From: Matthew Gream (matthew.gream@pobox.com)
Date: Thu Jan 13 2000 - 14:41:51 MST
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Ken Clements wrote:
> Sasha, this reminds me of code cracking arguments I have used in the past.
> People have remarked that some code has 6 orders of magnitude
> strength over
> any computation we are expected to have in 20 years, and I have
> argued that
> we would soon be able to transcribe the message to DNA, double it
> 20 times,
> wash it through a few jars of enzymes (or feed it to some GM
> bacteria), and
> reduce the cracking problem by 6 orders of magnitude.
More generally people should be aware of different types of computability
systems, and the appropriate fit of problems to systems.
DNA based computing devices are excellent for graph based problems (Adleman
demonstrated the travelling salesperson problem), as the problem class fits
naturally to the characteristics of the environment (i.e. the reaction
processes). Try a graph based minimisation problem on a classic
(linear/harvard/algorithmic) computer, and it'll take several orders of
magnitude more than it would take in a DNA based computer. Conversely, try
undertaking a numerical calculation in a DNA solution!
You can extend this to quantum computing environments, and parallel systems,
and so on. Even analog systems have their place in feedback control
applications, as they perform faster and more accurately than digital
systems. It follows a general principle of choosing the right vehicle for
the right path.
It is a most interesting development to see these platforms emerge. There
are many computing applications that can find time/space benefits by use of
computing environments more suitably matched to the application class (e.g.
optimisation processes in design and simulation systems). I can forsee,
somewhere in the future, buying a VME type form factor DNA computing card,
and beyond that, even smaller DNA "daughterboards" or chips. In the short
term, there's a _lot_ of work to do to reach those targets.
Perhaps someone more versed in computability metrics, universal (turing),
theoretical (graph, numerical, parallel) and realisable (dna, biological,
silicon, etc) machines could shed more light into this area.
Matthew.
- --
matthew.gream@pobox.com
Cambridge_UK/2000
-----BEGIN PGP SIGNATURE-----
Version: PGP 6.5.1i for non-commercial use <http://www.pgpi.com/>
iQA/AwUBOH5GZq1VD3MTsHhAEQKakgCbBDaVjIrqU0V3kquakBj4MNCkO24AoMec
03rGJhi/qA7UluALcLEsfRDc
=4ROx
-----END PGP SIGNATURE-----
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:26:14 MST