From: Mitchell Porter (mitch@thehub.com.au)
Date: Tue Mar 25 1997 - 20:03:40 MST
1) Quantum uploads
What are the implications for mind uploading if some form of
Q-mind theory is true? (e.g. the Penrose-Hameroff idea that
mental states are the quantum states of entangled sets of
neural microtubule electrons). As far as I can tell,
uploading, copying, and so forth are all still possible,
but there are all sorts of uniquely quantum-mechanical
considerations.
There is something called the No-Clone Theorem, which says
that there is no device able to make a perfect copy of an
*arbitrary* quantum state. However, there are ways to make
copies of particular states, and there are universal copiers
(only theoretical at this point) which only produce bounded errors.
When I started thinking about this, I guessed that there would
be two ways to copy a quantum state, a "classical" way and a
"quantum" way. The classical way would be to measure the original,
storing that information in some classical medium, and then
reconstructing the quantum state in some new substrate by
consulting the classical blueprint. (The original need not be
destroyed by this process, if "protective" or "non-demolition"
measurements are possible.)
The quantum way would avoid measurement and simply transfer the
information via quantum interactions; an example would be quantum
teleportation.
Conveniently, Seth Lloyd has just put out a preprint
("Quantum controllers for quantum systems, quant-ph/9703042)
which defines exactly this distinction (he calls it
semiclassical control vs quantum control). A quantum controller
can extract information from one quantum system and pass it
to another without disrupting the first.
I suspect that all the discoveries being made in quantum
computation, etc, will be very relevant to uploaders.
The quant-ph archive at http://xxx.lanl.gov is the place
where all the new announcements appear.
2) Fast minds
Long ago John Clark said that the first uploading would be an
evolutionary divide, for reasons which everyone probably appreciates -
an upload is expected by many to think orders of magnitude faster
than a biological human.
I've thought about this ever since, and so far I've thought of
three ways in which a fast mind might come about.
i) Emulation, i.e. uploading. Someone's brain is scanned and
emulated in a faster computational medium.
ii) Acceleration. The human brain is made to function at greater
and greater speeds (for example, through introduction of
new nanostructures into neurons).
iii) Invention. A "de novo" mind is created in a new medium,
rather than simply copied from the biological. This category
includes "invention" through digital evolution (genetic
algorithms, etc).
It seems to me that the third path is the most likely - perhaps
as a matter of market competition, and then natural selection,
amongst Internet agent software. The third path could involve
partial emulation, in the sense that neural architectures
discovered in biological brains might be found useful and
incorporated into software or hardware - but it seems unlikely
to me that all the conditions necessary for human mind uploading
will be met before there are "human-level" agents that have
evolved online.
-mitch
http://www.thehub.com.au/~mitch
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:44:18 MST