Friday, August 26, 2011

Research freedom

One of the things which is most important to me has always been the ability to decide the direction my research would take.  Working in an industrial research lab is then not a good match.  In industry the goal is financial reward and, in fact, financial reward in a very short time frame.  The free exchange of data and ideas is also limited by the desire to gain advantage over potential competitors.

Work at government ("national") labs is somewhat better but research objectives imposed "from above" are often irrational or based on poor values.  Congress seems to want to force NASA to build the SLS based on a "bigger is better" mythology (and to send contracts to congressmen's home districts). 20 ton science modules launched by Proton rockets were chosen for MIR even though the engineers wanted 7 ton modules (launched on the R-7) which could have been changed out frequently.

This leaves university laboratories as the best choice if you want to make your own decisions as to research goals and direction.  Even then you will find that many university departments want all staff to work in some one (of a few) directions that the department concentrates on.  To have anything approaching research freedom you must even choose your university employer carefully.  This reality placed many constraints on me.  The notion that "in america you are free" was simply nonsense.  These conditions drive some people out of science.

Many people have argued for the need for freedom in creative pursuits in general and science in particular.  What kinds of freedom are required and how much is needed is a matter of debate. "Various psychological studies do show that, in general, scientists .... are characterized by a need for autonomy..." DISCOVERING by Root-Bernstein, Harvard U. Press, 1989, pg 315

Wednesday, August 24, 2011

Bad Values

Congress wants to force NASA to build the SLS despite the existence of the Atlas V, Delta IV, and Falcon 9 rockets and at the same time it wants to kill the Webb space telescope despite the great successes of Hubble.  Science is exploration.  Sending people on sightseeing trips is not.

Friday, August 19, 2011

An AI at IBM

Modha, et al, report that IBM Research-Almaden is running a modular network of simulated phenomenological spiking neurons roughly equivalent in scale to a cat's brain or 4.5% of a human brain. (Comm. of the ACM, August 2011, vol. 54, # 9, pg 62) This "C2" simulator runs on an IBM Blue Gene supercomputing system but in order to reduce the power and space requirements in the future the plan is to employ newly developed neuromorphic chips.  Modha extrapolates to a human scale brain by about the year 2019.

I am not sure IBM will have the right functional brain modules identified by then or know how to wire them together.  Their work has been at a different level of abstraction. I see no provision for a value system for example.

Wednesday, August 17, 2011

Tax the rich

The news last night was that the wealthiest 20% of the population have 84 % of the wealth.  What they didn't say was how the rich get that way.  The two most common ways to get rich are crime and inheritance.  Anything involving a sharp mind or hard work is, at best, in a very distant fourth place. (Marriage is in third place.) And a great deal of luck is involved as well.

The republican party was bought and is owned by the rich. 

Friday, August 12, 2011

Solving the jobs shortage

When corporate profits and executive pay are at record highs and banks aren't lending like they should be the solution is simple; nationalize them (and hire and lend).

Tuesday, August 2, 2011

"Big science" (and technology) are very conservative

Because of the long lead time (designs must be finalized years or even decades in advance of use) and high price tags (multi billion dollars at risk) you won't see the very latest ideas in robotics in use on Mars nor the newest techniques in plasma science deployed on the ITER.  Rather, the newest work is confined to the university laboratories.

DeSTIN

Itamar Arel, et al's DeSTIN architecture is quite similar to my Asa H.  It complements my work on Asa H since it makes use of Bayesian statistics.