Re: Fwd: We Can Understand Anything, But are Just a Bit Slow

From: Ben Goertzel (
Date: Fri Apr 28 2006 - 13:31:30 MDT


Certainly, the goal of the Novamente design is to have a full semantic
understanding of the world around it.

Right now, the existing Novamente system is pretty dumb in practical
terms -- we are teaching it stuff like how to play fetch and the fact
that object still exist when it's not looking at them. However, the
learning algorithms in the system are very powerful (as has been shown
by applying them in data mining and other domains besides AGI), and we
believe that as we further build out the architecture and teach the
system it will get smarter and smarter, in the manner of a human baby.

But, even the current NM system's understanding is better described as
semantic than syntactic, since it is interpreting percepts and
generating actions based on it a desire to achieve goals using its
simulated embodiment, rather than just operating within a disembodied
formal system.

-- Ben

On 4/28/06, Woody Long <> wrote:
> > [Original Message]
> > From: Ben Goertzel <>
> > To: <>
> > Date: 4/28/2006 2:34:31 PM
> > Subject: Re: Fwd: We Can Understand Anything, But are Just a Bit Slow
> >
> > To give just a hint of how these distinctions manifest themselves in a
> > fleshed-out AGI design, in Novamente:
> Dr. Goertzel,
> As a fellow designer, I must say wow, you are well on your way. The more
> posts of yours I read here about Novemente, the more impressed I am with
> your development of machine intelligence.
> Without reference to the analogy, perhaps you would be interested in a
> phone conversation I had with Professor Searle yesterday. Here was the
> heart of it -
> WL: A syntactical machine can never be conscious. Correct?
> PS: Yes.
> WL: And semantical machines that have a semantic understanding of their I/O
> are conscious. Correct?
> PS: Yes. And that's the question. How do you get this semantic
> understanding.
> WL: Exactly! That's the key point. And my prototype of my invention is
> doing just that.
> PS: Then send it to me. (*In a stern tone that implied 'I will believe it
> when I see it.'*)
> Irregardless of whether this is true, I feel curious to ask you, though you
> may not want to discuss such details, simply put: Has Novamente crossed the
> great divide between syntactical, simulatory Classical Machines, and
> Post-Classical strong AI conscious machines? In other words, does it have
> a semantical understanding of its language inputs and outputs, or is it an
> extremely clever / ingenious syntactical simulation of intelligence? Either
> way, congratulations! It is solid evidence that machine intelligence is
> close at hand ...

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:56 MDT