On Thursday, May 27, 2021, at 11:05 AM, James Bowery wrote:
> Something else that has occurred to me about lossless vs lossy aka AIT vs SDT:
>
On Thursday, May 27, 2021, at 11:05 AM, James Bowery wrote:
> The bit string to be losslessly compressed by AIT has to come from somewhere
> and that "somewhere" has to be, in some sense, "embodied" in its environment
> so as to receive data from its environment. The structure of its input will
> not necessarily take in _all_ data available in its environment. In fact, in
> all but trivially meaningless cases, it takes in a very small fraction of the
> data available in its environment. This implies a kind of "genetic
> lossiness" or biological SDT that is enormously lossy based on the
> evolutionary utility of its embodiment. This seems to me to stop at one
> level of regress; it's not "AIXIs all the way down". The level at which it
> stops is reproduction as utility function for evolution.
This is interesting point. Sometimes I take in all that are gaven, sometimes I
eat just a taste and leave. The goal of a Lossless Compression or Perplexity
evaluation is to just test the compression of all the data, we would waste time
if had AI only use 3% of enwik8, why do that, we could just make the file 3% as
big then. In real life things do often get lost and not fully eaten or cloned,
but there is other clones, other men, other employees, other Earths, it's ok if
half us die on Earth... ruuuuuuuun!... So in the end things do get fully taken
in the ant eaters nozel pipe, the things that are left out the originals are
sometimes not taken in in other clones, because they are not the real patterns
in physics, so hands and feet and eyes and hair sticks in the evolution, the
deformed arm genes don't get sucked up in the end, some it was trash. Now, for
Lossless Compression, as said the goal of our test should use 100% the data and
request 100% all decompressed. I'm sure HP president will accept if 99% of
enwik8 (100MB) was compressed to 1MB and the rest missed. It would cost the guy
work to make it eat the other 1%! Especially if you made your code so big that
it spits out the last 1%, cuz the coder was so mean and nasty and refuted the
rules. Now you have to accept this. ? I'm unsure but real life is a bit of a
repair rejuvination thing, things revive elsewhere, our LC test may ask to use
100MB but I'm currently using enwik6 half the time. And that can make it work
better on enwik6 and not enwik8. Sure they will battle to get it work best on
enwik8, but today's AI would be half eaten and half as desired, see? tomorrows
may be closer to having all the lost parts. Our GOAL is not to lossy lose
stuff, we want the full and perfect, in real life or in older civilizations I
mean it is harder to keep more or take all desired, it's chaotic at first (big
bang), cold later, our LC test may miss some things it could have but we still
get a better algorithm the next day, and a better one next next day, things
from clones link together and less is lost, it forms back in full perfect from
other "Earths", by the end of a evolution.
------------------------------------------
Artificial General Intelligence List: AGI
Permalink:
https://agi.topicbox.com/groups/agi/T95f11a183fb9b6e1-M4e2cf0a4fecd85b9a6a88a41
Delivery options: https://agi.topicbox.com/groups/agi/subscription
