59 replies to this topic

[dnamechanic]

Hi basho,

Yes, go 2ch!

I am guessing that 2ch represents 2channel.

http://en.wikipedia.org/wiki/2channel

Shii's Song is beautiful, thanks.

[Live Forever]

Looks like Folding@Home finally crossed the petaflop barrier for real, due mostly to PS3 support:
http://fah-web.stanf...y?qtype=osstats

From the F@H blog:
http://folding.typep...ing-the-pe.html

I submitted the story to digg if anyone is interested.

Edited by Live Forever, 18 September 2007 - 05:02 AM.

[maestro949]

Nice! Dugg.

BOINC is at 1/2 a PFLOP.

Imagining 15 years from now...

There should be multiple (perhaps many) petaflops per PC / laptop / gaming console and maybe even in our mobile devices. They'll probably be powered by a battery the size of a stamp that last for a week between charges.

- About 100 million PCs are sold per year and this number could reach 1Bn by 2022. - Gartner says that cell phone production will reach 1Bn by 2009. Sales could easily triple by 2022.
- Gaming consoles sales were about 70m in 2007. This could easily be 1/2Bn or more by 2022
- TVs and several other gadgets will have general processors in them as well. That's another few billion processors with petaflop+ potential.

Networking and harnessing all of these idle Petaflops will be key to advancing the big NP complete problems of science. The challenge will be building the software algorithms and architectures that can harness all these exaflops* of horsepower at our fingertips. It'll take 15 years to design the software for this performance so we need to start now.

*An exaflop is 1000 quadrillion calculations per second, or 1 quintillion calculations per second.

[Live Forever]

Some of the stuff that they have been accomplishing post petaflop:
http://folding.typep...t-petaflop.html

It is mainly due to the software update to the PS3 that was released by Sony which dramatically sped it up.

[maestro949]

One idea we have is to make the calculation even more accurate.

Nice. The new projects and better accuracy are great example of how pushing the boundaries of science and tech can open new doors. These experiments in MD will attack the biological complexity problem from bottom up while we also attack simulating metabolic function and protein interactions via systems biology from the top down. We're seeing this steady march on all fronts. We have death cornered. There's nowhere to run mr. reaper. Might as well throw in the towel now.

heh, who am i talking to?

[Live Forever]

Nice.  The new projects and better accuracy are great example of how pushing the boundaries of science and tech can open new doors.  These experiments in MD will attack the biological complexity problem from bottom up while we also attack simulating metabolic function and protein interactions via systems biology from the top down.  We're seeing this steady march on all fronts.  We have death cornered.  There's nowhere to run mr. reaper.  Might as well throw in the towel now. 

heh, who am i talking to?

To the bastard Death who has plagued mankind for all of recorded history. (and before) That bastard a-hole is on the run. )

[nefastor]

Such a network of linked obsessed visual cortexes working in conjunction with imagination and problem solving skills could solve the most complex of biological problems in months or years where it will take an individual genius decades to work out the mathematical and chemical models.  Foster competition, teamwork and build a reward system and then we effectively have our genetic algoritms which will lead to the necessary evolutionary components as well. 

As for immediate implementation, there should be no reason why psychologists, neurologists, nor AI engineers should not already have access to large scale manipulations of variables on in-game performance (including teamwork, memory, reaction speed) as has already been done in Duke Nukem for research into depression. What other ways could human desires (especially in entertainment) be contributing to distributed cognitive advancements?

Hi guys,

Actually this is already happening, sort of. Many sociologists study the dynamics of the HUMAN population of MMORPG's (massively-multiplayer online role-playing games).

MMORPG's such as EVE Online (which, incidentally, runs on a supercomputer) have a limited population (tens of thousands) that can be extensively and accurately monitored in all aspects of their (socio-economical) life. Things like empire-building, trade, market evolution, territorialism happen in real time over several years and are much easier to work with than the actual real world. Plus, you don't need to pay the people you're studying : they are actually paying to play the games.

You can consider this sort of social study an experiment involving thousands of privately-owned computers AND the willing brains of their owners, in furthering science.

Nefastor

EDIT : PS : the processor of the PS3, known as "Cell" (an 8+1 core design) is actually not going to stay confined to the PS3. A year ago I saw a prototype server blade that featured two Cell's, and typical blade racks carry 14 servers. Picture a 19" shelf with the power of roughly 150 PS3. I don't know the ETA on those blades, though.

[Live Forever]

Explaining why streaming processors are so much faster than regular processors. (and relates it to Moore's Law):
http://folding.typep...ah-works-f.html

I have been checking the blog they have a couple times a week (they usually update it once a day), but I felt I should mention this entry in particular, as I found it very interesting.

[maestro949]

Astrophysicist Replaces Supercomputer with Eight PlayStation 3s

Too bad the PS3 sales are struggling. It would be nice to see a cell based console do well in the marketplace such that it can double as gaming rig and scientific engine. 1 Petaflop is nice but we'll probably need many thousand times this to do more sophisticated simulations. It'll be interesting to see what Sony does for the PS4. The cost of the unit clearly needs to come way down.

Edited by maestro949, 02 November 2007 - 06:55 AM.

[Live Forever]

Astrophysicist Replaces Supercomputer with Eight PlayStation 3s

Too bad the PS3 sales are struggling.  It would be nice to see a cell based console do well in the marketplace such that it can double as gaming rig and scientific engine.  1 Petaflop is nice but we'll probably need many thousand times this to more sophisticated simulations.  It'll be interesting to see what Sony does for the PS4.  The cost of the unit clearly needs to come way down.

Wow, that is a really interesting story. Thanks for linking to it, maestro. I am guessing that the 360 and Wii are different in their design so as to not make them as efficient a system for computing purposes?

Luckily, as computing hardware increases in power, the F@H project will naturally just expand. The orders of magnitude bigger that you predict are needed, maestro, will probably happen within the next few years.

[Live Forever]

They are up to a little more than 1,200 TFlops (1.2 PetaFlops) now: http://fah-web.stanf...y?qtype=osstats

The PS3s just by themselves are up to almost over a PetaFlop.

Also, the amount of active CPUs just keeps going up and up and up:


Spectacular. [thumb]

[maestro949]

I'm looking for another excuse to drop $500 on a PS3 but sadly I can't find one. Blue Ray DVDs? No thanks. Games? The reviews are just... meh.

[Mind]

Guitar Hero! That's the extra reason. Takes air guitar to another level.

[dnamechanic]

Guys at the [H]ardforum discuss topics related to folding machines (including the PS3).

This thread has a bit of info that could be helpful if looking to purchase a PS3.

[A941]

Now if we can take the next step and build a game where people are actively but indirectly contributing human brainpower to the biological research you then have an emergent concept and mechanism that can not only defeat aging but can solve any and all biological challenges.  You have the most powerful supercomputer that can be built.

Interesting idea!
Is such a project possible?
We should ask the people at F@H if this is possible by now!


We should also ask for more computing power in GAMERS FORUMS!
And we should ask F@H to programm clients for the WII and the XBOX!

[maestro949]

Guitar Hero! That's the extra reason. Takes air guitar to another level.

Heh. Jam away while the proteins fold. Hmmm.... not sure that'll get me to fall off the gaming addiction wagon.

Interesting idea!
Is such a project possible?
We should ask the people at F@H if this is possible by now!

I posted a question on the F@H boards but didn't get a response. Not entirely surprised as if it were easy, it would probably already be happening. It's still worth looking into though or simply keeping an eye out for opportunities.

The goal would be to find a sizable problem that is difficult to solve algorithmically but fairly easy to do with visualization and then have people chip away at it. Another might be to simply look for data points that are of value that can only be obtained by visualization and simply catalog them.

One idea for protein folding might be to let the algorithms do the parts they are good at such as folding the amino acids into secondary structures that can be done quickly with a high degree of confidence and then see if people can simply slinky around the remaining strand of amino acids to find a suitable structure that meets the minimum energy requirements, oh, and doesn't violate the laws of physics

Another idea would be in the exploration of protein interaction networks. In the not too distant future these will be all the rage. Though, unlike Beanie Babies, protein interaction networks will remain popular for quite some time as we'll be constantly comparing them in a multitude of ways for disease, drug and even aging research. Exploring and comparing these via algorithms though has an enormous computational cost but exploring them visually can reveal a wealth of data in regards to easy-to-spot patterns.

Edited by maestro949, 02 November 2007 - 06:57 AM.

[dnamechanic]

Along the line of comments posted several times in this forum by maestro949:

From the F@H blog:

October 25th, 2007: "Life In Motion"                                 
Bio-X has teamed up with Stanford's National NIH Center for Physics- based Simulation of Biological Structures to hold a symposium entitled, "Life in  Motion". The goal of this symposium is to educate students and scientists from different disciplines about the exciting uses of simulations driven by the laws of physics and mechanics across a range of scales, from molecules to organisms. The talks will be presented by a series of experts and innovators from around the world.

This represents steps in a direction that seems quite likely to yield progress in solving the big problems that we all face.

Stanford's F@H program represents more than a protein folding effort (though that is itself an ambitious undertaking). The F@H project is demonstrating to the world that home computers are more than something on which to write email and to play games. Of course other organizations are doing similar things, notably SETI (on which F@H was probably modeled). But F@H has taken distributed computing to a new level by teaming up with Sony, with the PS3, and with ATI, now a part of AMD, with their graphics processor. Here's a salute to both Sony and ATI/AMD.

I have noticed a positive trend in the gaming community (the guys with some the fastest personal computers) towards contributing to these projects. I often see comments that seem to imply to the noncontributors that they are "slackers". I think this is a very good trend.

Likely, one day it will be considered 'in' for all to contribute to a distributed computing project.

[Live Forever]

Looks like Folding@Home is going to be recognized in the Guinness Book of World Records as the most powerful distributed computing network in the world:
http://kotaku.com/ga...cord-317151.php

Neato.

[dnamechanic]

Thanks Live,

A good story and the other articles associated are good ones also.

This is another positive indicator. Good and well deserved publicity for the Sony/FAH collaboration. This sets an excellent example for the other major game console manufacturers, Nintendo and Microsoft. Hopefully they will soon have some hardware that will be amenable to participation in distributed computing (DC).

Also, as publicity increases more people will choose to chip in to DC with their home computers. Currently, there are so many problems to solve and even though considerable computing power is already available it is not being utlilzed.

Maybe it will soon be that all game hardware can participate in distributed computing. That could get the topic of the subtitle of this thread underway.

[A941]

Ive posted this here under computing but no one is interessted so i hope it will get more attention in this thread, it is similar to maestro949s idea:

Somtimes i asked myself is there a bigger inteligenc behind a crowd?
Is evry single person in a crowd acting like an individual or is he acting like a part of a collective (like the Borg do)?
And if they were acting like a collective, how is the internet working, could it be the first artifical intelligence made by single "natural" intelligences, Humans as neurons of a much larger beeing?
Is there a big A.I. allready "out there" ?
Could we build our own artifical intelligence made of human "neurons", could we build a distributed thinking project with computers connected with thousand other computers like a real neuron?
I think, all we need is an interface to those who will ask the Questions and a software which will manage the data streams.

[maestro949]

Ive posted this here under computing but no one is interessted so i hope it will get more attention in this thread, it is similar to maestro949s idea:

Somtimes i asked myself is there a bigger inteligenc behind a crowd?
Is evry single person in a crowd acting like an individual or is he acting like a part of a collective (like the Borg do)?
And if they were acting like a collective, how is the internet working, could it be the first artifical intelligence made by single "natural" intelligences, Humans as neurons of a much larger beeing?
Is there a big A.I. allready "out there" ?
Could we build our own artifical intelligence made of human "neurons", could we build a distributed thinking project with computers connected with thousand other computers like a real neuron?
I think, all we need is an interface to those who will ask the Questions and a software which will manage the data streams.

Emergent super intelligence via our ability to interconnect our ideas through the internet is already accelerating every aspect of technology and information at exponential speeds, and in my opinion, is just getting started. I believe this what you're referring to:

Symbiotic Intelligence

Wiki: Collective Intelligence

We can accelerate this by building and participating in collaborative and distributed efforts like F@H but I think the biggest opportunity for potential is when we combine both the power of human minds (and their numbers) with the planet's raw distributed computing power.

See Wiki: Human-based computation

We could not only apply this concept to protein folding and molecular dynamics but just about any complex problem that requires enormous computing horsepower and significant quantities human visualization and imagination skills. Every scientist on the cutting edge of the scientific event horizon should constantly look for ways to attack problems in this fashion. Biological complexity with all of it's 3D proteins and interaction networks seems to be a perfect target for this type of work.

Another way to look at this: If 1 person can build a 20,000+ piece puzzle, could a few billion people build a 100 billion piece puzzle?

[maestro949]

I just picked up a PS3. I played Ratchet and Clank for about an hour, watched the Blue Ray version of Spiderman 3 for 10 minutes but now it's folding proteins .

The Cell/B.E. Challenge '07 interests me more. Some of the work people are already doing with the Cell's multi-core architecture is quite impressive to say the least.

MIT has aPS3 programming course as well.

[Live Forever]

Uh oh, sounds like you bit the bullet and got one. Don't get too addicted. )

[Mind]

I would live to have a game system, the games are so enteraining and with nintendo games you can exercise your body and mind (more so that with mmorpg's or fps games). Alas, I am afraid I would get too addicted and waste too much time.

[dnamechanic]

I just picked up a PS3.  ...but now it's folding proteins .

Great, Maestro.

It is very good to have another PS3 on TLM F@H team.

The Cell/B.E. Challenge '07 interests me more.  Some of the work people are already doing with the Cell's multi-core architecture is quite impressive to say the least.

...MIT has aPS3 programming course as well.

Happy to see your interest in this. As you know, the PS3 is making quite a stir in the technical world.

The Institute of Electronics and Electrical Engineers (IEEE) has published several articles on PS3 related topics (on both the cell processor and programming for the cell). Articles related to the processor have been very positive. But, apparently the parallel processing is daunting for the programmers, some think that game developers are just not up to the task

Or maybe a better way of putting it is: relatively, it seems so much easier to develop games for competitive products offered by Microsoft and Nintendo, that PS3 games are just not cost effective.

This situation probably represents a golden opportunity for a new generation of programmers.

[maestro949]

Uh oh, sounds like you bit the bullet and got one. Don't get too addicted. )

Heh, I don't have time. I'll be dead shortly and would rather work at the seemingly futile task of designing a future where a) I won't be killed by a horrible aging process and b) time is no longer an issue. Then I could play all the games I want

apparently the parallel processing is daunting for the programmers

As a software engineer I can attest to the difficulties regarding distributed algorithms but it's rather insignificant when compared to the complexity of biology. I firmly believe that a present day human will not be able to model biology's complexity through traditional software engineering practices. Pieces perhaps but the ultimate engines that crunch the numbers and run the predictive analysis tools will need to be evolved through genetic algorithms. Ironically, just like biology, I suspect we will not fully comprehend how these algorithms will work. It doesn't matter though - as long as they do.

[maestro949]

14 hours in and it's almost completed it's second work unit already.

The user interface is really nice. It allows you to select different views of the atoms in the protein. There are some YouTube videos that illustrate.YouTube videos.

[Live Forever]

14 hours in and it's almost completed it's second work unit already. 

The user interface is really nice.  It allows you to select different views of the atoms in the protein.  There are some YouTube videos that illustrate.YouTube videos.

Really buzzes along it appears. Sweet! [thumb]

[Live Forever]

It appears that PS3s have passed the petaFLOPS barrier all on their own now. (1.3 total petaFLOPS for the whole project):

http://fah-web.stanf...y?qtype=osstats

[Live Forever]

They are now using the data they are getting for a variety of other means, as well as helping out other groups by freely sharing their GPU technology and stuff they have developed:
http://folding.typep...nghome-and.html

These guys are really great.

If you haven't yet, please think about joining our Folding@Home team to increase life extension visibility and help out a great project at the same time.