A Proposal for a Popularly-Funded Prize in Computational Chemistry

(c) 1997 Will Ware

Recently on sci.nanotech, there was a complaint that too much of today's discussion of nanotechnology is amateur pontification, not informed by any experimental or computational results. So I discussed an idea of offering a prize of a few hundred dollars to anybody who could animate a few picoseconds of physically accurate behavior for a serious nanotech design (for instance, Eric Drexler's fine motion controller design recently posted on the IMM web site). To win the prize, somebody would need to not only create the animation (in the form of JPEGs or a Quicktime or AVI movie, or whatever), but also post as freeware all the source code that was involved in creating it.

The more I think about this idea, the more I think it's a good thing. Whoever wins the prize gets some fame and a small immediate financial benefit, and the longer-term benefit of being able to sell videotapes of the animation. Other people win by the free availability of molecular modeling source code capable of doing useful animations. The only person who doesn't win is whoever coughs up the few hundred dollars of prize money. But there's a way to make this more equitable: the cost of the prize money can be distributed among all those who would consider themselves beneficiaries if the prize were won.

Here's the plan

After some public discussion, a consensus is reached about what the criteria should be for winning the prize. These criteria should be very clear and very binary, with no opportunity for ambiguity or differing interpretations.
Either:
- A small panel of judges is assembled. Their job is to decide when somebody has won the prize. They must reach unanimous agreement in order for the prize to be awarded. Or:
- Somebody cooks up a test suite to make the judgement of whether a body of software passes the criteria. Anybody should be able to use the test suite to verify for himself or herself whether the prize criterion has been met, so the suite must be easy to run on a variety of popular computing platforms.
People who want the prize to be won indicate their interest by publicly pledging some amount of money to whoever wins it; call them "investors".
Some people will try to win the prize, by submitting their animations and source code to the judges. Call them "contenders".
When the prize has been won, the judges make an announcement, and the investors mail their checks to the winner.

Nobody pays more than they want to pay. Investors may remain anonymous if they wish, announcing their pledges thru the judges but having their names withheld, or not announcing them at all. (The obvious benefit in announcing one's pledge, or notifying the judges so that they can announce it, is to encourage people to try to win the prize.) Assuming market forces really work, the value of the prize should fairly accurately represent the value to the nanotech community of the work involved in winning the prize.

There is nothing in this scheme to prevent investors from welching on their pledges. They are bound only by the honor system, and prize contenders must understand this when they decide to try for the prize. For this reason, investors should only pledge amounts that they are sure they're willing to pay, even if they don't like the judges' decision. Investors should refrain from comparing their pledge amount to those of other investors, and choose their amount based only on the benefit they feel they'll receive from the prize having been won.

Another thing I like about this scheme is that even though none of the investors is paying more than he or she wants to pay, there is nevertheless no fundamental limit on the amount of the prize. In principle, the prize money could be big enough that a contender could take time off from their day job to win it. (This constitutes an obviously dangerous gamble, and is not recommended.)

Possible Prize Criteria

The animation must be physically realistic. This could be verified by specifying some very simple test cases where a few atoms bounce around for a few femtoseconds. If the contender's software doesn't produce the right trajectories, it is assumed not to be an accurate molecular modeler.
The animation should run for a sufficiently long time that some interesting behavior is visible. This might mean a few tens of picoseconds. There might be a lower limit on how many atoms were involved in the animation, or it might be required that a particular structure be used (e.g. Drexler's fine motion controller).
The source code that wins the prize must be freely redistributable, under one of the standard licenses (GPL, LGPL, Perl license, BSD license, public domain...)
The source code must compile and run on at least one of the following platforms: SunOS, Linux, Win95, MacOS.
The format of the animation must be a standard format (Quicktime, AVI, a series of JPEGs) or something that can easily be converted to a standard format. (Refer to a later section for more about graphics and animation.)

Why is this a Good Idea?

Making this kind of software widely available and easy to run, on widely available computing platforms, will raise the level of nanotechnology discussion. This will make it both more interesting and more useful in guiding the development of nanotechnology in benign directions.

The total size of the prize is itself an interesting number. Like a stock price, it will periodically move up or down (investors may decide to pull out, or alter their pledge amounts). Like any price, it gives information that is not available in any other form. In this case, it's the estimated worth of the software to everybody who might want to use it.

Another interesting thing is that this is an experiment in compensating the development of free software. Lots of free software has been written, some of it very good, but financial compensation goes not to the author, but to the people who sell floppies and CDROMs and support contracts. There is no existing financial mechanism for compensating authors of free software; this might be a good model for that.

The Animation Part

This part is easy. In fact I've written some software that does a big part of the job, and the rest can be done with easily available freeware and shareware. The input to my animating program is a standard molecule file format called XYZ. The output is one or more line-interleaved RGB files, and these can be converted to all the popular image and animation file formats using ImageMagick (a freeware image processing toolkit), and AVI movies can be made with a shareware program called AVI Constructor. Video production houses can transfer AVI movies to videotape.

The XYZ format for the 3D positions of atoms at a particular moment (what you'd call a frame in animation or television) looks like this:

          178
 Energy:   -1209.775535419484    
     C             5.47239             0.00000             6.36850
     C             5.33518             1.21772             5.66100
     C             4.93045             2.37438             6.36850
     C             4.27848             3.41198             5.66100
     C             3.41198             4.27848             6.36850
   (...173 more atoms...)

The first line gives the number of atoms. The second line is a comment; it can say anything you want. Then you have one line for each atom, giving the name of its element, followed by X, Y, and Z coordinates in angstroms.

A multiple-frame XYZ file is simply a concatenation of frames with no blank lines between them. It's a simple easy-to-parse format that compresses well (I get a factor of seven compression using gzip). It gives an easy way to represent the trajectories of a bunch of atoms over time, which is what you'd want to be able to do to win the prize, now that the animation part is trivial.

Given the existence of this software, the problem of winning the prize is reduced to producing a multiple-frame XYZ file. The prize should therefore concentrate on making the trajectories as physically realistic as possible.

What is Already Being Done?

I briefly discussed this idea with Al Globus, who is doing nanotech simulations of fullerene-based gear structures at NASA. He points out that there are a couple of groups already doing work like this:

I'm glad to hear about these efforts, but I guess the thing I'd like to see is software that allows the average nanotech enthusiast sitting at home with his Linux or Win95 box to do this stuff. When the software is capable of running on reasonably standard home PC hardware, and when enthusiasts are educated enough to understand what they're doing with it, then we can expect both a profusion of interesting nanotech designs, and a livelier and more interesting tone in informal nanotech discussions.

I should add a last point: I haven't looked carefully at the software discussed on these two websites. It may already do everything I'm looking for, and already be able to run on vanilla Linux boxes or other common home hardware. Maybe it already exists.

But even if that were the case, there could still be something useful to do: write a book or a manual (that isn't too intimidating) that explains how to use the software and what everything means. This would fill a welcome niche between complete ignorance and the comprehensive but intimidating Nanosystems. If that turns out to be the case, perhaps there's some way to structure a popularly-funded prize that incentivizes the writing of such a book. That sounds tougher; judgement of a book's quality is going to be much more subjective, and subjectivity will probably make a popularly-funded prize fail miserably.