There are not prior conditions of talking. Posterior testimony and its evaluation tell the tale of how you came to say what you said, if there is any tale to be told at all. The idea that there are first principles of talk, i.e., a philosophy of talk, or (to bow to erudition) a philosophy of language, is an academic abstraction.
In order to have such a philosophy, you have to recast talk into something that it is not. In order to do this you cannot be paying attention to what people say when they are talking. That is why I have said elsewhere that, unlike Will Rogers, a philosopher never meets a man he ever likes, because he does not listen to him talk. (November 17, 2006)
The context of understanding in this example is determined after what is said. Each remark fits(if it fits at all) with a prior remark after you hear it. Not before. And you may have to ask what a current remark has to do with a prior one. Or vice versa. When would you ask that? Again, later. Not prior.
==============
Dash dropped by a couple of posts ago and made an interesting comment about Chomsky providing the math, if you will, for Derrida and Foucault. This is a reasonable view unless you see it the way I do:
I want to say that there is not any mathematical theory to approach the conversation because, in dialogue, the question of what this(what someone says) has to do with that(something else that someone says) is answered after the fact, not before it.
This, of course, is the case, if it is to be answered at all. Sometimes there is no answering. Sometimes conversations just ramble for a time, reflective of a mood, for example. Sometimes catching up on family matters or old times brings a host of disjointed topics. Sometimes work imposes a common order into which other matters flit.
There is simply no model for this. If you try to think of a conversation as a time series with possible numerical values you quickly come a cropper. It’s not that there are too many things that can be said. There are ways to handle infinity with models: continuous functions, for openers.
It’s just that here the possibilities are indefinite. (See post of October 13, 2006)
Misunderstandings alone can kill you. Double meanings. Nonsensical fun. Boredom. Absentmindedness. Belches. Etc. A conversation can shut down anytime.
Math has to be able to model what is going to happen. It has to have a kind of ‘before the fact’ force, based on some research, or it is pointless.
It cannot have this force here because what this has to do with that in a conversation is determined after the fact, if it is determined at all. What shuts a conversation down(or picks it up, for that matter) can be anything.
Think of the sales pitch that suffers immediate death when you unknowingly speak unfavorably about your hoped-for customer’s father.
Think of the resurrected sale pitch that had almost gone sour, reinvigorated by the mention of a woman’s mother in her teens, as the head cheerleader, during the first time the team had reached the playoffs in 20 years.
“I remember your mother.”
“You do?”
“She was one awesome Tex-Annie.”
“Yes, she was.”
“Finest of the fine. Let me show you…”
This is why artificial intelligence will always be artificial.
