gimps schmimps

Jones, Spike (
Thu, 12 Aug 1999 12:57:34 -0700

	GIMPSers, after reading a few of the replies on the list
	and privately mailed to me, I must comment to clear up
	a misconception.

	***DO NOT quit your day jobs*** if you are joining GIMPS
	to try for the 100K EFF prize.  I mentioned the prize to a
	certain extent as a joke, for one's individual chances of
	getting it are small indeed, as I commented before:

	1.  Currently GIMPS is systematically searching numbers
	of a couple million digits, so these are not even eligible for
	the EFF prize, which is for a 10 million digit prime.  The
	GIMPS wolfpack will not be at the 10 million digit mark for
	about another 4 to 5 years, assuming it does not dissolve
	in chaos, as lone wolves defect hunting for the big game.
	The EFF prize may actually *contribute* to the breakdown
	of the previously stable systematic search for Mersenne primes.

	2.  Current common desktop computers (400-500 Mhz
	P2s and P3s) take about two *years* to analyse a *single*
	candidate, and that candidate has about 1 in 300,000 probability
	of being a winner.

	So, the mathematical expectation (prize value times the
	probability of winning) is only about 30 cents, for 2 years
	of computer time.  Consequently, I do not recommend
	joining this project for the money.

	I *do* recommend joining it for developing parallel distributed
	computing, however.  The GIMPS software is open source
	C code, and can be modified, so that you can use the background
	computing infrastructure and I/O, then modify the math to
	do whatever you want.

	This also provides an interesting study case for Greg Stock's
	E4 talk.  I suspect many of the current participants in GIMPS
	have not calculated their person probability of winning the
	cash prize.  Many of us however, were in it for other reasons
	before the prize was announced.

	Stop CPU waste.  A busy computer is a happy computer.
	Let us create a womb for the singularity, then hope to hell
	a monster is not born therefrom.  {8-]  spike