# RE: The Geddes conjecture

From: Marc Geddes (marc_geddes@yahoo.co.nz)
Date: Mon Jul 18 2005 - 22:28:41 MDT

--- "J. Andrew Rogers" <andrew@ceruleansystems.com>
wrote:

> Marc Geddes wrote:
> > Prediction: Calculations of determinstic outcomes
> > based on fixed initial conditions
> >
> > Projection: Calculations of a *range* of possible
> > outcomes based on changing conditions.
>
>
> What is the bloody difference here? None worth
> arguing about that I can
> see.
>
> Let me rephrase for you: Your system produces a
> list of outcomes
> ordered by their computed probability. This is your
> first-order
> prediction. If you pop the first element off the
> list, you can build
> another list, a second-order prediction based on the
> first-order
> prediction. Oops, I mean "projection". Get as
> fancy as you want with
> your probability lists, it is a distinction without
> a difference.
>
> In short, the difference is a few lines of code in
> implementation. And
> that is a difference not worth discussing here as
> far as I can tell.
>
>
>
> j. andrew rogers
>
>

Hmm, it seems my last post completely passed every-one
by again *sigh*

Let me try one last time to explain why I think the
distinction betwen *Projection* and *Prediction* is
important. I will also sketch why I think the
Bayesian probability framework is seriously incomplete
(the two points are related).

O.K, I want every-one on the SL4 list to try to
*project* their own futures - what do you think you'll
be doing in:

(a) 1 hour from now?
(b) 1 day from now?
(c) 1 year from now?
(d) 10 years from now?

To carry through this process, you have to include in
your projections the effects of possible courses of
action that YOU took at earlier times. Projecting
forward your life is *not* like projecting forward
possible outcomes of a chess game for instance. The
reason is that the world can exert feed-back on you.
To project yourself forward in time to possible
outcomes, you have to extrapolate not only the results
of your possible actions on the world, but also THE
FEED-BACK RESULTING FROM THESE ACTIONS. So the
possible futures you're projecting for yourself are
becoming more and more dependent on previous earlier
choices you made as you move the projection forward in
time.

And this is what breaks Bayes. It took me a long
time, but I finally pin-pointed the exact flaw that
smashes the Bayesian framework. The flaw in Bayes is
that it assumes that the objective external things
being reasoned about can be completely isolated from
the internal *process of reasoning itself*.

Is it possible to assign objective probabilities on
the signfigance of external observations without
reference to the internal procedures (cognitive
processes) used to make these observations? Bayes
assumes that it is. I say, it is not. The same
external observation can result in a different
probability based on the *procedure* (cognitive
process) used to make the observations.

Comprehende?

```---
THE BRAIN is wider than the sky,
For, put them side by side,
The one the other will include
With ease, and you beside.
-Emily Dickinson
'The brain is wider than the sky'
http://www.bartleby.com/113/1126.html
---