Showing posts with label Mersenne. Show all posts
Showing posts with label Mersenne. Show all posts

Monday, September 11, 2006

44th known Mersenne prime: M32582657

Just like in the case of the 42nd and 43rd known Mersenne prime, The Reference Frame is the only place in the world where you can learn about the newest Mersenne prime before it is officially revealed.

The newly found Mersenne prime by the GIMPS project is
  • 2^{32,582,657} - 1
and it is larger than the previous, 43rd known Mersenne prime. This means that the newest prime above is becoming the largest known prime.




Congratulations to Mike Cobb: but he won't get the $50,000 award yet because the number is shorter than 10 million decimal digits.

Wednesday, April 19, 2006

BBC climate software confuses 200,000 computers

This story is a good example how the climate models work in the most optimistic case.

The idle time of most PCs is wasted. About 30,000 people like me run software such as MPRIME - the search for the greatest prime integers in the world. It is a well-defined activity and there are very good reasons to trust this software.

Actually, there exist other programs above the BOINC platform, and some of them can be found in this list:




However, some people don't like things like LHC at home too much. Instead, they want to save the world and help the humanity. So they download the third program in the list above, namely

You can also join the group of 200,000 enthusiasts, the saviors of the planet, if you click the link above and continue with "Taking part in CPDN". This community will calculate the date of the armageddon. ;-) But wait a minute. The Reference Frame has been saying that the existing climate models are not trustworthy and those who run them often fail to respect basic principles of science.

So can this software be scientifically useful at all?

So far, the answer is, of course, No. The first big fault was announced by Nick Faull of Oxford:

The users were told that the software is defective, heavily overshoots the 20th century warming, its aerosol configuration files are completely corrupt, the simulation completely crashes in 2013, and the crunching must start from the very beginning. Two hundred thousand of disappointed users were shown that they trusted charlatans and amateurs.

Why is it that we can learn about these rather lethal bugs? It's because the BBC model was, unlike many others, put into test. It was tested by 200,000 people who can actually look what's going on, investigate, and if their computer game crashes in 2013, they may suspect that something is wrong about it. The creators of the software simply could not try to hide the failure.

Today we know that the model was giving an unreasonably high warming for the 20th century. But as Nick Faull explained, this is not how they found the problem. They found it because the computers were terminating violently in 2013. It follows that

  • Only after a climate model crashes on 100,000+ computers with an "unrecoverable error", it becomes acceptable for consensus science to suggest that there could perhaps be something imperfect about the climate model.

We should expect that the BBC team is going to design a new climate program that will not crash, and the people will again download it; the update may occur automatically for the current users. The program will not crash but its scientific value will be more or less identical to the first one: zero. These programs generate bogus numbers that have never been tested, and if these bogus numbers agree with the scientists' environmentalist preconceptions, the scientists improve them and publish them. Climate change science is a junk science.

Now consider the fact that BBC is a relatively well-known institution and the software was designed for hundreds of thousands of people. Under normal circumstances, similar climate models are run by a few rather unknown people who underestimate the latent heat of ice by three orders of magnitude and who can't calculate the probability that one of 5 equally likely outcomes from a list of 32 alternatives occurs.

Just by counting the people who test the software, it is easy to guess that the probability that all critical problems with a climate model and the underlying science are found and corrected is smaller by several orders of magnitude if these models are not checked by hundreds of thousands of people. If the authors know that no one is watching what they're doing, the chances are that no one can ever figure out anything. This unchallenged approach, an approach based on "consensus" and mutual respect between politically synchronized people, is simply not science. It is a religion and a very expensive one.

It is not a religion that builds cathedrals that will be admired by people in 2500. It is a religion that slows down the economies.

Even with thousands of people who have been getting "Unrecoverable error" messages for some time, essentially since the project started in February, it took weeks before the error was admitted by BBC and three days before you could learn about it on leading anti-junk-science blogs, including Climate Audit.

Update

The announcement under the "Error discovered" link above has been updated. The original message of Nick Faull, the project coordinator, was deleted and "superseded" by a comment by Myles Allen, the principal investigator. I guess that it is Allen who is in charge of the money, which is the reason why an honest explanation of Faull was replaced by a not-really-honest statement by Allen. For example, Faull fairly says (see the ClimateAudit article):

  • Unfortunately, all the data returned to us so far has been affected by this problem. ... since such an important component is missing. In order to do the experiment we intended, we unfortunately have no choice but to start models again from the beginning.

On the other hand, Allen says: "Your crunching to date has absolutely not been wasted." He continues by saying that it is a "routine work" to separate the influence of one component of a coupled system: they're doing such things routinely. Yes, indeed, they are routinely committing scientific malpractice because if it were possible to separate the influence of one important component of a coupled system, it would have to be a linear (or another simple enough) system and it would not require 200,000 computers for being evaluated.

Of course that some terms contribute almost linearly but if a piece of computational software is more than just a waste of CPU time, it should be exactly designed in such a way that if a bug is found, the results become worthless. Unfortunately, the computers, CPU time, and salaries of many consensual scientists are not tools to improve science but rather tools to increase a false impression of credibility of not-so-good science.

Sorry but the statement of Mr. Allen is what I call a lack of scientific integrity. If money is at stake, things can be re-interpreted, can't they? If Mr. Allen succeeds in continuing with the corrupt data, then I will personally consider him to be approaching the boundaries of the criminal justice because what he's doing to earn some money is a kind of fraud. Imagine that there is a multiplication bug in the program that looks for Mersenne primes and the creators announce that they can fill in the missing prime integers anyway - they're doing it routinely. ;-)

Summary of the Inquirer

The Inquirer has calculated that the grid has contributed more to global warming than to scientific knowledge.

Tuesday, December 20, 2005

43rd known Mersenne prime: M30402457

One of the GIMPS computers that try to find the largest prime integers of the form
  • 2^p - 1
i.e. the Mersenne primes has announced a new prime which will be the 43rd known Mersenne prime. The discovery submitted on 12/16 comes 10 months after the previous Mersenne prime. It seems that the lucky winner is a member of one of the large teams. Most likely, the number still has less than 10 million digits - assuming that 9,152,052 is less than 10 million - and the winner therefore won't win one half of the $100,000 award.

The Reference Frame is the only blog in the world that also informs you that the winner is Curtis Cooper and his new greatest exponent is p = 30,402,457. (Steven Boone became a co-discoverer; note added on Saturday.) You can try to search for this number on the whole internet and you won't find anything; nevertheless, on Saturday, it will be announced as the official new greatest prime integer after the verification process is finished around 1 am Eastern time. If you believe in your humble correspondent's miraculous intuition, you may want to make bets against your friends. ;-)




Actually I am so incredibly sure that you should bet thousands of dollars if someone is ready (and has the courage) to argue that I am mistaken and the exponent won't be 30,402,457. Trust me. Note that we predicted the previous Mersenne prime correctly, too. The new greatest prime looks like follows (more than 9 million digits are omitted):
  • 315416475 … 652943871

The exponent may count the number of curves of a given genus in a particular elliptically fibered Calabi-Yau manifold. Or something else.

Thursday, February 24, 2005

New record prime

The Mersenne numbers are the numbers of the form "2^p - 1". Several centuries ago, some people used to think that these numbers were always primes if "p" itself is a prime. While it's true for the first few values, the general claim is, of course, false. Primality of "p" is a necessary condition for primality of "2^p - 1", but not a sufficient one. (It's necessary because if "p=qr", then "2^q - 1" and "2^r - 1" are divisors of "2^p - 1".)

In fact, only 41 Mersenne primes are officially known. The highest one, "2^24,036,583 - 1", is also the greatest officially known prime integer in the world. It was found in May 2004. Usually, the exponent roughly doubles if you want to find a new Mersenne prime. However, the Mersenne prime numbers look surprisingly dense for the exponents between 20 million and 26 million.

An international network of computers GIMPS whose website is at

is looking for new Mersenne primes. You may join. One can see that the main page of this server announces that a new record prime may have been found. It would be the 42nd known Mersenne prime number. I predict that the exponent "p" will be

  • "p = 25,964,951"

Note that if you find the first Mersenne prime with at least 10 million digits, you will win at least 1/2 of the $100k award from the EFF foundation. It takes roughly one month for a 3GHz computer to test (using the Lucas-Lehmer test) a number like "2^34,362,227 - 1" which is how a candidate for a 10-million-digit prime number looks like.

Note added later: Of course that the word "predict" was partially a joke. I knew that the exponent would be what I wrote. Congratulations to Dr. Martin Nowak from Germany (by the way, it's almost a Czech name) to the discovery of the 42nd known Mersenne prime.