On Thu, 16 Feb 1995, Patrick O'Neil wrote:
1> On Thu, 16 Feb 1995, Eugen Leitl wrote:
1> > I am an Evolutionist and use GAs on my computer. Yet if I am
1> > an individual suffering from fatal intreatable cancer, arisen
1> > from essentially the same mutation/crossover process which is
1> > the driving force of evolution I don't care a fig for adaptation.
1> > It is entirely clear that a population of immortal individuals
1> > does not evolve.
1> > Unless it wishes to. Do you care for Darwinian or Lamarckian
1> > evolution? Or a pleasant mix of both of them? You're free
1> > to choose. That is the nice point in the Downloading concept:
1> > freedom to do anything you did before (including the death
1> > _option_), plus More. How much more, is only limited by the
1> > amount of computation at you disposal and your imagination.
1> > It's Man-made Heaven, all right.
1> Unless you actually LIKE the feel and touch of a woman's skin, or all the
1> wide scents of nature, the wide flavers of organic matter, and don't mind
Theoretically, I can simulate any sensory input I wish. Especially the
olfactory sense is easy to do since the bit rate is so low. Only vision will be
really hard because of high information flow. I can reduce bandwidth
by 1:120 (4 Mio fibres, a la 50 bps/fibre) if I use a renderer which
directly produces retina-preprocessed output.
1> relying on factories receiving a constant good supply of silicon, arsenic
Semiconductor integration density is too low, power dissipation
too high. Only molecular circuitry will suffice, which would be
ideally biochemically produced. I will need GaAs only for semiconductor
lasers to pump my CAME circuitry and only small amounts of GaAs/Si
with a bit of optic cabling to provide the necessary communication
bandwidth. The whole box would be about the size of a fridge. Water-
1> and the like, and the random EM or entropy induced memory wipe or system
Bit mutation is not a problem since I'm a NN. They are highly error-tolerant.
Moreover, I'll do redundant representation and incremental remote backups
(optical tape, not cheap yet sufficient storage density).
Even if you bomb the site, a remote copy will be activated.
1> crash...and reliable physical humans around to make sure things keep
There can't be any crash since the hardware is very special and has only
a "Virtual Machine"/Soft Circuit layer: no software, it's all
architecture, running a special data structure (me).
And since it is a packet switched hypergrid, it's pretty error proof.
And one can be pretty sure that I won't rely on humans: maintenance
can be done by myself or even some dumb machinery. Going into deep
space (Moon/asteroid belt) would be a very sensible idea to do
as the necessary technology (automata which can reproduce themselfes
in hard vacuum relying on common ore/materials found in the belt,
with a touch of comet ice/organics maybe) becomes available (of
course one can do the stuff all by oneself, time being no problem ;).
It's all SciFi, of course. In 50-100 years it could change, though.
1> running correctly.
1> I'll take physical, biological sex and biological interaction with the
1> universe around me any day over plastic and silicon. In any case, the
You can't _feel_ the hardware you are running on. The only things
you will notice are input artifacts, arising from imperfect
technology. As the technology improves, reality will start increasingly
1> mind is not an occupant of your brain that can be uploaded or
1> downloaded. Who and what you (we) are is irreversibly coupled to our
This is a hypothesis. Recent advances in neuroscience seem to indicate
this is not true. Since my position is also based on a hypothesis (but backed
up by hard data) the only proof would be a simulated human. Or will you
settle for an ant/Drosophila m.? A mouse, may be?
1> biochemistry and evolution. In addition, our brains are very unlike
1> digital computers and they cannot be classified as analog either. Neural
On lowest level they use digital pulses, but frequency-modulated (FM is
an analog coding) which are converted into chemical signals
(neurotransmitter), which are translated into electrochemical signals
Essentially, the brain is an analog machine.
1> network computation is inspired by certain aspects of brain structure but
1> it hardly even approaches the real conformation and function of a living
I admit that current NN models are very primitive. You
can't expect to be able to model 10-20 billion neurons
(some 50 types of them) overnight.
But since we live in the 90's, the decade of the brain,
we learn fast.