Talk-Polywell.org

It is something I have been thinking about (also in context of the Google- Tri Alpha- cooperation). At some point simulations and AI will become so good, that it will affect how we develop fusion reactors. At least we will be able to test ideas without having to build prototypes anymore. But I think that it will go even further than that at some point in the more distant future.

The current sim is based on First Principles, and is very detailed. It has produced some really good work. It is well beyond their sim work of the past.

ladajo wrote:Yes, I would say it is the most advanced 3D particle plasma sim out there. Part of the work has been to create the software. It runs on a multi-cell super computer. The outputs are impressive (and pretty).

I must have missed something. I've been glancing over the recent entries in this thread, and I am getting the impression you are a lot more intimate with these doings than the rest of us.


Did you somehow get a backstage pass on this organization and I missed the announcement?

It is not news to some that I have intimate knowledge, not sure how you missed that. Granted, a select few have know that much longer than others. I remain limited in what I can share for now. I am also surprised that no-one has gone patent hunting of late.
Hopefully, within the next few months, more will be in public. Be patient.
Really, the next big issue is resourcing the next major phase. And this remains dependent on completing the sim work. Phase II Sim Runs(as I call it) are underway now.

ladajo wrote:I am also surprised that no-one has gone patent hunting of late.

I admit that I have looked every now and then to see if the current patent application has been accepted, but I have not found any news on that. I have also not found any new patent application.

There is more than one on this forum that is close.(not me but others)

Simulations are very useful. And NOT to be trusted without validation, for which you need to run physical systems. My experience in computer simulation tells e that even with all the assumptions right there are ways for a simulation of a complex system to go wrong.

I second that. NEVER let yourself believe that the simulation is reality. They are much more helpful after experiments, used to discover where the current understanding of reality is amiss.

Of course simulations need to be confirmed by experiment, especially in a field like this. But looking into the future, they could be used to rapidly test millions of combinations before going into actual experimental verification, which could vastly accelerate nuclear fusion (and other) research.

My experience with models is mixed. Run tests, get loads of data, tune model to fit, make a change and all the test need running again for new tuning parameters. In other words, some complex systems have extreme sensitivity to boundary conditions.

On the other hand if you run the model, get results, test and data looks good, then you may have something. It is not uncommon for those that have not worked with complex systems to not grasp that some systems are just nasty.

I would be interested in why a low profile is desired? I have an extreme distrust of secrecy. For instance in real estate it a common practice to use delay and promises to secure another property, that once secured screws the the other. Or, secrecy could be used to enable a foreign country or investor to do a deal contrary to American interests. Or, it may be used to prevent a bidding war. Etc.

Aero wrote:I second that. NEVER let yourself believe that the simulation is reality. They are much more helpful after experiments, used to discover where the current understanding of reality is amiss.

Yup.

ladajo wrote:Yes, I would say it is the most advanced 3D particle plasma sim out there. Part of the work has been to create the software. It runs on a multi-cell super computer. The outputs are impressive (and pretty).

Explain to me the algorithm, please?

Is it an all-pairs method? If so, does it use fast multi-pole method or some variation of?

is it particle in cell, discrete element...?

how are the "cells" broken up? different spatial regions? different regions of phase space?

how are the outputs "pretty"? what is being outputted? a simple spatial representation?

so many questions...

You will have to wait. Sorry. I can share that the coding effort was collaborative between some of the top minds in the field.
There is nothing else like it right now. It is the cutting edge for plasma sims. and proprietary.

ladajo wrote:You will have to wait. Sorry. I can share that the coding effort was collaborative between some of the top minds in the field.
There is nothing else like it right now. It is the cutting edge for plasma sims. and proprietary.

Then I bet it uses my distributed dynamic octree idea for massive scalability with near O(n) performance.

ladajo wrote:I can share that the coding effort was collaborative between some of the top minds in the field.

Wait, by top minds do you mean top physicists or top coders?

big difference...