Hardware for Uploading
Much debate has been generated recently over what sort of hardware
will be required to support an uploaded person. Will any
(Turing-machine equivalent) computer do, or are specialized devices
necessary? The issue is more important than it may first appear,
because it places constraints on the conditions under which uploaded
people may operate. For example, Moravec has proposed that uploads
may be able to function in computers constructed on the surface of
neutron stars, and other exotic environments; clearly, if the mind
requires certain properties in its supporting hardware, then many such
scenarios become impossible.
The "strong" uploading position is that the mind is simply an
information processor, equivalent to a Turing machine in principle
(though highly parallel). If this is true, then any computer with
sufficient speed and capacity can implement the mind of an uploaded
person. Of course, the speed and capacity of current hardware is far
below what is required, and typically more exotic components (optical
or molecular) are imagined. Nonetheless, in principle even a personal
computer (or a Turing machine made of tissue paper!) could support an
upload if it is given sufficient storage capacity, and if one is
sufficiently patient.
The "weak" uploader asserts merely that the mind can be supported by
an aritificial device, but makes no assertions about the nature of
that device. It may, for example, require quantum interactions and
nonlocal effects which cannot be implemented in a Turing-equivalent
machine. The device may have to be geometrically similar to a real
brain, or have components which respond to magnetic fields, or
whatever. Robert Ettinger gives a commentary on this issue in which he
argues that such nonsymbolic processes are necessary for generating
our subjective consciousness.
In either case, no one actually supposes that an upload would be
implemented on anything resembling today's computers. The speed and
storage capacity required are simply too vast. Some developments
which may prove relevant include:
- Protein-Based Computers. Recent work with
bacteriorhodopsin, a light-sensitive bacterial protein, has been
applied to optical data storage and computing. Cubes of material can
theoretically store nearly 100 trillion bits per cubic centimeter
(compared to about 100 million bits per cm^2 in two-dimensional
media). Furthermore, this data can be read, written, and acted on in
a highly parallel manner. For an excellent introduction, see the
recent Scientific American article (Birge 1995).
- Nanocomputers. In
Nanosystems, Drexler describes
the molecular equivalent of simple mechanical computers, with switches
implemented by interacting rods and knobs. With conservative
assumptions, he estimates a switch density of roughly 15 trillion switches
per cubic centimeter in a CPU with a 1 GHz clock speed, processing about
a billion instructions per second (1000 MIPS).
- Optical Computers. Certain materials change their
optical properties based on the light passing through them; that is, light
changes the way they affect light. This allows us to build optical
equivalents of transistors, the basic component of computers. In
principle, an optical computer might be smaller and faster than an
electronic one. But the chief advantage is that such a computer can pass
signals through each other without interference, allowing for much higher
component density.
- Quantum Computers.. In late 1993, Seth Lloyd described a general design
for a quantum computer which could, in principle, actually be built. The
computer consists of a cellular automaton-like array composed of quantum
dots, nuclear spins, localized electronic states in a polymer, or any
other multistate quantum system which interacts with its neighbors. The
units are switched from one state to another by pulses of coherent light,
and read in an analogous manner. Lloyd has shown that such a computer
could perform both as a parallel digital computer, and as a quantum
computer in which (for example) bits can be placed in a superposition of
0 and 1 states. Technical difficulties in building such a computer
include finding systems with long-lived localized quantum states, and
delivering accurate pulses of light. Recent developments have been
bringing quantum computers ever closer to reality; see
Science
Magazine articles on Quantum Computing.
hardware.html . . . . . . . .
1/27/97
. . . . . . .
Joe Strout