Hi Gil,
Evolutionary programming. It's a whole nother topic. It has some currently popular implementations in Genetic Algorithms, and Genetic Programming. GAs were designed, as far back as the 50s, to solve problems. An early implementation served in a learning checkers game. Today, GAs find their way into all sorts of programs, including a stock market trading program I built (see http://www.thinkalong.com ). GP writes algorithms, of sorts, that are also used to solve problems. When I use the term Evolutionary Programming (EP), I am most interested in the way Doug Lenat used it in AM and Eurisko. I built aspects of it into my program TSC. The idea is to implement an engine for evolving knowledge (whatever that is).
I suppose I didn't explain EP, though, did I? Here goes. Borrowing from Chuck Darwin, we are led to believe that some algorithm that uses survival of the fittest is provides a mechanism by which things *learn* to occupy various niches *out there*. Who defines the fitness function? What does a fitness landscape look like? It is clear that when the frog invented a sticky tongue, life as the common fly knew it changed forever; It's fitness landscape got a major jolt. Only those who had faster twitch responses survived to propagate their kinds. Faster twitch reponse (jumping quicker) meant the fitness landscape of the frog was drifting as well. Mother nature, herself, was at work in that fitness landscape.
So, we write a program that messes around with the *dna* of concepts (things like attributes). We use a random number generator to make selections of which attributes to mutate (or combine with some other concept), then we propose some test (environment) and try to see if we can *break* our new concept. Eurisko, for instance, would take, say, a heuristic rule (a concept) (e.g. a concept is interesting if its average example (defined elsewhere) is interesting). Mutate that heuristic to read something like a concept is interesting if its extreme example is interesting. Yup, it's a lot of *wrist action* and arm waving, but it works.
While speaking of knowledge, I spent my vacation reading Lakoff&Johnson's latest book _Philosophy in the Flesh_ which is all about *embodied knowledge*, and, of course, metaphorical thinking. They make a very convincing case that knowledge must be embodied--it must exist in human experience; that, of course, supports the case you make below. It strongly suggests that where we are really heading with an OHS/DKR is a well-groomed library, search engine, and collaborative tool. Topic Maps meet Microsoft Office, I suppose. Thus, given a rather convincing case by L&J, perhaps our use of Knowledge Representation really involves encodings of information that we turn into knowledge when needed. That, of course, leads directly to all this semiotics stuff of C.S. Peirce. He spent a lot of time thinking about how humans use signs and create meaning from them.
My work on the Topic Maps committee is all about giving users the ability to create views to suit their own needs. That's just a start on the matter. Now, for instance, Mozilla lets you make your browser look and feel like you would prefer it to look and feel. Tiny steps. An enormous fitness landscape we are traversing.
Cheers
Jack
From: Gil Regev
Hi Jack,
Sorry for responding so late. This would be a good thread to start but I must admit that I don't know what evolutionary programming means exactly. How about clarifying the term as a starting point?
On the subject of knowledge representation, I tend to believe that true knowledge only occurs in the presence of people so I have a difficulty admitting that a machine can represent knowledge. For the moment I still cling to the idea that machines can only represent information and data. For sure, a machine can optimize the way it represents the information so that it is more understandable to people. We can see that with all the people involved with the mapping of Web sites and databases. But when information comes from other people rather than from a machine, I think that it is better to enable the people involved to generate their own representation rather than to give full power to the machine to find the "correct" representation. What if we gave the power to Web site designers to create their own graphical representation (site map) of their site rather than trying to mechanically create such a representation after the site is already built?
Gil
-----Original Message-----
From: Jack Park [mailto:jackpark@verticalnet.com]
Sent: samedi, 8. juillet 2000 01:41
To: unrev-II@egroups.com
Subject: Re: [unrev-II] "Trees of Knowledge" Map vs. DKR Enables Knowledge Mapping?
Thanks, Gil.
Your comment about software dictating the knowledge representation. Howard Liu and I have been talking about the appropriateness of evolutionary programming to the construction of knowledge structures. Indeed, it would be worthwhile to start a thread on that topic here.
Jack
From: Gil Regev
Hi Jack,
There are a few other important points in the products Trivium have:
1. UMap takes a set of keywords you're searching for and gives you a graphical representation of the results, but more importantly, I think, it represents all the additional keywords it found in the pages containing the original keywords and the relationships between these keywords. This helps the users to expand their understanding of a subject matter by helping them find new concepts (keywords).
2. Gingo is about mapping peoples' skills. Levy's and Authier's book about knowledge trees was about fighting what the french call exclusion, the fact that the unemployed become excluded from society. Knowledge trees are supposed to give a sense of what it is they know and start looking at how to improve themselves by learning what they don't already know. The authors also describe how "knowledge" can become a currency that can be exchanged. You exchange "knowledge" with others in order to get recognition and possibly future "Knowledge". I am not sure that the term "Knowledge" is appropriate in this context if we define it as something that exists only in the context of a person. In this sense you can only exchange information but not knowledge.
3. There's also the concept of "patents" which are kind of exams that you create in order to test other peoples' knowledge. Interestingly, you get a "blason" when you conceive an exam. The Knowledge tree of an individual is the tree of "blasons" that the individual has accumulated. The knowledge tree of an organization is more complex and normally consists of the accumulation of the individuals' or departments' knowledge trees.
The thing I have a problem with with regard to both these products has to do with their fixed representation that is decided by the software rather than by the user. In some ways it defeats constructivism because it prevents users from creating their own representations. Otherwise, these are very very interesting products.
Gil
-----Original Message-----
From: Jack Park [mailto:jackpark@verticalnet.com]
Sent: vendredi, 30. juin 2000 17:21
To: unrev-II@egroups.com
Subject: Re: [unrev-II] "Trees of Knowledge" Map vs. DKR Enables Knowledge Mapping?
Wow! This thread is getting longish, and fun. TriVium appears to understand
the issues. There is more on the Trees of Knowledge at
http://www.connected.org/learn/levy.html
What's important, at least to me, is that they center a presentation on a
single focal point. By way of analogy (which, itself is an absolutely
necessary feature in any knowledge engine), Doug Lenat's Eurisko program
used a concept he called *focus of attention* (he wasn't alone -- other AI
jockeys used that term) as a means of computing the *priority* placed on
some task on an agenda.
Now, we're getting to the meat, IMHO, of the matter. We only learn when we
are interested and we tend to agendize things that are interesting to us,
leaving the 'c' jobs for later -- perhaps except for those few who are
compulsive about getting 'c' jobs out of the way. Eurisko's architecture
applied an agenda and task structure that kept Eurisko working on those
tasks with highest priority. Each task cycle saw a slight decay in priority
of all tasks (forgetting), and each task, while being executed, had the
ability to modify the priority of any task still on the agenda (feedback).
Thus, it became possible to use *focus of attention* as a means to keep some
train of thought running for a long time (greatly to the probabilistic
detriment of those not running). That, of course, explains <gg> why the
excuse "I forgot" is valid.
I am saying here that an agenda-based architecture, one with feedback and
decay mechanisms, comes closest of anything I have seen yet to a
biologically inspired architecture. My program The Scholar's Companion
implements just such an architecture <note> not bragging here, just stating
that I have some experience with this approach </note>. All of which is to
suggest that it would be really nice to see an English version of the books
mentioned at TriVium (else I'll have to dust off my 40-year old high school
French ;-(
Merge a couple of threads together here and we're liable to have an image of
the DKR.
Cheers,
Jack
From: John J. Deneen <JJDeneen@ricochet.net>
> So the following is some interesting info from TriVium.com, relative to
our design
> requirments for the DKR to enable knowledge mapping:
>
> On 6/15/00 at SRI, after Gil Regev demonstrated his collaborative concept
mapping
> applet (http://icapc4.epfl.ch/knowarepub, http://icapc4.epfl.ch/g99space),
he
> suggested checking out the "The Trees of Knowledge" technology (i.e.,
Umap, Seek-K,
> and Gingo) at TriVium.com based on my comments about Cartia.com Relational
Topic
> Mapping (RTM) technology called "ThemeScape."
> (http://www.cartia.com/products/index.html)
>
> Proprietary Technology
> http://www.trivium.fr/new/techno.htm
>
> ...."For information, type "Trees of Knowledge" or "Gingo" (the first
real-time
> visualization software for organizational information and competencies) in
the
> search engine of your choice. You will then have access to interviews,
case studies,
> and the opinions of various members of the Internet community
(journalists, leaders,
> citizens, critics, etc.) on the subject. This information is constantly
updated,
> given the increasing power of this theory, and its practical applications
to daily
> life. Reconnect often to remain up-to-date on the latest developments. For
a global
> vision of different opinions, why not make a Umap map of all the
information?"...
>
> The map
> The Umap map is a mosaic of colored pieces on a uniform background. Each
of these
> pieces represents a thesaurus word; each word of the thesaurus finds its
place in
> the map.
>
> What is the map?
> Placed between the windows of the thesaurus and the document group (or
body), the
> Umap map indicates the relative proximity of thesarus words, beginning
with their
> relative significance in each text.
>
> Significance of the map
> By grouping the words in proximity zones (small islands, near-islands,
concentric
> layers, etc.), the map offers an intuitive approach to someone familiar
with the
> logical topic connections that exist between certain texts.
>
> What use is it
> By spotting common topics within certain texts, one can quickly select the
texts of
> interest, or remove those that are momentarily unneeded. ....
>
> Why Cartography ?
> http://www.trivium.fr/new/carto.htm
>
> Concept & Ideas
> What is exactly Knowledge Management and why do companies need KM
solutions today ?
> http://www.trivium.fr/new/index_2.htm
> http://www.trivium.fr/new/gingo/main.htm
>
> Rod Welch wrote:
>
> > Bill,
> >
> > Sorry have not been able to respond sooner.
> >
> > The aim of a knowledge management effort, is to map a share of important
> > connections showing cause and effect that the mind forms when it
encounters
> > information during a meeting, looking at a picture, reading a book,
walking
> > across the street, i.e., input from sight and sound that constitutes
human
> > experience.
> >
> > This uses writing in a different way, to set out our personal
understanding of
> > the why and wherefore of events. In particular we want to identify our
mistakes
> > by checking alignment, and make corrections in small communication
miscues
> > before they become big problems. The DKR rigged in a certain way can
hardwire
> > the relationships so we can get them back when needed to improve upon
> > spontaneous impressions. The DKR provides context that leverages the
value of
> > information. It enhances research by providing a routine template of
structure
> > that extends traditional punctuation and rules of grammar for imparting
meaning
> > to information.
> >
> > It is not all a bed of roses. When we begin making connections, the
result
> > looks confusing to some. Jack and Doug are working on tools to improve
the
> > view.
> >
> > There is a lot more that can be accomplished with an engine of knowledge
to
> > enhance traditional work practices, e.g., reporting, engineering, law,
medical
> > practice, accounting, scientific research, etc. In short, the DKR
provides an
> > environment and tools (OHS) for getting a share of our knowledge into a
form
> > that allows it to be tested for accuracy, and applied consistently,
promptly
> > when and if needed. The big distinction between this idea and the
popular
> > notion of storing "knowledge" from books, magazines and so on in a
repository,
> > is that each of us have a lot of knowledge from our daily experience
that we
> > primarily rely upon to do our work and live our lives. This latter body
of
> > stuff is what we want to improve, and in doing so, the formal stuff in
books
> > will get better also.
> >
> > Hope this helps.
> >
> > Rod
> >
> > Bill Bearden wrote:
> > >
> > > Rod,
> > >
> > > You bring up an interesting and valid point with which I am currently
> > > struggling. I have been reading (and trying to understand) some of
> > > Malhotra's extensions of Churchman, esp.
> > > http://www.brint.com/members/online/200603/kmhitech/kmhitech.html.
There,
> > > Malhotra quotes Churchman:
> > >
> > > "To conceive of knowledge as a collection of information seems to rob
the
> > > concept of all of its life... Knowledge resides in the user and not
in the
> > > collection. It is how the user reacts to a collection of information
that
> > > matters."
> > >
> > > This sounds very much like what you say.
> > >
> > > But if knowledge can not exist outside of the mind, how can a DKR be
> > > possible? By this definition, neither book nor computer can contain
> > > knowledge. I believe in the concept of the DKR. Therefore, I can not
accept
> > > a definition which fundamentally prevents its existence.
> > >
> > > So, with your definition, my previous comment about knowledge being
> > > everywhere is not valid. But I would guess that your definition
invalidates
> > > lots of things that have been discussed.
> > >
> > > Bill
> > >
> > > > -----Original Message-----
> > > > From: Rod Welch [mailto:rowelch@attglobal.net]
> > > > Sent: Saturday, June 24, 2000 6:57 AM
> > > > To: unrev-II@egroups.com
> > > > Subject: [unrev-II] 2020 Hindsight: A Fictional DKR Narrative (long
> > > > (sorry))
> > > >
> > > >
> > > > Bill,
> > > >
> > > > Just, on your comment that "knowledge is generated all the time.
> > > > It is all
> > > > around us in books, etc..."
> > > >
> > > > My sense is a little different.
> > > >
> > > > "Knowledge" resides in the minds of people, and so is constantly
> > > > being formed
> > > > out of the information that is all around us in books, TV,
> > > > meetings, and so on,
> > > > as an interplay between our experience, and the mental ability to
form
> > > > consistent pattersn connections or patterns of cause and effect.
> > > >
> > > > Rod
> > > >
> > > >
> > > >
> > > > Bill Bearden wrote:
> > > > >
> > > > > Rod,
> > > > >
> > > > > >
> > > > > > Bill,
> > > > > >
> > > > > > Very thoughtful illustration you set out in your letter
> > > > today... <SNIP />
> > > > >
> > > > > Thanks.
> > > > >
> > > > > <SNIP />
> > > > >
> > > > > > ... your open source query
> > > > > > might be aided
> > > > > > by explaining how that approach saves time, improves
productivity, and
> > > > > > earnings. Those criteria have proven to be good generic
starting
> > > > > > points for
> > > > > > evaluating tools and work methods. ...
> > > > > <MORE-SNIP />
> > > > >
> > > > > I agree that the traditional "value" metrics are useful.
> > > > However, I question
> > > > > how well they apply to something truly new. I doubt that
> > > > electric lights or
> > > > > telephones were cost effective replacements for existing
> > > > technology right
> > > > > when they were introduced. If DKRs ever prove truly useful, it
> > > > may only be
> > > > > after there are lots and lots of them hooked together and people
are
> > > > > immersed in them as a normal part of their lives. That is a
> > > > long ways off.
> > > > >
> > > > > And anyway, my text was an exercise in speculation as much as
> > > > anything. It
> > > > > was just me trying to describe part of a system I see in my
> > > > head (if I close
> > > > > my eyes real tight after I've had a couple of beers :-).
> > > > >
> > > > > <SNIP-SNIP-SNIP />
> > > > >
> > > > > > In the meantime, it turns out that using a keyboard, computer
> > > > screen, and
> > > > > > special tools seems to augment human intelligence beyond what
can
> > > > > > be expected
> > > > > > from reliance on voice recognition and pictures, for reasons
> > > > in the record
> > > > > > reviewing Andy Grove's book on 980307...
> > > > > >
> > > > > >
http://www.welchco.com/sd/08/00101/02/98/03/07/161449.HTM#L351552
> > > > > >
> > > > > > Generating knowledge is hard work. ...
> > > > >
> > > > > Yes, but knowledge is generated all the time. It is all around
> > > > us. Capturing
> > > > > it and encoding it so computers can store it and people can
> > > > learn it is the
> > > > > problem as I see it. IMO, this process will remain very
> > > > difficult until we
> > > > > have more immersive user interfaces. Until then, it *might*
> > > > make sense to
> > > > > try and "scrape" knowledge from existing stores (e.g. books,
databases,
> > > > > source code, etc). Truly integrated information systems
> > > > probably yields more
> > > > > short term bang for the buck. Again, just my speculation.
> > > > >
> > > > > > ... But people don't mind hard work, if it
> > > > > > yields rewards and is fun. Games are an example. People "work"
> > > > > > awfully hard at
> > > > > > golf, tennis, running, exercise, and computer games, because
they
> > > > > > get immediate
> > > > > > satisfaction of varying kinds from the experience.
> > > > > >
> > > > >
> > > > > I love the point you make about "work" and games. IMO,
Csikszentmihalyi
> > > > > explains fairly well why that is in his book, Flow. I was excited
to see
> > > > > Flow mentioned on the L3D philosophy page
> > > > > (http://www.cs.colorado.edu/~l3d/philosophy.html). L3D
> > > > (LifeLong Learning &
> > > > > Design) is the "mother" project of Dynasites, to which John
> > > > Deneen submitted
> > > > > a link yesterday.
> > > > >
> > > > > <FINAL-SNIP />
Community email addresses:
Post message: unrev-II@onelist.com
Subscribe: unrev-II-subscribe@onelist.com
Unsubscribe: unrev-II-unsubscribe@onelist.com
List owner: unrev-II-owner@onelist.com
Shortcut URL to this page:
http://www.onelist.com/community/unrev-II
Community email addresses:
Post message: unrev-II@onelist.com
Subscribe: unrev-II-subscribe@onelist.com
Unsubscribe: unrev-II-unsubscribe@onelist.com
List owner: unrev-II-owner@onelist.com
Shortcut URL to this page:
http://www.onelist.com/community/unrev-II
Community email addresses:
Post message: unrev-II@onelist.com
Subscribe: unrev-II-subscribe@onelist.com
Unsubscribe: unrev-II-unsubscribe@onelist.com
List owner: unrev-II-owner@onelist.com
Shortcut URL to this page:
http://www.onelist.com/community/unrev-II
------------------------------------------------------------------------------
------------------------------------------------------------------------------
Community email addresses:
Post message: unrev-II@onelist.com
Subscribe: unrev-II-subscribe@onelist.com
Unsubscribe: unrev-II-unsubscribe@onelist.com
List owner: unrev-II-owner@onelist.com
Shortcut URL to this page:
http://www.onelist.com/community/unrev-II
This archive was generated by hypermail 2b29 : Sun Jul 23 2000 - 15:50:53 PDT