Deadly Sins of Real AI

From: ben goertzel (ben@goertzel.org)
Date: Tue Apr 02 2002 - 09:30:47 MST


***
Okay, it's easier for you to go down the current path you're going. Where
is that path leading? How far are you currently along that path?
***

I break our progress down into 3 phases

Phase 1: Implementation

Phase 2: Basic teaching (via experiential interaction in a shared
perceivable/manipulable environment)

Phase 3: Significant self-modification of learning mechanisms & data
structures

We are currently about 1/3 of the way through Phase 1, I'm afraid.

With the Webmind AI Engine we got about 80% of the way through Phase 1 and
did some experiments in Phase 2, but the other 20% of Phase 1 was
unfeasible in the framework we had at that point. I believe we have
solved those problems and that the completion of Phase 1 now is "just" a
matter of implementation and testing and detailed software design based on
our completed mathematical/high-level software design. (And, lest someone
get excited: When I say "mathematical" I do not mean that we have proved
the system will work mathematically, just that some key parts of the design
are formulated in terms of equations.)

***
 From what
I've seen so far you seem to think that this is the path leading directly
to
Real AI.
***

I think it is *a* path leading directly toward Real AI. I am sure there
are many other paths too. For instance, in my recent paper Hebbian Logic,
which I sent you, I sketched a possible alternate path.

***
  But I'm no longer sure what you mean by Real AI. Obviously Real
AI includes something that a human would intuitively recognize as a person,
but what else does it include for you? Does it include everything that I
would call "a functioning deliberative supersystem"? Because I would
consider that the first grand milestone, or epoch, not the last one.
***

Eli, a search for "deliberative supersystem" on Google does not turn
anything up. So I suspect this is not a phrase you have used in your
online writings. If you will define it for me I'm willing to discuss it
with me.

Based on my intuitive interpretation of the phrase, I would say: Yes, I
intend to create a functioning deliberative supersystem. (Wearing a tight
blue uniform like Superman if you like, to show off its hunky little buns
;)

I think that I have clarified my goals for you many times. First a
roughly-human-equivalent intelligence that can be taught and conversed
with, then (growing out of this) a self-modifying intelligence that can
bootstrap its way to tremendously superhuman intelligence. If successful,
this will lead either to the Singularity or -- perhaps, I hope not! -- to a
system that understands the hidden obstacles to the Singularity that we
stupid humans are missing...

I don't use the same exact vocabulary as you to describe my long-term
project goals, but I should think that by now you'd understand them
qualitatively.

***
I realize that I have no code, but I have put my design online and let
people judge it, with another major update coming Extremely Soon Now. I
haven't seen your design yet, I haven't seen your code,
***

You have put nothing online that I would consider a concrete AI design. A
lot of interesting ideas, to be sure, though.

I don't think your online writings are any more explicit than my various
online and published articles on Webmind. (For instance, the chapter on
Webmind AI Engine in my latest book CREATING INTERNET INTELLIGENCE).

As for seeing my design, as you know, we're currently dickering over the
details of the nondisclosure agreement. So if you really want to see it,
you probably will get to soon. I look forward to hearing your detailed
critiques.

**
 I haven't heard any reports of what it is
exactly that Webmind or Novamente do - except for text categorization and
stock market observation and so on, which you could be accomplishing in any
number of ways aside from what I would regard as "Real AI".
***

In fact, these practical functions of Webmind Inc. products *were* achieved
using relatively simple algorithms, not using "general intelligence."
 These simple algorithms were inspired by our work on general intelligence,
however.

***
 I can see that you've gotten
over the Deadly Sin of AI in which you only use one idea, but there are
other Deadly Sins.
***

Would you favor us with a list of the "N Deadly Sins of AI According to
Eliezer" ?

Sins that merely cause severe illness rather than death may also be
included.

;-)

ben



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:37 MDT