Ben,

On Mon, Jul 9, 2012 at 7:35 AM, Ben Goertzel <[email protected]> wrote:

>
> On Mon, Jul 9, 2012 at 10:23 PM, Steve Richfield <
> [email protected]> wrote:
>
>> Hi all,
>>
>> I wonder whether most people here grok that the world is fundamentally a
>> gargantuan set of simultaneous nonlinear differential equations for us to
>> solve in order to achieve our goals?
>>
>
> We all know that is the way physics conventionally models the world... but
> not all of us think that's the most useful model to use for AGI purposes
>
> One could just as well say "the world is fundamentally a gargantuan
> computer program" -- this view is also consistent with all known physics...
>

There are some BIG weaknesses in this POV:

1.  Differential equations describe a system. Errors in the equations map
to errors in the system being described, and can be directly corrected by
adjusting the equations to eliminate the differences between
simulated/solved and observed results. Of course, computer programs can be
similarly adjusted, but ONLY when they are also in effect paralleling
reality, and so we are comparing like things.

2.  Perhaps you can explain. I know of no way of making computer code
consistent with known physics, except through EXTREME programming effort
that in effect makes it solve the same differential equations, albeit
potentially by programmers who don't even realize what they are doing.
Again, we are comparing like things.

In short, sure you can do pretty much the same with a computer program -
which is EXACTLY not only what I am proposing, but my proposal leads fairly
directly to the detailed mechanisms needed to accomplish this. Wonderful
how great minds think alike.

It appears to me that the primary point of failure in our discussions is
that you haven't yet "seen the forest for the trees", that in the long
distant future when you FINALLY get your code working the way you want it
to work, that it absolutely MUST in effect be learning and solving vast
systems of differential equations, regardless of whether or not its
programmers ever realize it. Of course, realizing what you are doing goes a
LONG way in directing your efforts in a productive direction.

The (only?) remaining wrinkle in what I have been saying is that Von
Neumann style computers are REALLY REALLY poor at solving systems of
differential equations, even arguably worse than old vacuum tube analog
computers, and hence similarly poor at executing software that "pretends"
to solve differential equations, even when it is written by programmers who
don't (yet) realize what they are trying to do.

Sticking your fingers in your ears while loudly saying "NO NO NO" while
ignoring the math, doesn't alter the fact that you are in effect trying to
write code that solves systems of differential equations. Of course, people
who ignore the underlying math are usually doomed by it. Do you know of any
exceptions to this rule?

There are other radically different architectures that natively solve
systems of differential equations, so why pretend when you can more easily
do the real thing, unless you are simply trying to quickly produce
brain-dead demonstrations of pseudo-AGI, akin to the original Eliza?

We could simulate such a machine on current hardware. The simulation would
be slow, but at least it would run at SOME non-zero speed, so that real and
useful research could begin.

Agreement aside, I haven't yet heard words suggesting that you have grokked
what I have been saying. Perhaps you can better relate to Sergio's POV in
his posting, made while I was writing this.

Right now, all I am looking for is an indication that you understand what
Sergio and I have been saying. Only then can we carry on a productive
conversation about it.

Steve



-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968
Powered by Listbox: http://www.listbox.com

Reply via email to