Re: [sl4] Is there a model for RSI?

From: Krekoski Ross (
Date: Sun Jun 22 2008 - 22:49:24 MDT

yes, correct-- especially since the number could be conceivably quite large.
But if its just a program that outputs every possible program, its almost a
trivial distinction. -- the source code itself would be a negligible
contribution to complexity with respect to the argument. Theres just no real
way of knowing in advance if we want a specific target program that would
result, and if we did have a target, we could code it with the same
complexity as is required with the argument itself. -- less actually since
we dont need the overhead of the original program itself.


On Mon, Jun 23, 2008 at 12:31 PM, Peter de Blanc <>

> Matt Mahoney wrote:
>> Nope, the machine can't read its own source code.
> Doesn't need to. Here's my machine, based on one of the Quine programs in
> the Jargon File. The 0 at the end is the machine's model number.
> ((lambda (x y)
> (list x (list (quote quote) x) (+ y 1)))
> (quote
> (lambda (x y)
> (list x (list (quote quote) x) (+ y 1)))) 0)
> J. Andrew Rogers wrote:
> > Eh, you might want to double-check your reasoning; it could be picked
> apart
> Then do so.
> Krekoski Ross wrote:
> > you can describe the complexity of each descendant with the original code
> plus an argument stating the number of iterations the program runs for.
> There is no change in complexity.
> Your first sentence is true, but you're incorrect about there being no
> change in complexity. The extra complexity is in the number.

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:03 MDT