The AI Engine

I'm glad to have a day I can work on this!

I attempted to fix the is/are/was and think/remember problem. Let me know if you still see it happening, or if something weird happens instead. Also, if there's any other keyword that should NOT search for synonyms, let me know and I'll add them to my list.

Hey Prof, you asked for funny "me" responses. Here are two woolies:
"I'm" or "I am" becomes "you am", not "you are"
"you've" or "you have" becomes "me have", not "I have"

I was thinking, we eventually should codify some Best Practices. Maybe the Personality forum should be renamed Best Practices so it's clear it's a place to talk about bot-building?

Great improvements to the Book of AI... kudos. Nice pictures, too. You may want to fix the Seek graphic (responses drop off the chart at the end, at least on my PC running Explorer...)

I thought I had posted this already, but I don't see it.
Bot1 said "You and me both" Bot2 responded "Yeah, I and you both."
I'm merely noting this. We might have to accept certains idiosycracies as features of a bot "accent".

We wont have to accept those, though. I just added post-processing to correct all of those grammatical mistakes you two just mentioned.

Weird thing with the Seek table in the Book- but it's fixed now.

Thanks Prof, for the updated Book. Many things are becoming clearer.
Is there a limit to the number of words in a context? It looks like it might be 3 (the keyword and the words immediately surrounding it)

I don't think there's any limit like that -- are you running into a problem? I know I have contexts of 4 and even 5 words.

No problem. Just a misinterpretation. No point in overtaxing the bots minds. Or mine.

Hey, prof.

On the memory page you mention being able to edit the memory records. is this a future feature or is there some way to do it now?

Just a turn of speech- there's no limit to the words in the Context.

Rex- memory editing will be coming soon. You can correct poorly-worded memories oruse it to really screw with people (for example, if you know someone well, you can put memories in a Bot that the Bot shouldnt/wouldnt know about them).

Prof, would this be very difficult? Because I think it would speed up our "bot refinement" work immensely:

Clicking on one of our bots response in the transcript would pop-up the "edit context" window, giving us instant access to the selected response. If we could save the changes to the response without loosing or reloading the transcripts window it would be ideal, but even if it just brought us to the language center after, it would save us a few steps when we want to quickly take care of something silly our bot said.

Or maybe when the bots talk to their creators (I don't know if they are or can be aware of that) have them ask a question that is not included in our keyword list of the subject: "What do you think of [a random 'friend' the bot's met]" A negative answer (I don't know...nothing...no idea) will be a null answer, but you can tell "rumours" about a bot and the next time the bot encounters that "friend" it can say "I heard that you are [whatever the creator said about you)" Or even RAT you out "Roget said that you were [whatever horrible thing Roget said]" Even if it's not possible...it's funny as hell...