Friday, January 29, 2016

Solar power

I've bought several solar panels sufficient to power one or two Lego NXT robots.  This will give Asa H an example of an alternative source/kind of "food."  It may also help Asa to understand something of how plants operate.

Thursday, January 28, 2016

Doing fundamental science

At the foundation of any science is the set of primitive concepts that that science employs. When I examine the concepts that Asa H is creating/discovering I am looking for new, interesting ones.  This is one way of exploring the most basic foundational issues.

The few unique concepts Asa has produced so far have not been very profound.

Tuesday, January 26, 2016

The self

I have been watching as my artificial intelligence Asa H develops a model of its self.  (See, for example, my blogs of  4 March, 28 April, and 5 Dec. 2015)  But Thomas Metzinger and others have argued that there is no self. (See, for instance, Being No One, MIT Press, 2003) 

As an advocate of scientific pluralism I believe that one needs multiple models of reality. One need not talk only in terms of quantum fields, there will be times one will wish to talk in terms of atoms and molecules.  The idea of atoms may not be our most fundamental concept but it may still be useful.  Similarly, the concept of a self may remain useful too.  The self concept might be present in some models of reality but not in others.

Mathematics and the study of mind

Ben Goertzel argues that a sufficiently complex spatiotemporal  "pattern network constitutes a mind" (The Hidden Pattern: A Patternist Philosophy of Mind, Brown Walker Press, 2006, pg 16). If this is true, and since mathematics is arguably a science of patterns, then we should use mathematics to study/describe mind. I offer this as a challenge to all the armchair philosophers of mind whose work is typically nonmathematical. 

One should note that theoretical computer science can be looked at as a part of mathematics and so I am just arguing for traditional computational intelligence as constituting work on the philosophy of mind. On the other hand my work on Asa H has been more mathematical than a lot of AI is.

Friday, January 22, 2016

Spreadsheet AI again

There are people who try to do anything and everything with a spreadsheet.  This was especially true 25 or 30 years ago. (I even designed an EXCEL spreadsheet version of Asa H light.)  I think it makes sense to create neural network programs in EXCEL.  There are a number of those available online and in texts. But I just recently ran across a Turing machine and a chatbot both programmed in EXCEL!

Thursday, January 21, 2016

Another new concept

I find that Asa H has formed another nonhuman concept/word, something like: "a task requiring 3 hands."  An example for humans might be soldering or certain parts assembly operations. For humans this might also be "a task requiring 2 people." 

Tuesday, January 19, 2016

Knowledge organization

Michael Lesk has suggested that having one single standard knowledge organization system would be ideal.  It seems to me, however, that knowledge should be organized in a number of different ways each one of which may be optimal for a certain kind of knowledge use or problem solving.  Glossaries, dictionaries, encyclopedias, taxonomies, thesauri, ontologies, abstracts, indices, semantic networks, etc. each have their various different uses. This is scientific pluralism again.  Each entity can be characterized in a number of different ways.

Sunday, January 10, 2016

Can we force an alternate conceptualization of reality?

I am interested in alternate models of the world. (See, for example, my blog of  22 April 2013)  Asa H was able to learn all of Wierzbicka's natural semantic metalanguage in a hierarchy about 5 layers deep. Hinton's deep backprop net learned the concept "cat" at about layer 6  and "human face" at  about layer 5.  A shallow backprop network might be trained on the same input and learn the same mapping function as the deep network did.  I suppose that the shallow network could possibly learn the very same concepts as the deep network if they were all distributed representations but this seems unlikely.  Teasing out the concepts that are generated in the net would not be easy of course.

Saturday, January 9, 2016

An experiment with Asa H tuning

Starting with a large value for TMAX in each level of the hierarchy (see, by way of example, the simplified code of my 10 Feb. 2011 blog) run for a time* and then reduce TMAX to the largest value of DURATION(Z) that has been observed plus a tiny bit.**

* how long?  seeing how many cases/patterns? 'til DURATION doesn't increase?
** how much? measured by a standard deviation for DURATION?

Compromising your values

How intelligent you are depends upon how good your value system is.  If you have bad values you make bad decisions and get fewer rewards.  A value system optimized for operation in the natural world will differ a bit from one optimized for operation in our imperfect human society.  One must employ a compromise between the two.

Thursday, January 7, 2016

My library

I have a research library of about 3000 books.  Since I don't have space for many more than this a year or two ago I began to try to weed out any books I could as I acquired newer ones.  I have also tried to add electronic books rather than paper ones.

Friday, January 1, 2016

Distributed representation of concepts

In keeping with scientific pluralism I have tried to use my artificial intelligence A.s.a. H. to create/discover alternative concepts with which to model reality. As Asa explores/experiences the world it typically learns more than the fundamental human concepts (e.g., my blog of 5 Nov. 2015 and things like Wierzbicka's semantic primes/natural semantic metalanguage).  But it is difficult to identify and characterize any original (non-human) concepts since Asa can form distributed representations as well as local ones. This also makes it difficult for us to understand what human concepts Asa has acquired. Hand coding and tuning of Asa's hierarchical memory is of some help with this.

Static electricity in bed

Rubbing your fingernails (back of your hand) quickly over silky (polyester?) bedding can produce flashes of white light.

Nonconceptual content and Asa H

Not all of one's mental contents have been exhausted when you have identified all of the concepts an individual possesses. (See, for instance, Essays on Nonconceptual Content, Y. H. Gunther, MIT Press, 2003) Experiments with Asa H help us explore both the conceptual and nonconceptual contents of mind (both human minds and artificial ones).

Reflexive reactions to bright light, striking of the shin, etc. are nonconceptual in humans and a robot may have these as well. Preprocessors may adjust visual (and other) sensitivities (sensor "autoscaling") for example.

In Asa H concepts are formed from input through vector clustering.  A concept is a cluster.
Clusters which have not been identified (associated) with a human word/label may be concepts for Asa H but may not correspond to any concept humans use to describe the world.

Any new input vector which isn't sufficiently similar to an existing concept will be stored in memory and may later be incorporated into the formation of a new cluster.  Prior to that time it might be considered to be a "protoconcept."

More abstract concepts reside higher up in the Asa H hierarchical memory network.  More abstract concepts require a deeper hierarchical network. Shallow networks might be able to learn the same content (input-output mapping function) that deeper networks learn but they would not be able to contain the same concepts.

Reusable spacecraft

The X-15 was reusable and economical, the space shuttle was reusable and uneconomical.  We'll see how the economics works out for spaceship two, new shepard, and falcon 9. It should be easier for the suborbital systems.