[ExI] Current uploading sales brochure?

[Eugen Leitl]

On Tue, Apr 23, 2013 at 04:18:20PM -0400, Alan Grimes wrote:
>
>>> claims about performance were not well justified. I would definitely
>>> prefer to work from more primary sources such as technical papers rather
>>> than have to rely on other people's fiction.
>> What do you want to know? Your earlier question about clocks to synchronize
>> low-level stuff: you don't need clocks for that. What else?
>
> Point 1: that seems to contradict, or rather be strongly incompatible  
> with the often repeated claim that "clock speed" was a variable that  
> could be tweaked or otherwise manipulated.

Asynchronous free-running systems do have a natural refresh rate,
but it's not due to a global clock. You can synchronize at higher levels
by using a system of locally coupled oscillators. Look at how
biology does it. A good book to look at is
http://www.amazon.com/Sync-Order-Emerges-Universe-Nature/dp/0786887214/

> Point 2: All game engines I know, to the best of my understanding, are  

Game engines run on classical computers, even if it's GPU-accelerated
physics. Classical computers are largely useless for production quality
neural emulation.

> strongly based on at least two critically important clocks:
>
>     1. The sound system's sample-rate clock which dictates what voltage  
> to send to the speakers/headphones at each instant.

I don't see how this is relevant for anything.

>     2. The frame rate clock off of which the entire rendering engine  
> hangs...

Again, this is how classical computers work. It's not a natural
law, look at asynchronous, clockless machines.

> If any of these clocks (both mentioned above and not mentioned) are not  
> perfectly synchronized with each other and the systems simulating the  
> brain, which also has a number of local and global synchronization  

If you're simulating the brain in realtime or faster, you're using
a massively parallel system which directly represents patches of
circuitry. An advanced design would not have a clock, but even if
it had, it would be transparent to the simulated system due to
time scale differences.

> problems, then the subjective experience will suffer and the oft and  
> emphatic claim that it will be "indistinguishable from reality!" will  
> not hold.

If you're working at low level of theory, and your simulation is
not accurate, then you'll get a lot worse than just rendering
artefacts. E.g. total energy would not be conserved, and your
system would suffer an excusion into regime not representing 
anything happening in biology. Think of this as a grand mal on
steroids -- you won't be there to watch yourself twitch.