I don’t know who will win the Presidential sweepstakes this November, but I hope that it is either a person who understands computing technology, or someone who has a batman who does.
But I’m not holding out hope.
I was having a beer with a good friend, someone whom I greatly respect with regard to computing insights. He was talking about the ARM11 core and the fact that it had a vectorized floating point unit. We had been talking about the awesome computing power available in the Cell microprocessor from IBM/Sony/Toshiba and he offered the ARM11 as counterpoint and then said that it had the same processing power as a Cray XMP.
The ARM11 processing core will be used in cellphones. The Cray XMP was the state of the supercomputer art in 1982 and was used to design nuclear weapons. The measure of floating point performance is the FLOP, or floating point operations per second. The Cray XMP was rated at 400 megaflops for a two processor machine. The ARM11 has a rating of 4.2 peak gigaflops, an order of magnitude better than the Cray. The peak rating of the ARM11 core would be hard to hit, but it could certainly hit 400 sustained MFLOPS.
Our conversation about the Cell microprocessor was really more about the Sony PlayStation 3 and using some PS3s to make a supercomputer cluster. The Cell processor in the PS3 has a rating of around 100 sustained GFLOPS, so a cluster of four PS3s would deliver three orders of magnitude more processing power than a CRAY XMP. The PS3 sells for $400, with three Blu-Ray movies.
As a measure of the power of the PS3, researchers at Stanford University have ported their Folding@Home project to the PS3, and spare CPU cycles are being used to work on protein folding problems. This distributed network is now recognized as the most powerful computing network in the world.
Here is the issue: as my esteemed friend, who is ABD’d in nuclear physics, pointed out, it is a rather trivial task for people to go back to first principles and model nuclear reactions on the abundantly cheap computing hardware now available. Solving the differential equations about placement of explosives required to achieve fission is eminently achievable. Solving the mechanical engineering problem of refining fissionable materials is less easily done. But if the goal is not to achieve a big explosion but to just spray highly-toxic, long half-life fission byproducts in a major metropolitan center, that is not that hard to do.
Which brings us back to the Bush Administration. The people at the top don’t really understand or care about the technology required to be a threat. Before 9/11, the Administration was moving to listen in to all traffic flowing between to off-shore points. The focus of this Administration has been people, not technology. They started a war with Iraq that was based on personalities (”this is the guy who tried to kill my dad“) and were willing to sacrifice intelligence assets to cover up their lies.
We should recognize that the ability to design nuclear weapons will be in the public domain in the very near future. The spread of cheap computing assets will make that happen. The bulwark against the spread of real nuclear weapons has two parts: 1) interdiction of fissile material trade; 2) helping to build countries/economies/systems where the possession of nuclear weapons is not an asset. Our intelligence services are working on the first one. The flap of that tent was lifted during the Plame affair.
The second part is harder and is akin to that joke test question: define the universe and give two examples. The causes and effects of the things that drive countries toward nuclear proliferation are like a mesh where there is no clear beginning or end. But the sure lesson that has been learned is that if a country has nuclear weapons, they won’t get invaded by the US.
This is the wrong lesson to learn.