Friday, December 31, 2010

Project - Mc Lab / Magic Chemist, in a Box.

Maybe I was a bit naive when I drafted this. Or maybe I am just lacking the right letters after my name.

===========================
Project - Mc Lab
Magic Chemist, in a Box.
John L. Sokol - Jan 2009

Magic Chemist, in a Box is the simplest way to explain something that is right out of Star Trek. Magic Chemist is a supercomputer that explores chemical space and detects new chemicals, and formulas then catalogs them in a database.
Once complete, Magic Chemist is capable of performing; a quantum simulation of the atoms in the chemicals interacting, determine the properties, existence and variation of these molecules, how they interact with other molecules. It can report their physical properties, chemical changes and stability.

More importantly it can intelligently search for chemicals that are optimal for different applications. In addition Magic Chemist will perform stimulated experiments without a lab or wet chemicals, thus preventing; fumes, toxins, disposal, regulations and human resource needs. This can and will be done with incredible speed and accuracy.

This could lead to large improvements in super conductors, semi conductor, solar cells, carbon sequestration, water purification, bio-fuels, batteries, super capacitors, paints, dyes, food and medical applications.

Is There Currently Anything On The Market Like This?
Initial research shows that there is nothing currently like this on the market. The fields of computational chemistry and computational biology use these types of simulations, however they are usually run on PCs or small computers in hopes of providing some understanding of things that have already been discovered or isolated in other research. The field of computational chemistry date back to the late 1930's and by the 1950's it was used to understand the benzene ring. Today there are about 20 off the shelf applications for running these types of simulations.

One of the largest is "protein folding at home" http://folding.stanford.edu/, it uses Distributed Computing. A process where people around the world participate by leaving the application running on their PCs thus donating time to the project.

What is Unique About This Project
None of the current projects are pragmatic. In addition, they are not searching for practical profitable formulas that can revolutionize entire industries. In addition, there hasn’t much attention has been given to chemists by computer people. In general they seem to just get to borrow some CPU cycle on a super computer but don't generally get to purchase that level of hardware for just their application. However, this is currently being done by geneticist and biologists.


What Are The Steps To Achieve This?
  1. Design and build hardware capable of performing the required simulations very quickly. I estimate I can build create a 2 to 4 Petaflops for $1M USD.
  2. Log these results. (Chemical Space) of the infinite possible molecules and chemical interactions possible. This information can continuously be data mined for gems. In addition, revenues could be generated by selling copies or charging for access to the database. This would require a copyright and/or patent as well.
  3. Set the system up in a feed back loop to a targeted search for the most likely chemicals to try.

Project Financing
I am seeking $500,000 for an initial design feasibility study.

This initial design feasibility study would include further investigation of the hardware design, investigation of the code available and what needs to be written. The next step is to get demonstration code working to understand the CPU requirements of the system. We would next hire several technical consultants in the fields of physics and computational chemistry to assist and verify the accuracy of our design and plans. Ideally later these would be early hires in to our company.

The next round of financing would require another $30M in funding to proceed towards the completion of final design.

I am expecting that the first major discoveries could be within 3 to 4 years after funding, less than 12 month after turning on the system. I would like the first target to be "the" optimal dielectric for use in Super Capacitors that would replace batteries technology in almost everything from laptops to Electric vehicles. They charge almost instantly, never wear out, and are non-toxic.

Exit Strategy
Because of the nature of the project, advantages don’t lie within typical business strategies. In essence the plan is not to exit because of the value of information and discoveries produced.

This is a Golden Goose.
Supply electricity and some manpower in one side and out pops a stream of new discoveries that can be sold, licensed or rolled in to another company to produce and sell it. You will not want to patent this internal technology, publish or press release this. It would be best to keep all traces of the existence of this secret.

Compare this to a stock predictor that works, as long as no one knows about it, you have the advantage. In some respects this is like a pharmaceutical company but without the regulations. It's a long research cycle with huge payoffs. But after we discover something we could know very quickly if it will pay off or not commercially.

===============================
Responses and Addition discussions. 
===============================

Dear John,

I don't have the reference, but about a decade ago I saw an estimate  on how many petaflops it would take to solve the Material Science Inverse Problem:  i.e., input the desired characteristics of the  material, be it "transparent aluminum" from the Star Trek movie, or a super dielectric or cheap room temperature superconductor that would found an industry and make you a billionaire, and have some  supercomputer cluster grind away with numerical solutions of quantum mechanics and output the chemical formula and structure of the desired super-material.

I don't think that we're quite at the computational power needed, yet, but I'll bet we're getting close.

There are dozens of Quantum Mechanics software packages that do minimum energy and minimum entropy and so forth.

I've written a (not accepted) paper on those relevant for Computational Biology, and what is needed to break through.

We are still learning what questions to ask, as well as the design  requirements for the needed software.

This is at the fractal border between theoretical and applied and computational research.  Institutions with good track records keep submitting grant applications for this.  We, as de-institutionalized  outsiders, are pretty well kept away from the feeding trough. Except that great enough research papers can be done once in a while, and published, and then one can bid on who gets the administrative overhead for research in which one is Principal Investigator.
But cool research is cool research, and I'll do it even when somebody pays me.

Good questions, John. I know you as a prolific and brilliant inventor whose ideas you can reduce to practice, and hence you are deeper in Pragmatism than Theory.  But I've found you a quick study at Theory,
too.

Best,
JVP


The rewards for something like this can be priceless for the Human race.  From Carbon sequestration to solar panels and all the way to construction of the space elevator.

Yea, I am winging it here on the theory, it is over my head. That why I have friends like you around.
I don't think that we're quite at the computational power needed, yet,but I'll bet we're getting close.

Digging up computational power, now this is something I can do.
And petaflops aren't that hard to reach these days, I can get there on 1 million if I build an application specific machine based on FPGA's or some sort of Cell processor like they are using for computational biology.  I agree using off the shelf PC parts wouldn't get us that far unless we did the SETI at home, where we take advantage of people's unused PC's computing power.

There are probably all kinds of short cuts we could also take too for reducing the calculations to be able to eliminate molecules that are clearly not suitable, this would also speed things up a whole lot.

I find most college code to be terrible, and inefficient unless they are specifically CS students and even then it's rare.

But if I can get some working code from some place, I can clean it up, optimize it and even work on some custom HW platform to run it.   Also Genetic Algorithms could really speed the discovery process.

I mean how much computation would be required to simulate say an H2O molecule and test it's properties?

John


===============================
Links
===============================

Youtube: The Parallel Revolution Has Started: Are You Part of the Solution or Part of...

http://nextbigfuture.com/2008/08/thermoelectrics-and-refrigerators.html   
Quantum Simulations Group at Lawrence Livermore National Labs
simulations [modeling material processes using quantum molecular dynamics methods]
For example, nanoscale materials could improve cooling technologies in military equipment and reduce the size of gamma radiation detectors being developed for homeland security.
http://physci.llnl.gov/Research/qsg-old/index.html


Why sustainable power is unsustainable
http://www.newscientist.com/article/dn16550-why-sustainable-power-is-unsustainable.html
Founder of IdeaLab TED speech
Used Genetic Algorithms to develop solar collector design.

Bill Gross of Idealab Talks About his Dream of Cheap Solar Power
http://www.ted.com/index.php/talks/bill_gross_on_new_energy.html
This company
is doing something similar to what you have suggested, except that they do this internally, for the pharmaceutical industry.

GPUs Used To Crack WiFi Passwords Faster
http://it.slashdot.org/article.pl?sid=09%2F01%2F15%2F1334222


http://en.wikipedia.org/wiki/Quantum_chemistry_computer_programs

http://en.wikipedia.org/wiki/Ab_initio_quantum_chemistry_methods

http://en.wikipedia.org/wiki/Computational_chemistry

http://www.lecb.ncifcrf.gov/~toms/glossary.html

Applications of Quantum Mechanics
http://scienceblogs.com/principles/2009/01/applications_of_quantum_mechan.php

No comments:

Post a Comment