[extropy-chat] limits of computer feeling

[Rafal Smigrodzki]

On 3/14/07, Anders Sandberg <asa at nada.kth.se> wrote:
> Rafal Smigrodzki wrote:
> > But, with self-modification there will be much more profound
> > differences in productivity. The mind that builds itself to do nothing
> > but to maximize access to resources, and to multiply as quickly as
> > possible, ideally adapted to the year year 2030, will not have a
> > "slight" edge in economic fitness over those still adapted to the year
> > 200 000 BC - it will be a gaping chasm.
>
> So you claim. Are there any good evidence for this? I think it would be
> worthwhile to check with economists for example of how much an advantage a
> fully economically rational workaholic would have over an ordinary worker,
> as well as what kinds of enhancements are likely to be how powerful. There
> is a lot of assumptions here we can play around with, and it would be nice
> to have some basis for our intuitions.

### Well, let's consider how much time does the ordinary worker devote
to maximizing access to resources and to procreation: On average, most
humans work less than 5 hours a day - after subtracting the emailing,
daydreaming, schmoozing and other activities that occupy a large part
of the nominal workday. Of the resources gained by the average citizen
of a developed country only a small fraction is devoted to survival
and procreation. You could easily survive on less than 3 000 $ a year
and use the rest for producing offspring, but almost nobody does that
here.

Its fully survival and procreation-oriented competitor could work
perhaps 16 hours a day, make 3 times the money and at least 10 times
the offspring, even assuming no additional advantages aside from work
ethic.

For now, our situation is highly unusual in evolutionary terms - we
have a generation time much slower than the rate of acquisition of
resources, and leftover adaptations that actually *reduce* the number
of offspring as we get richer above a certain point. This is
responsible for the increase in resources per capita in the past 200
years but it will change soon - just wait until the average time to
produce a new generation of offspring changes from 25 years to 25
minutes, or however long it takes to copy a really big hard disk. At
this junction the optimized mind, unless constrained by very powerful
legal or other bonds, will take over.
----------------------------------------

>
> > The speed with which you will be able to adapt yourself to the
> > prevailing conditions will no longer be limited by the snail's pace of
> > evolution - you will be able to adapt as soon as you understand what
> > is needed, work out the mods to your mind, and reboot. In other words,
> > the posthuman minds could be in evolutionary equilibrium all the time.
>
> Assuming working out what is needed takes less time than it takes for the
> economic state to change. That is a pretty tricky assumption.
>
> If we make the analogy of professions, today I can figure out my ideal
> profession, get training for that and have a reasonable likeliehood (say
> 50%) that it has not changed beyond recognition in the meantime. A few
> decades and centuries earlier it was even more likely. So currently we are
> seeing the reverse trend: fewer and fewer people work in exactly the job
> they optimized themselves for. Increasing the ability to adapt and learn
> may help, but that will in itself speed up change in the economy.
>
> My guess is in fact that a posthuman civilisation will always be in
> evolutionary disequilibrium.
>
### I have said "evolutionary disequilibrium" in a bit vague way. Let
me elaborate: there are certain situations when the frequency of
alleles changes only very slowly across generations. Deleterious
alleles are removed at the same speed as they appear, alleles
responsible for various strategies are present in constant ratios, or
sometimes oscillate around an attractor. This is an evolutionary
equilibrium. A disequilibrium occurs usually when the environment
(including other species) changes significantly. In this situation the
frequency of alleles starts rapidly changing, with a velocity
dependent on the selective pressures and the diversity of alleles.
After some time the system arrives at a new equilibrium, with new
alleles at different frequencies. A special case of disequilibrium
occurs when a new ecological niche is reached - when the slow
accumulation of change allows at some point a totally new way of life,
for example when the development of tracheae allowed the proto-insects
to crawl out on land that no creature crawled on before. During the
period of colonization of the niche many adaptations, specific to the
ancestral niche, become maladaptive and their alleles are soon
removed.

I believe that we are in evolutionary disequilibrium caused by
evolution of language, and more recently the complex societies capable
of supporting science. The changes in our environment are profound,
and therefore most of what our genes tell us to do on a daily basis
are actions that are totally maladaptive as measured by the fitness
function. It took only a few decades (invention of the condom, later
the contraceptive pill, ) to sever the connection between the vast
majority of sexual acts and procreation - conventional mammalian
evolution may take centuries to reshape our minds around this
development. But, unless evolution itself is abrogated (by a singleton
AI or other mechanisms), the relevant alleles will be eventually
removed, bringing the minds into equilibrium again.

Now, a disequilibrium as described above can only persist if the rate
of change of the environment is significantly higher than the speed of
removal of outdated alleles. Given self-modification, removal of
maladaptive alleles could be six orders of magnitude higher than in
normal mammalian evolution. To persist in disequilibrium the posthuman
economy would have to experience dramatic jumps in efficiency,
equivalent to the invention of language and the scientific method,
every few generations of posthumans, or once every few hours. I tend
to think that this is highly unlikely.

Rafal