home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cis.ohio-state.edu!rutgers!igor.rutgers.edu!planchet.rutgers.edu!nanotech
- From: srs@oasis.icl.co.uk (Steve Strong)
- Newsgroups: sci.nanotech
- Subject: Re: Evolution and nanotech
- Message-ID: <Aug.29.01.54.01.1992.13244@planchet.rutgers.edu>
- Date: 29 Aug 92 05:54:02 GMT
- Sender: nanotech@planchet.rutgers.edu
- Organization: ICL, Bracknell, Berkshire, RG12 4SN, UK
- Lines: 59
- Approved: nanotech@aramis.rutgers.edu
-
-
-
- Ian (ian@inf.ethz.ch) wrote:
- :
- : I contest that ! Certainly the production of function is a chancy
- : business but the loss of function . . . imagine the mutation
- :
- : ADD H010102,H100023 -> RTS
- :
- : Where RTS is Return from Subroutine (you knew that). This instantly
- : removes the entire second half of the routine, for example, removing
- : half the proof-reading from the program.
-
- I feel that it would be quite improbable that a random error would
- cause one instruction to mutate into another valid instruction. If
- it does turn out to be probable, then by simply making the distance
- between two instructions greater you can reduce the chance of random
- mutations to arbitarily low figures.
- For example, in Machine Code MKI, we have the following four instructions:
-
- Instruction Binary Code
-
- ADD 00
- MULT 01
- STORE 10
- GET 11
-
- With this, any mutation would generate another valid code, which is
- not very good. Now consider Machine Code MKII:
-
- ADD 00110011
- MULT 11001100
- STORE 11000011
- GET 00111100
-
- With this code, at least 4 bits have to be corrupted *in the right way*
- for another valid instruction to be produced.
- This is really getting quite unlikely. The whole thing is a form of
- error detection, and this really is a very basic version.
-
- : >the error correcting mechanisms in living things are fairly poor.
- : >Simply using a good mathematically based error correction code in the
- : >nanomachine's internal program could reduce the "mutation" rate so low
- : >that a mutation event is expected only once every few billion years or
- :
- : I'd have to see some figures before I accept this.
-
- Using methods as described above, or better ones, you can set the
- mutation rate to an arbitarily low figure. How about one mutation
- per 10**15 years? What about one per 10**20? You can just keep on
- going. Obviously you trade off a bit of storage space for things
- like this. TANSTAAFL.
-
- Steve
- --
- ------------------------------------------------------------------------------
- Steve Strong - email srs@oasis.icl.co.uk | Wasn't me. I didn't say that.
- voice 44 344 424842 extn 2279 | Definitely got the wrong guy.
- ------------------------------------------------------------------------------
-