[ExI] sciam blog article
rafal.smigrodzki at gmail.com
Sat Mar 26 00:06:56 UTC 2016
On Thu, Mar 24, 2016 at 1:03 AM, Robin D Hanson <rhanson at gmu.edu> wrote:
> You can code an awful lot of complexity into even 100MB of code, and if
> that is non-modular spaghetti object code instead of modular documented
> source code, it could take an awful long time to figure out.
### Some parts of the brain, such as the midbrain and structures inferior
to it, are non-modular, spaghetti-like and hardwired in details -
genetically determined and running on completely different principles from
the cortex. The cortex and parts of the basal ganglia are however highly
modular and most likely running a relatively uniform underlying algorithm
that determines both short-term function and the longer-term processes,
such as rewiring of the cortex.
The amount of genetic information needed to shape the hardwired parts is
most likely larger than the information that shapes the cortex, despite the
latter's much larger size and immensely larger ability to learn. I think we
can safely assume that most of any normal individual's cortex' complexity
comes from the process of learning, rather than being genetically coded.
I would expect that a compact description of AlphaGo's underlying algorithm
is many times shorter than the description of that AI's final shape after
playing millions of Go games.
Thus we may have a close analogy between the learning brain and the most
recent crop of AI designs. If so, then there is no magically complex master
algorithm in the brain but rather a simpler genetically coded process
underlying a huge amount of learned information in any individual mind.
Contrary to what Horgan is trying to tell us, the biggest obstacle to
uploading so far is not lack of knowledge about the master algorithm, but
rather the lack of methods for collecting detailed synapse-by-synapse
information on a whole-brain scale. Once we can scan the brain and record
the physical features that determine individual synapse activity, the rest
should be relatively smooth sailing. Of course, finding ways to abstract
the informational structure of the mind from the physical structure of the
brain will be non-trivial but I don't see why it shouldn't be possible.
Horgan's article is silly. It's a grab-bag of stories aiming to shore up a
fervent belief, not a scientific analysis of a problem.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the extropy-chat