Thursday, May 19, 2005
The next generation knowledge tools
Two communications from Dr Richard Ballard, Founder of
Knowledge Foundations, with foot notes by
Paul Prueitt, and links to additional comments that might be made by BCNGroup
members
Second communication à [80]
Paul & Friends:
Paul knows well the way to lure me away from Mark 3 programming. That is the chance to read and comment on John Sowa's work, which is always eminently scholarly and well tested in practice.
See
The Challenge of Knowledge Soap (Sowa, 2004)
http://www.jfsowa.com/pubs/challenge.pdf
John has been one of the most important of my teachers. Still at times, he thinks that I have taken quite a different road than he might have wished.
The critical difference between us was that he was a philosopher - logician and I was an experimental physicist on the way to finding a proper philosophy for natural law. It was absolutely vital to him that logic finish its historical quest to master natural languages and mine was to find a most fundamental representation for capturing every natural law, including physics -- but also economics, aesthetics, politics, ethics, etc. [1]
The dividing line between us came historically before I was born via relativity and quantum mechanics. Before that the mathematics of a continuous variable and Newton's mechanics made logic and mathematics appear to be the Queen of Sciences. But by the early 1930s American positivist philosophers had written logic off as a philosophical mistake. Logic and mathematics were created to describe a fixed, deterministic, and self-consistent world, but the laws of physics had established by the 1930s that reality would always be uncertain. No truth or falsity, the answer to every real question was "maybe". [2]
For humans what is meant is that within such uncertainty, there were no predestined deterministic paths possible, our own actions could change history decisively [3].
The key to that power is in using our experience to create modestly predictive theories of how the universe truly works in every particular instance.
Through this articulation we learn thereby how best to achieve any goal we sought.
Meaning (semantics) is attached to the capacity to provide successful prediction (theory), not to the lesser capacity of self-consistent feature description (logical grammar -- linguistics).
Note by Prueitt: this choice is one
discussed by Tim Berners-Lee and references on page two of the Preface. It is the key differentiator between the first school and the second school.
Nature was not self-consistent and consistentency was never a successful human survival or business strategy. [4]
Logic would allow us to build machines that did our bidding in a provably self-consistent, but unthinking, un-learning, repetitive way. But human thinking and learning might allow us to use our most successful theories of every kind to achieve ever more powerful and successful survival goals.
The signature of intelligence is the capacity to use lessons-learned to change behavior, with no need for significant change to the machinery or functionalities of the biological or mechanical host.
Here forward we need machines to start thinking and learning like us or in our place.
Note by Prueitt: It is here that Richard and
I are at deep odds over religious statements.
It is my feeling that this statement is unnecessary to anything that
Richard wants to do, and is in fact inconsistent with what one expects to find
in the future. The second school
recognizes that some of the founders are at odds over this type of
statement. Ben Goertzel, for example,
feels that the issue of more complex than the discussion allows us to recognize.
The differences between John and I has now several names. John uses the notion of logical algorithms as one pole and Knowledge soup as the catch all for everything else still in a muddle from his perspective.
The terms I favor has been the linguistic - semantic gap.
Paul has offered the notion of two schools and this is fine with me
Paul, Mills, and Dennis know more than most about what this work means. Andy and others can learn as much as they might like, that is why I will be coming to Washington prepared to deal with any question and work with any in Paul's circle.
For those in R&D mode, the challenge question is "how can your technology learn and change its own behavior with or without human intervention." Most stop at some process of pattern recognition and call this pattern recognition “cognition”, arguably every living thing with or without a brain can do that. [5]
The adaptive discrimination problem was what we were doing between 1985-1990 Star Wars efforts (Space-based Survellience and Tracking SSTS($2.5 million) and BattleManagement Requirements Analysis Testbed $5.6 million. We could learn and teach totally new target discrimination signature patterns after only 40 space-based missile intercepts using statistical and genetic algorithms. That process starts with an algorithmic theory and learns different data weights (information).
But how can an algorithm that thinks itself provably correct learn its correctness is meaningless or flawed and change its own underlying theory and assumptions all by itself -- learn programming?. It needs self-awareness of its own successful and unsuccessful theory use.
Note by Prueitt: Ah, does it really? I think that this is where the “religious
belief” that Richard expresses about living computers is grounded. The work on “perceptional measurement” that
I cite, and which Peter Kugler is the leading expert on, pushes this
reinforcement theory back onto the inputs to the program. If these inputs can be in a situational
setting and from a learned human with very little dysfunctionality in the
software interface, then we have the second school notion of HIP (Human-centric
Information Production).
Let me assert logic can never do this. Knowing that may be intuitive, as it for Paul and others. Like Paul, some may know there is a theory of Knowledge, but like him have yet to grasp what it all means.
Few are prepared yet to prove formally that Knowledge = Theory + Information, but I can.
Theory to KFI is not simply logic, it offers quantitative measures suited to describing absolutely any theory that might influence or change human or machine behavior. That is new science -- it gives me my predictive physics, but also provides the same for economics, aesthetics, politics, and even fantasy, imagination, myth, fear, and hate -- take your pick.
First of all, the Mark 3 is not an aggregate system of systems or object oriented. [6]
(short section moved to a separate bead [79-1])
The same coding system works for ontologies of 100s of millions of concept models -- the estimated size for libraries containing all written knowledge. These technologies scale linearly (proportional to information content [Shannon limits]) with anything and everything changing at will. Every part can learn and change at the bit level. It was this requirement that led us to require complete independence from any predefined word sizing.
What would it mean if all human knowledge was in the wristwatch I got for my first or fifth birthday and I could operate with legally defensible levels of professional practice in a few months or years of certified training. Locating anything I wanted without search. Semantic webs contain only one copy of every idea and its unique coded identifier locates it instantly with no search.
Note by Prueitt: The uniqueness of concepts has been a key element of Tim Berners-Lee’s
notion of URI use within the Semantic Web concept that he has discussed. The concept is very hard to get set right, because
often the locating of unique identifies for concepts seem to violate the notion
of response degeneracy, and this notion is not one that most are familiar with. We are however aware of the need for
flexibility and agility. My concept of a tri-level
architecture addresses this issue in a way that few have made
comment on.
Paul's and my second school is all about this kind of science-lead capability, not continuous technology evolution by incidental feature improvements. [7]
I can bring you copies of all of this, if you want to consider teaming. Our interest is technology transfer to you. You keep your customer -- and gain a new business line. That is the deal that has let us pick and choose the projects we like for 20 years.
Write and give me a heads-up on what you all might like to talk about.
Dick
[1] It is not always possible to completely represent the other side of a life long discussion.
[2] This history is one that is common to how John Sowa, Richard Ballard and I and others see the history of science. My own representation of this history is communicated in a chapter called “Is Computation something new?” and my representations about the possible answers to the question:
“mathematics is to physical engineering as “x” is to social science”
what is x?
is given in the Preface to Foundations to the Knowledge Sciences:
http://www.bcngroup.org/area3/pprueitt/book.htm
[3] I quote research literatures in cognitive and quantum neuroscience as to the notions of “response degeneracy” and tipping points, where a symmetry induction in the physical phenomenon causes a delay in what would other wise by a conservation of thermodynamical laws. Richard is approaching the same phenomenon, and he points out as I do and as John does, that this response degeneracy is not handled well by classical logics and the primary traditions of artificial intelligence. Peter Krieg is one of the founding members of the BCNGroup and represents a similar computational capability as the feature that Richard points out here about the Mark 3. This work by Krieg is called poly logics and is available in a commercial system from www.pilesys.com. SchemaLogic’s SchemaServer also provides some degree of control over the double articulation found at these tipping points.
[4] For course this statement is rejected in principle if one is a fundamentalist of any type; religious, economic or scientific. Fundamentalism is the belief that one set of facts and rules are at core truthful, and those facts or rule that are inconsistent with the chosen set is not true.
[5] Of course, the computer program is not completely a thing in the sense that it is a construction for a set of abstractions. Things that are physical are not abstractions, and are seated in physical reality in what I have called the “pragmatic axis” (section 3, Preface: http://www.bcngroup.org/area3/pprueitt/book.htm)
[6] Richard is comparing and contrasting his system with the proposed Roadmap system or the Knowledge Sharing Foundation system.
[7] Yes, we agree