From: Ben Goertzel (ben@intelligenesis.net)
Date: Sun Oct 29 2000 - 19:32:59 MST
>
> I don't think Ben Goertzel was talking about ownership of facts, just
> ownership of the facts in ready-to-think format for your favorite AI.
Well, this sentence raises an interesting question (which you address in a
slightly
different way later on in your message)
To what extent will it be possible to create/sell a "lobe" of facts/know-how
(i.e. declarative + procedural knowledge), that will be usable by
MANY DIFFERENT AI programs, not just one
To the extent that a "mind standard" for declarative and procedural
knowledge
interchange can be created, I think we'll all benefit
This is one way to get some of the benefits of open-source AI even in a
commercial-AI
context..
For instance, we've created a couple languages that we plan to release into
the public
domain in the not too distant future
KNOW -- a declarative knowledge representation language, sort of like a more
flexible,
fuzzy version of predicate logic
Psynese -- a procedural knowledge representation language: basically a
generic functional
programming framework where the functions involved may be black-box objects
obscuring
proprietary functionality
AIP -- Agent Interaction Platform ... a general framework for different Net
agents to
interact with each other according to languages of their choice (e.g. KNOW,
Psynese)
We also have Webmind-specific languages such as MindScript (a JavaScript
extension)
and MindSpeak (more conversational in nature), which are by their nature not
portable to
different AI architectures
Presumably in the future, if there were many different AI's all able to read
in KNOW or
Psynese files, then one could sell declarative + procedural knowledge
broadly, to users of
any compliant AI. One will have 'mind format converters' ;>
> more likely, the "format" is
> a third of
> what you need to know to create an AI, and another third is composed of
> thoughts stored in that format.
Yes, a KNOW or Psynese file doesn't tell you how to create a mind that can
use the
knowledge is contains ... it is only useful if you already have an AI system
of some
kind to load it into ...
> Another variable strongly affecting the saleability of knowledge
> is the degree
> to which the knowledge can be made independent of the particular
> context of
> the AI that created it, or the extent to which that context can
> be packaged
> with the knowledge. We aren't just talking about things like what type of
> fluctuations indicate anomalies, but the fact that real knowledge
> grounds in
> experience, and experience is perceived in ways that link into
> the rest of the
> knowledge base. So to export knowledge plus grounding, you have
> to export the
> experience, and to export the particular experiences, you may
> need to export
> the rest of the experiential and knowledge base.
This is a serious issue, which gets at the very basis of culture and
communication.
The concept "cat" doesn't mean the same thing to you as to me, as to my
3-year-old daughter. Yet we can all talk about cats together with little
confusion
Similarly, when a Webmind exports KNOW and Psynese files reporting its
knowledge, unless
it exports its ENTIRE mind, it's not exporting enough for someone to totally
understand
the knowledge it had. On the other hand, if it's communicating with an
adequately similar
mind, it can presumably do a fairly good job.
What we do in producing English sentences is, partly: We have a thought in
mind, and we seek
the closest word/phrase (i.e. the closest "culturally common idea") to
express it. Similarly,
an AI can take a piece of its insular knowledge, and seek the closest
"culturally common idea"
to express it.
I once invented an elaborate scheme to permit this, involving a network of
"culture servers"
that maintain the "average knowledge state" of all the AI's in the world --
a kind of dynamic
"language" for inter-AI communication. So when one AI wanted to communicate
with another, it would
a) choose a hunk of its own brain to send, b) use the culture server to
translate it into
communally comprehensible language, as closely as possible.
Well, we haven't implemented this yet -- it's still a couple years off --
but I still think the
idea is good
The main point is: language allows us to export knowledge without grounding;
and, among AI's,
language can be more dynamic than among humans, because of the different
technological platform
> The same thing goes for selling sensory modalities for source code or
> financial accounting packages or commodities trading. Hey, Ben,
> aren't you
> glad I didn't patent that business model back in 1998?
You couldn't have; I patented it in 1997 ;D
ben
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:35 MDT