[ExI] proto-bitcoin

[Eugen Leitl]

On Wed, Apr 24, 2013 at 11:47:06AM -0400, Alan Grimes wrote:
> Eugen Leitl wrote:
>> The problem with "virtual" is that it's encrusted with concepts
>> likehttp://en.wikipedia.org/wiki/Virtual_machine
>> and makes people ask nonsensical things about underlying
>> operating systems or hypervisors. Instead, the mental model should
>> more resemble FPGA state, evolving onboard. It still
>> wouldn't be accurate, but it's a start.
>
> That's one of the things that bugs me about your philosophy.
>
> You are like "don't worry, computers won't be the ungodly monstrosities  
> of bloatware and legacy design in the future because all of computer  
> science will take a left turn and suddenly become elegant, robust and  
> operate on almost entirely new principles."

They wouldn't if they didn't have to. Fortunately, they do.

If you don't yet understand why, consider looking at how
much crunch you need today on a classical (say, Blue Gene/Q)
architecture to run one liter of neural tissue in realtime.
Now consider that you're going for a speedup,
1 ms: 1 us, 1 ns, 1 ps. Look at how much happens in
biology within 1 ms and how many "atomics ops"
you'd need to do to manage it within 1 ns...1 ps.

If you have trouble, here's a hint: if you start
with current clusters granularity scales down to
the point where 3d torus clusters become cellular hardware,
and really simple hardware at that -- with zero
software but state. It's just one big complex kernel
being applied on the entire 3d volume.

> I'm sorry, that just doesn't hold water with me. I'm going to need to  
> see some papers/books/articles that are specific enough to convince me  
> that A, it will work, B, it will be useful, and C, it will happen.

Nobody ain't got no time for that. Try researching it on your own.

> Better yet, I would like some kind of physical artifact that I can (in  
> principle) play around with... (my computer generally doesn't do  
> anything I want it to except e-mail and sometimes web browsing...)
>
> Failing either of those, I'm going to have to group you in with people  
> who believe in magic and rapture cultists. =\

Pretty much exactly the opposite. Since you already like
Minecraft (and probably seen the neuron emulation made in
Minecraft) it shouldn't take a large conceptual leap to
look what a hardware implementation with <1 um^3 cells 
of a much more complex translation rule could do.

> Furthermore, you will have to show how your proposal is either  
> compatible or incompatible with the selling points uploaders typically  
> trumpet.

I couldn't care less about what "uploaders" trumpet.
We're at an early stage, where we can do primitive single
cells and where a nematode is perhaps 30-50% done -- this
is all classically done, still.

If you want to help, look into building an integer gas model
for systems which are normally done with differential equations/
matrix algebra. For a hint, see prior work in
http://scholar.google.com/scholar?q=fluid+dynamics+with+cellular+automata&hl=en&
but don't bother with single bit sites, use small
integers instead.

For what you're trying to model, see https://www.coursera.org/course/bioelectricity
https://www.coursera.org/course/compneuro