Thursday, November 17, 2016

Beta > 1 plasmas

I have suggested that fusion researchers should put more effort into the study of beta > 1 or wall confined plasmas. Magneto-Inertial Fusion and the Magnetized Target Fusion projects at Los Alamos National Laboratory are recent examples of this.  Unfortunately, theoretical studies of such systems may be employing overly optimistic models of the magnetic thermal insulation. One might well expect such systems to have stochastic field lines.  If that is the case then we might want to employ turbulent thermal insulation as suggested in my papers: Current Science, pg 991, 1988 and Bull. Am. Phys. Soc., Nov. 4, 2009 (51st Annual Meeting of the APS Division of Plasma Physics).

Wednesday, November 16, 2016

Whiskers

A Spectra Symbol Flex Sensor (part FS-L-0112-103-ST) makes a good whisker for A.s.a. H.  The whiskers can be stiffened and strengthened by inserting the sensor into a long thin spring.

Saturday, November 12, 2016

Moral calculus and values

In blogs like those of 21 Sept. 2010 and 26 Oct. 2016 I have discussed value/moral systems for A.s.a. And other artificial intelligences.  Jonathan Haidt's moral foundations theory (see his book The Righteous Mind) suggests that humans have a set of primary values:

1. Care versus harm
2. Liberty versus slavery
3. Fair versus cheat
4. Loyal versus betrayal
5. Authority versus submissive
6. Sanctity versus degrading

Haidt believes political liberals value the first 3 more while political conservatives value the last 3 more, thus explaining how some people can vote for someone like Trump. They experience somewhat different realities. (See my blog of 21 July 2016) This set of values overlaps only slightly with those found in my 21 Sept. 2010 blog.  Perhaps I should try to implement more of them? Or should we be less concerned with what humans value (sticking to scientism)?

Thursday, November 10, 2016

Sense of smell

To give A.s.a. H. more of a sense of smell I have bought a set of MQ gas detector modules (MQ-2, 3, 4, 5, 6, 7, 8, 9, and 135) to add to A.s.a.'s conventional smoke detector.

Wednesday, November 9, 2016

In a mirror, darkly

There were several Star Trek episodes in which members of the Enterprise crew are trapped in an evil mirror universe.  For some time now I have felt that I was stuck as a slave in that Terran empire. Today I can't keep this feeling private any longer.

The political party that gave us a crook and a moron has now given us a mad man. And the people that elected a crook and a moron have now elected a mad man. Only a fool gives a mad man control of nuclear weapons. Human values aren't what they should be.

All species become extinct.  Is it now our turn?  Just how many mad men can the world survive? "The end of days" could be a self fulfilling prophecy. Are we about to see, first hand, the solution to the Fermi paradox?

I had hoped that humankind would survive long enough to establish a robust independent society of mechanical intelligences on earth.  I now think that that is less likely. Perhaps it will occur elsewhere or in some other Everett world.

The United States presents a substantial threat to the world as a whole and its survival.

Tuesday, November 8, 2016

Voting in Kansas 2016

At my polling place. Gone are the no guns signs.  In their place were multiple no cellphone signs.  I stand there wondering if I can do concealed carry of a new sort. At least my papers were in order, photo ID, proof of voter registration card.

Friday, November 4, 2016

Surprise

Some people believe that an AI needs to experience emotions. Surprise could be implemented in Asa H as a low similarity measure seen at time t after having n time steps during which the similarity measure had been high. Surprise might then trigger attention, bringing more sensors to bear, increasing the depth of case base search, reducing time spent on extrapolation and other learning, simulation, and planning, etc.

Thursday, November 3, 2016

Multiple exemplars for concept formation

Embodiment in general and embodiment in Lego robots in particular only give Asa one, or perhaps a few exemplars of any given concept.  But an intelligence will want multiple exemplars from which to generalize, enrich, and flesh out detail. The work presented in my blog of 1 Oct 2015 may not be enough.

Tuesday, November 1, 2016

Innate concepts again

Which, if any concepts should be "hand coded" into Asa H's concept hierarchy? I.e., "Core concepts."  Susan Carey's work might suggest: object, agent/agency, number/magnitude, cause, and language as possibilities worth considering. (The Origin of Concepts, OUP, 2009)

a simple Quinian bootstrapping in Asa H

In A.s.a. H. a concept's name is associated with whatever sequence of features happens to have been active at the moment the name is learned, for example: collision=(sense near, contact, decelerate, hear word "collision"). If this same name is heard again, but under somewhat different conditions, then the concept may evolve/change a bit. (See The Origin of Concepts, Susan Carey, OUP, 2009 for a discussion of a much more elaborated Quinian bootstrapping.)