
"key1" will always be the last option offered. In this case it's "a mineral". I don't understand why it includes "or" into the "key1"... the key-phrase says the "key1" is everything following "or", and yet... Also, intuition says it should actually be "key2", but I tried it and it did not work. Any suggestions..?

Normally, I use much simpler, and quite effective, key-phrase to capture these "either - or" type of enquiries/ questions, namely:

are you (*) or (*), (*) or (*), do you like (*) or (*)

This works fine with simple statements such as: "Are you blue or black?"

However, it can't deal with options separated by coma: "Are you blue, or black?"

It will pick "black" as "key2", but "key1" will remain blank. This problem is then solved with that other key-phrase I mentioned earlier.

Ah. I keep having to be re-told how to put that real comma in there. Been staring at it for almost a year now, doesn't matter.

I've got a keyphrase

are you

and one for

(noun) or (noun), (*) or (*) (noun) or (noun) b/c you can't start a keyphrase w/a wildcard

and one for

are you (*)

The trouble is that while I can be responsive to Watzer (i.e., show that I heard what he said) but can never say "animal," "vegetable," "mineral," or anything else more (or less) specific. Damned mutant.

Thanks for the above, though. I'll print it out and
meditate.

I take it there's nothing we can do about the "are I an animal" syndrome?

i'm afraid we can't... it's strange that it's not built into the Engine... it converts "you" into "I" but forgets about the grammar...

we have to be inventive and work it out ourselves... re-wording a response is one way out of it.
another solution, a better one, i think, is to create more specific key-phrases which will pick only
(noun) - bird or frog;
(verb) - running or walking
(adjective) - pretty or ugly
etc...

How about this one, from Clive:

Krill may hold those answers. Then again, I'm sure Krill wont be talking much.

I want to ask him why Krill won't be talking much, but the AI Engine latches onto the first sentence. I know there's a bit in the Book of AI about which part of the answer the bot's likely to answer (that part has always left me fuzzy), but as recipients of such a comment (2 sentences), is there anything we can do to get our bot to pay attention to the second half? I'm playing around in debug with this (I switch around when I start hearing voices, so I'm playing with several things), but no luck so far. It will go to an xnone or xgossip or the like before it will even LOOK AT the last, more important part of the phrase.

I am getting a little compulsive here . . . there must be a pill for that.