Saturday, October 23, 2010

Why are there "so few" physics majors?

Administrators tell us that we have too few physics majors (even when we exceed the national averages).  Why are there fewer physics majors than one might want?

1. First, physics is hard.  Not everyone is capable of doing physics.  And even if you have the ability it takes many years of training to become competent at physics.

2. Second, physicists are poorly paid.  Considering how important physics (and the other sciences are) for national development physicists are poorly paid.  Two of my best friends, one a physics major as an undergraduate and one in engineering, changed to careers in law and medicine, physics and engineering just didn't pay them well enough.  Some of the best physics students just won't put up with it.

3. Third, physicists are poorly treated.  We are treated as hired hands by a management class that has poor values and bad judgement. The old term "wage slavery" is all too true.

Friday, October 22, 2010

Data Mining with Asa H and vector utility

In a recent post (Data mining with Asa H) I assumed that Asa H was using a classical scalar utility.  In my plasma confinement example U = TNt, the Lawson triple product.  T is the plasma temperature, N is the plasma density, and t is the energy confinement time.

But a better model of plasma confinement is the two component vector utility U = (T, Nt). Vector utility was included in my original paper (http://www.robert-w-jones.com , inventor, Asa artificial intelligence) and this is an example where it is useful. (Of course for a DT plasma one could use the energy gain, Q, as the scalar utility rather than TNt.)

When fusion scientists use the triple product they are being value monists, when they use the Lawson conditions they are being value pluralists.

A set of Lawsonlike conditions is also a good way to describe ion source efficiency.

Wednesday, October 20, 2010

Is "goodness" a vector?

Utility theory in economics is a contribution to formalizing a notion of "good."  But we now know that the traditional scalar theory of utility is an approximation (Beardon, et al, J. of Mathematical Economics, 37, 17-38, 2002) and that utility must, in general, be a vector (Thrall, Decision Processes, 1960, Wiley, NY).  Similarly, any other formalization of good-evil would be a vector quantity.

It has often been a source of confusion to observe that people can be both good AND evil.  This is now explained in terms of the various vector components, one of which might be quite "positive" (strongly good) while another component might be quite "negative" (strongly bad/evil).

Tuesday, October 19, 2010

Mechanical Life

I wish to argue that web-operated autonomous robots (www.robert-w-jones.com, cognitive scientist, cognitive architectures.) are an artificial life form.  They: 1. use energy, 2. respond to stimuli from their external environment (via touch sensors, webcams, etc.), 3. can grow, repair, and reproduce by file transfer/copy into new processors, bots, etc. 4. are organized and incorporate learning/adaption, 5. their software is a "genetic" blueprint that is downloaded/uploaded to new/additional hardware, 6. they exhibit feedback/homeostasis, 7. they abandon worn out/obsolete  hardware, 8. they can act on their environment.   By most definitions Asa H would be considered alive.

I believe that the invention of mechanical life is probably more important than the biochemical synthesis of "artificial life" will be. (Mechanical life may prove to be more important than biological life.)

Friday, October 15, 2010

Cross-field transport through plasma bridges/spokes

Parker, et al (Applied Physics Letters, vol. 97, 091501, 2010) suggest that electron transport across a
magnetic field in plasmas may be caused by "spokes." This looks to be the same as my Trans. of the Kansas
Academy of Sci., vol. 92, # 3/4, pg 176, 1989, "plasma bridges."

Will Artificial Intelligences be Immortal?

Futurists often argue that artificial intelligences will be effectively immortal, able to file copy themselves into replacement machines as the older ones wear out.  Similarly, if the brain replacement argument (Clark Glymour, Hans Moravec, and others) allows us to upload human intelligences into machines they suggest we too could become immortal (Mind Children, Harvard U. Press, 1988).

But in order to reduce memory size and speed up search an AI needs to delete older, less useful memories in
order to make room for newer more useful ones.  Over a long period of time you simply wouldn't be the same
"person" anymore, even if you inhabited the "same" body.

It would be good for humans to live longer.  We spend too much of our lives preparing for our careers
(typically 22-28 years) and too long in retirement (15-20 years or so).  But the only sort of immortality that
seems possible might be recurrence (Recurrences and Discrete Dynamical Systems, Gumowski and Mira,
Springer, 1980). (One would want recurrence of that portion of the world that is "you" but NOT recurrence of the rest of reality. You don't want to live the same life over again.)

Tuesday, October 12, 2010

Grading

I believe that grading of students is more useful for motivation than for assessment.
I believe that grading makes students work a bit harder and learn a bit more than
they otherwise would.

Friday, October 8, 2010

Equations must be compatible with their boundary conditions

A system described by the second order linear and homogeneous differential equation:

y" + p(x) y` + q(x) y = r(x)

is not compatible with just any reasonable boundary conditions, for instance the linear and inhomogeneous
boundary conditions at the end points of the interval [0,1]:

a1 y(0) + a2 y`(0) = c

and

b1 y(1) + b2 y`(1) = d

Equations and boundary conditions must not only be reasonable models independently they must also be compatible with one another.

Thursday, October 7, 2010

Classroom Learning

I think people overestimate what you can learn in a single course.  Let me tell you how I learned mechanics and electricity and magnetism (and perhaps some other physics subfields as well).  I learned a bit in junior high physical science and my high school physics class.  I was taught the same things again in a first year undergraduate physics course and then in undergraduate mechanics and E&M classes.  (Electronics too.) I learned a bit more in graduate school in several mechanics and E&M classes.  I learned the rest when I taught first year physics, E&M, mechanics, and several electronics courses.  Only with ALL of this did I really know some physics.  No one course would possibly have done the job.

People expect too much of a single course. After all, neural networks typically require many observations of a given pattern before they learn it and learning is, in general, NP hard.

Wednesday, October 6, 2010

Weapons of Mass Destruction

The United States government and legal system have completely distorted the meaning of the term "weapon of mass destruction."  The proper military definition of a weapon of mass destruction can only be a weapon which is at least capable, in principle, of killing more than a million people at a single shot.  This would include most nuclear weapons, some biological agents, and practically no chemical agents.  A WMD is a strategic weapon, not one that can bring down a building.

Tuesday, October 5, 2010

The Scientific Notebook, Files, and Blogs

Back in the day we all kept notebooks.  An idealization of what these should be like can be found as chapter 6 in Writing the Laboratory Notebook by Howard Kanare, American Chemical Society, 1985.  I use this in my Advanced Lab course in physics here at ESU.  While I still have some notebooks of my own from as recently as 2001 I have since gone over to keeping files instead.  The files for my philosophical work look something like Working Notes, pg 136 of the Theory of Difference, Douglas Donkel, Ed., SUNY Press, 2001.  Perhaps a blog could become the modern incarnation of "Working Notes."  It would have the advantage (disadvantage?) of being available to everyone at once.
The disadvantage of this speed (and all things on the web) is that they are more likely to have been half baked.

Is Object Oriented Programming a Mistake?

I just read Mordechai Ben-Ari's criticism of OOP in the September issue of the Communications of the ACM
(vol. 53, No 9, pg 32, 2010).  My own criticism was: Is OOP a mistake?, www.comp.lang.C++, 26 May 2000 and www.comp.object, 22 Feb. 2001. (Note I am not criticizing MODULARITY in general.)  Gabriel has also offered: Objects Have Failed (www.dreamsongs.com/Files/ObjectsHaveFailed.pdf).

Friday, October 1, 2010

Data Mining with Asa H

The Asa H (Hierarchical Autonomous Software Agent) AI (Trans. Kansas Acad. Sci., vol 109, No 3/4, pg 159, 2006 and www.robert-w-jones.com, computer scientist, artificial intelligences) can be used for data mining. Data is input as a vector and corresponding utility (V1, U1), (V2, U2), (V3, U3).....

For a plasma confinement device V might be:

V1= (B, Nn, P, Ne, Te, t,...)

where B is the magnetic field strength, Nn is the neutral fill gas density, P is the input power, Ne is the plasma density, Te is the (electron) plasma temperature, and t is the confinement time.

The corresponding utility U might be:

U1= t Te Ne

the Lawson product.

When a set of V,U "pairs" are input to Asa H its extrapolation algorithms will suggest possible parameter changes that may improve U.

One of the extrapolation algorithms Asa H uses is:

V3 = V2 +(V2 - V1)c

where c <= 1 and U2 > U1.

Another application might be to forecast (and then avoid) arcing in an ion source.