On 12/12/2018 12:25, Martin Gregorie wrote:
> nobody understands how a neural net actually
> recognises the situation/image/sound/whatever and worse still, there is
> not the slightest possibility that a neural net will ever be able to
> explain why it made a decision.

There is not the slightest possibility that a scientific theory will be
able to explain why *it* works either. Except in terms of eiother a more
fundamental scientific theory or ultimately a metaphsyical assumption.


Why is gravity?


There is no ansewer to that. Gravity as a model of behaviour works. But
we dont know WHY it works, excpet that if it didnt work in the way that
it does the world wouldn't BE te world of our experience.


Our models and gravity is a human imagined MODEL - not a 'thing in
itself out there' -;  are *consistent with experience*. In the end that
is as good as it gets.

Neural net is an attempt to model inductive reaosning - that is find
p[atterns in [phenonema, and thsoe [patrterns may become theories.

In that sense I would expect neural nets to be able to come up with some
kind of 'well this is the pattern match that works' kind of explanation
if interrogated correctly.

So I dont think neureal nets are any more or less inexplicable than
human logoc and perception


--
Canada is all right really, though not for the whole weekend.

"Saki"

--- SoupGate-Win32 v1.05