Re: The wirehead problem

Don Geddis <don@xxxxxxxxxx> wrote:
Tim Tyler <seemysig@xxxxxxxxxxxxxx> wrote on Sun, 05 Oct 2008:
Don Geddis wrote:
Tim Tyler <seemysig@xxxxxxxxxxxxxx> wrote on Sun, 05 Oct 2008:
Of course organisms cannot necessarily maximise their fitnesses -
rather they attempt to maximise their *expected* fitness, just like
other expected utility maximisers.

But that's "expected" in kind of an odd way. "Expected" by the very
long search process of evolution, in a particular environment.

It seems like the ordinary sense of the word "expectations" to me.
Expectations are beliefs about future events. Some of the expectations
are unconscious - but I think that's permissible with the conventional
usage.

I still think it's a bit odd, in a creature like humans, that has both
conscious beliefs, and also these subconscious instinct "beliefs".
Especially when the combination might be contradictory.

Let's at least leave it like this: when trying to explain the behavior of
humans in the modern environment -- e.g. altruism towards friends, or
infidelity in marriage -- it's tempting to say "you're just acting to
maximize your fitness, and these actions help you do that."

This "explanation" is seductive, but in truth much too simple. Your
instincts are designed to maximize fitness in a different environment,
not this one. Your instincts have many "mistaken beliefs" about the
current environment. But it doesn't do any good to ask the guy: "what
are your beliefs? what are your expectations about this action?", because
his conscious response is only half the picture. And yet clearly that
conscious deliberation drives a lot of the person's actions.

You eat a cookie. Why? Because, in the ancestral environment, food with
similar characteristics (although not the same; _nothing_ as yummy as
cookies _ever_ existed during evolution) kept you alive during
starvation. If I ask you, "why are you eating the cookie?", you'll say
"because it's yummy, and eating yummy food makes me happy."

I don't think there's a clear explanation that says, "you are eating the
cookie because you are maximizing fitness." Or even, "you are maximizing
fitness according to your beliefs." Because, clearly, none of your
conscious beliefs expect the cookie-eating to maximize your fitness.

So the only explanation is, "your instincts expect cookie-eating to
maximize your fitness in the ancestral environment, even though they
never encountered cookies during that development time. It's more, a new
thing in a new environment gets classified into a bucket along with other
things that used to, long ago, keep you alive."

Cookie-eating is just another perfect wire-heading example. It's using our
intelligence, to do what we are built to do - maximize reward, with no
regard to what evolution tried to make us do - eat only stuff that helped
us survive.

I find this conclusion stated clearly by the phrase, "organisms are
adaptation-executers, not fitness-maximizers." Despite the fact that I
don't really disagree with your criticism of the phrase.

Yes, because we are built by a reinforcement learning process (evolution),
we execute what we were built to execute. The long term process of
evolution tuned us to be maximize the fitness function of survival, but we
were "tuned" by making us execute the functions that were shown to best
help us survive over millions of years.

And so finally, your comment that started all this was something like: "I
don't want to become a wirehead, because I'm trying to maximize my number
of grandchildren." I don't think that explanation holds up. Human
evolution never had the chance to consider becoming a wirehead. So
there's no guarantee that your design protects you from the dangers of
this new environmental opportunity.

We are safe from the wirehead problem only for as long as we can't find
ways to wirehead ourselves - or as long as we are conditioned to believe it
would be "bad" (as Tim is - and 99.9% of everyone else as well).

Is there an example of a human being modified for pure pleasure like the
rats yet? Or is is it something that has so far never been tried?

Is there any evidence that any drugs would create a state of pure pleasure?

For example, if we knew how to do this, why not do as a standard procedure
for anyone that wants it and can afford it - especially for people that are
about to die from some nasty painful death like cancer? Give them the
surgery to wire-head themselves and let them die in a state of pure
pleasure. We do this to some extent with drugs like morphine already
(under the table) but I suspect morphine does more to stop the pain than to
actually create a real wire-head pleasure effect.

Of course, if you start doing it for cancer patients, then it's going to
creep back into society for anyone that just doesn't want to live anymore
and then the next thing you know, more and more people see it for what it
is - a free (or lost cost) ticket to heaven without having to wait for a
painful death to get there and next thing you know, most the human
population is gone from the earth and we heave solved the over population
problem and global warming!

And the only people who are left are very happy being alive so you have
greatly increased the average happiness of the human population! :)

--
Curt Welch http://CurtWelch.Com/
curt@xxxxxxxx http://NewsReader.Com/
.

Relevant Pages

  • Re: The wirehead problem
    ... rather they attempt to maximise their *expected* fitness, ... in a particular environment. ... It seems like the ordinary sense of the word "expectations" to me. ... Is is valid to say the robot was "maximizing it's fitness function" buy ...
    (comp.ai.philosophy)
  • Re: Sergey Gavrilets and the adaptive landscape
    ... >>> problem to continuing evolution but its width. ... >> page assumes a static fitness topology. ... That in a fluctuating environment organisms respond by randomizing ... of measures are intimately related in the context of biological evolution. ...
    (sci.bio.evolution)
  • Re: pre-tuning to baldwin effect
    ... mean that the fitness level of an organism is hardly determined by the ... You seem to want 'fitness' to mean 'ability to evolve progressively'. ... the ability to cope with different types of environment ... the environment doesn't change and there isn't much evolution, ...
    (sci.bio.evolution)
  • Re: Absolute or just relative fitness?
    ... >> would argue that there is a general overall increase in absolute fitness. ... > is, in some sense, a fitness that would result if the environment ... > of evolution, such a concept is purely theoretical. ... The argument that fitness is increasing suggests that organisms are ...
    (sci.bio.evolution)
  • Re: The wirehead problem
    ... they attempt to maximise their *expected* fitness, ... search process of evolution, in a particular environment. ... It seems like the ordinary sense of the word "expectations" to me. ... You eat a cookie. ...
    (comp.ai.philosophy)
Loading