Yet Another Fresh Start


Geez, I’ve got to pull myself together. Make a website that doesn’t actively repel just about anyone who I might care to impress. It isn’t about you, it’s about me. I’ve got to do something else than be crazy.

So… back to research, then. I was re-reading Zen and the Art of Insanity — the unpublished and unfinished second edition — when it struck me that while I was learning Lisp I kind of forgot a few things about intensions and extensions. Now the question is this: how do we create a program that has meaningful functionality based on what the theory predicts?

Some notes.

Firstly, how about summing actual and potential quality to each other? So that… mate, for example, would be a good reaction because… patterns of reciprocal quadrants can be summed to each other like the sign of the x coordinate were the same. That would explain why mirror doesn’t increase quality so much.

How about waste, then? It’s perhaps a badly named relation. Waste doesn’t have to be waste, it can just be bringing the discussion to a new level. Still it’s kind of antisymmetric. But… waste and distract? They are all pretty negatively named.  And they are named that way because of something someone told me… that some things weaken each other…

Looks like in mating both people get score (horizontal) but in vertical relations only the initiator of the change in ontology gets score… if… he is of equal or higher rank.

Maybe just sum them as vectors. You get classical quality from vertical relations and romantic quality from horizontal relations.

This is pretty hard. The underlying formula for the relations is still somewhat absent. Just a bunch of fragmented sketches.

***

Another symptom of tiredness: I remember something but forget it right away.

Oh yeah. Each instance of the bot should have a mood that is a Sinn. But, if we have so many moods how do we design the UI? Can’t rely on text anymore. Need graphical UI. But suppose we have one. There are so many possibly relevant modifiers here, such as:

* What the bot actually feel vs. what it shows

* What the bot thinks it shows vs. what it actually shows

We can’t differentiate those two meaningfully here without a lot of other kind of work being done first.

But what was it that I saw there… the light… within intensions and extensions? At least it’s a fact that the AMOQ Lisp program doesn’t currently use that information for anything.

Let’s keep things simple. The bot expresses each said thing with exactly one Sinn. The Sinn may be a nonrelativizable one. How does the Sinn affect the thing that is said?

In order for effects to be meaningful the bots need to have some goals. They need to form these goals randomly and they must be as simple as possible. They can do this by associating themselves to each other as a group and then favoring the interests of the group.

Suppose the bot is an AI of a 1st person shooter game. Which priorities will it have?

Objective 1: gaining objects.

Objective 2: healing friends, hurting enemies

Objective 3: winning, ie. capture-the-flag-style or something else

Objective 4: let’s leave that one.

How about other objectives, such as… subjective objectives?

Subjective 1: knowing the surroundings

Subjective 2: knowing the state of friends and enemies

Subjective 3: knowing about victory

Subjective 4: let’s leave that one for the player?

And…

Normative 1: knowing how to move

Normative 2: knowing how to tell friend from foe (this means the bot will mistake a painting of a friend for a friend?)

Normative 3: knowing strategy

Normative 4: …

So, this leaves us with bots who may be unable to move or who may think a stick is a friend. Hmm… interesting.

Aesthetic 1: recognizing similar events unrelated by causality (this will prompt the bot to NAME the event and will affect its emotions) (the bot will try to find a cause?) (an example: team member strikes nonhostile, and all nonhostiles in area turn hostile. bot does not know why this happens if it has no prior conception of it)

Aesthetic 2: recognizing reactions to coincidences, ie. what sinn is usually used of them

Aesthetic 3: recognizing when there is no previous model of what to do

Aesthetic 4: …

So.

This means the bot has to have coded as data:

what it sees, as bitmap

what are objects (this means no typing between objects and terrain in code?)

what are options (what can it do, ie. open/close, eat, drink)

(should be able to fight in different directions when wielding two weapons, but too complicated?)

There should be a drawing option.

Also:

what are concepts? The bot needs to be able to conceptualize things.

This means the game should revolve around issuing verbal instructions, commands or whatever to the bot. Sinns. So… if the bot is instructed to LIE… it will use a wrong concept for something. If MUST it will keep doing something. If MAD it will behave randomly. If NEVER it will refuse to do something. If DOUBT it will search or expect lies. If ALWAYS the bot will always do something?

If THINK it will find a cause for something. Isn’t that REASON? If CHOICE it will make a choice according to priorities. If WHY? it will report a cause for its actions. LANGUAGE means it is taught or it teaches someone a new word. If TRUTH it will not believe what it’s told.

This is all rather complicated. How do we systematize this?

Or should we just stick to text?

07:03 <J_Kahvi> hmm, tässä kannattaa välillä vähän miettiä mihkä suuntaan tätä hommaa viedään. tää alkaa kehittyä nyt jo enemmän liikkuvan, näkevän ja toimintoja tekevän tekoälyn suuntaan kuin botin joka heuristisesti tunnistaa sävyjä
07:04 <J_Kahvi> tuohon heuristiikkahommeliin on periaatteessa jo olemassa kaikki tarpeellinen ellen nyt aivan ole erehtynyt
07:05 <J_Kahvi> loppujen lopuksi siihen ei yksinkertaisimmillaan tarvita muuta kun markover, kaiken teorian voi sivuuttaa. tekee vain markoverilla tiedoston ja sitten matchaa normaalissa markovin ketjuihin perustuvassa lauseenmuodostuksessa sinnejä toisiinsa
07:05 <J_Kahvi> toki siitä saa tuota monimutkaisemmakin
07:06 <J_Kahvi> esimerkiksi juuri lisäämällä roolit joissa kukin puhuja toimii. silloin botin pitää osata myös tunnistaa kuka sanoo. ja ehkä myös että kenelle.

Tuo “kenelle”. Siinäkin on klassinen and a romantic component. Classical would be who the person is looking at or speaking to. Romantic is whose Sinns does it pertain to. The bot should choose the interpretation that has the greater amount of quality.