Using QTI to generate algorithms

From: David Barrett-Lennard <>
Date: Wed, 20 Apr 2005 07:56:42 +0800

Consider the following experiment...

On a computer we generate a million numbers at random, and we write a
program that tests whether the numbers are stored in ascending order, and if
not causes the experimenter to be killed.

A device that measures the polarisation of photons from a light source is
used to generate a random bit string, stored on the computer to be executed
as a program to attempt an in-place sort of the million numbers. After the
program terminates (by executing a HALT instruction) the test procedure is
run to validate the program.

Assuming QTI the experimenter will always find that a "sorting algorithm"
was generated. Often this will work on *any* list of numbers - not just the
one that it was "trained" on.

An interesting question is whether there will be a tendency to discover
short and concise algorithms. In any case, the experiment could be repeated
many times to allow the experimenter to seek short programs - by giving it
less space to work in.

Some really difficult problems could be solved with this technique - eg
"Find a proof to ...", and the test code simply validates the line of

Evidently QTI gives the experimenter the equivalent computing power of a
Quantum computer. ie exponential order problems can be solved in linear

Perhaps the US should build such a machine, making use of its nuclear
arsenal so the whole world can share in the benefits of such enormous
computing power [tongue in cheek].

- David

Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (
Version: 6.0.797 / Virus Database: 541 - Release Date: 11/15/2004
Received on Tue Apr 19 2005 - 19:59:11 PDT

This archive was generated by hypermail 2.3.0 : Fri Feb 16 2018 - 13:20:10 PST