Hi,

On Mon, Dec 15, 2003 at 10:15:29PM -0500, cmoel888@... wrote:
> Hi, Juha.
>
> The model is interesting but like all software and state-machine
> models it ultimately fails in anything but static analysis
> because it lacks the ability to follow a complex operation in real
> continuous time.

First of all, I don't think there is any proof that our 'time' (a sequence
of moments) is 'continuous', even though time may be continuous in
some physical models for example. But this discussion perhaps does
not belong to this group.

Secondly, I don't think brain uses continuous time somehow to
solve the frame problem.

>
> None of these models or systems solves the frame problem,
> e.g., Yale Shooting, because they all rely exclusively upon
> frames.
>
> Frames (groups of states) do not communicate with each
> other due to the fact that each frame is static. Frames can
> be "read" but they, themselves, do not observe or
> communicate and therefore can't be aware of, nor draw valid
> inferences about developments that unfold in a dynamic
> process. Each operation performed upon frames results
> in yet another static frame.
>
> There is no general solution to the analysis of continuous
> processes via frames.

The human nervous system does not observe the environment in a
continuous manner at all. Even the eyes do automatic saccadic
movements and during the movement the receptors of the eye don't
launch at all. If I watch an egg falling, my nervous system does not
sample every position of the egg during the falling. And when I do
some cognitive task, my brain does not iteratate through all possible
logical states to find out what to do next. My brain finds the right
thing to do because my brain is thus wired. The wiring is largely due to
my earlier experiences. To simplify, if the brain gets to a certain
state, the pre-established connections ensure that the next state (or
a sequence of states) is often good and intelligent one from the human
viewpoint.

Animals and most humans don't have a frame problem. Some philosophers
and AI researchers may have.

Juha

>
> Charles Moeller