Re: Representationalism rescues reinforcement learning
- From: "J.A. Legris" <jalegris@xxxxxxxxxxxx>
- Date: 28 May 2007 18:56:47 -0700
On May 28, 8:16 pm, Neil W Rickert <rickert...@xxxxxxxxxx> wrote:
"J.A. Legris" <jaleg...@xxxxxxxxxxxx> writes:
So here's the thing: in the real world, an organism that learns by
trial, error and (yawn) reinforcement, is likely to get eaten before
it is lucky enough to stumble on the appropriate response. But, it if
has a virtual environment in its head where it can test various
responses before committing to any, it has a leg up on the challenges
of existence, which appears to be just what we mammals have managed to
evolve - internal representations of the real world with little
homunculi going at it, and just possibly, another level or two of
homunculi contained therein (not so much for the good of the theory,
but just to irritate antirepresentationalists a little bit more). And
get this - it's TESTABLE!
That might be a bit of a strawman. It depends on what is learned,
and on what is reinforced. The argument works against simplistic
versions of reinforcement learning.
In any case, you evade the issue of how these alleged internal models
are built. If, for example, they are built using reinforcement
methods, then you are only arguing for a more sophisticated version
of reinforcement learning.
I have no problem with reinforcement learning as a general phenomenon
- I am trying to rationalize a role for internal representations that
could provide a selective advantage in a dangerous environment. In a
harmless environment, a simple reinforcement algorithm will eventually
optimize a response after a sufficient number of trials. But in a
natural environment, this may be too risky, so the "organism" reduces
the risk of stumbling into an unrecoverable state by pretesting its
responses in the safety of a simulated environment. The problem of
obtaining a useful simulation is another issue, which, as you suggest,
might be just another reinforcement learning task.
--
Joe
.
- Follow-Ups:
- Re: Representationalism rescues reinforcement learning
- From: Neil W Rickert
- Re: Representationalism rescues reinforcement learning
- References:
- Representationalism rescues reinforcement learning
- From: J.A. Legris
- Re: Representationalism rescues reinforcement learning
- From: Neil W Rickert
- Representationalism rescues reinforcement learning
- Prev by Date: Re: Beyond Freedom and Dignity: A Philosophical Review
- Next by Date: Re: A request for information please.
- Previous by thread: Re: Representationalism rescues reinforcement learning
- Next by thread: Re: Representationalism rescues reinforcement learning
- Index(es):
Relevant Pages
|
