Re: new mersenne?

From: Spike Jones (spike66@attglobal.net)
Date: Thu Nov 15 2001 - 22:02:37 MST


> > > > Spike Jones wrote: and a neeeeewwwww Mersenne prime has been
> > > detected.
> >
> > On page 87 of The Spike, Damien Broderick speculates:
> >
> > "...we can estimate that there's nearly an 80 percent chance that
> > a three million digit prime will be discovered by November 2001."
>
> Anders Sandberg wrote: So we can safely say Spike is on schedule? What
> about your estimation
> of when it would be discovered, did it fit in perfectly or does it turn
> the fit?

Good question Anders. I have been working over my models in
order to predict when the next record will be set. I have a
few observations to report, but it is unclear the extropian angle.
So if you do not like math or computers, please read no further.

Other than the fact that the new Mersenne prime detected yesterday
was a nearly 1.5 sigma long gap between primes, the event is not
far from my expectations. In Damien's book I calculated about an
80% chance by 11-01, but in fact I was a little cowardly in my
model of the increase in computer speed. Had I known in December
1999 that we would have covered this much ground in two years,
I would have calculated a 87.3% probability of a Mersenne in that
interval.

A complete report is being prepared for my website. This is
a short preliminary report:

I estimated the number of participants about right, and the
numbers of machines doing factoring and double checking
was predicted fairly accurately. The speed of computers
as a function of time I derived from Ideas Futures. I assumed
that those who traded in those memes knew what they
were doing, and they nailed it almost right on. But where I
really messed up was underestimating the impact of the
Pentium4. The P4 instruction set is twice as fast as a
similarly clocked P3 in doing the FFTs used in the Lucas-
Lehmer algorithm. So I blew that one. But I have a new
version of the model that takes these developments into account.

I am currently tuning my model using new information on the
speed of processors and a really cool Matlab tool Ive been
experimenting with, a homebrew Butterworth filter thingy. But
preliminary estimates indicate that mersenne prime number 40
has about a 50% chance of being discovered by November 02.
If discovered earlier, the word "by" means the start of November.
If discovered later, the word "by" means the end of November. {8^D

I expect to sharpen this prediction in the next few weeks, so
stay tuned. In the mean time, if you want to join GIMPS, there
is an extropians team. For some time we were running neck
and neck with the Meeeeensa team. Go to the link below and
follow the instructions. The password to join is maxmore.

http://www.mersenne.org/prime.htm

spike



This archive was generated by hypermail 2b30 : Sat May 11 2002 - 17:44:19 MDT