Monday, January 30, 2017

Alternative facts

It really is possible for there to be alternative facts. One specialist intelligent agent can learn the proposition P while a different specialist agent, operating in the same environment, can learn NOT(P). See, for example, The Logic of Reliable Inquiry (K. T. Kelly, OUP, 1996, page 381-383) "minor differences in the order in which they receive the data may lead to different inductive conclusions in the short run. These distinct conclusions cause a divergence of meaning between the two..." "...there is no version of the truth that all users converge to." This happens with Asa H specialist agents. One can also see it happen in human society, between communists and fascists, for example. (See also my blog of 21 July 2016) This was not what happened the other day with Kellyanne Conway, however. She was simply lying.

Saturday, January 28, 2017

Quantum fields

I am trying to read Lancaster and Blundell's Quantum Field Theory for the Gifted Amateur (OUP, 2014). I would be willing to accept quantum fields as the ultimate reality* if only there weren't the issues raised by general relativity. (see, for example, Monton's Against 3N-Dimensional Space in The Wave Function, Ney and Albert, OUP, 2013)

*"Therefore even matter itself is an excitation of a quantum field and quantum fields become the fundamental objects which describe reality." Lancaster and Blundell, page 2.

Friday, January 27, 2017

Interspecies communication

All formalisms are idealizations.  Like any other formalism natural language provides only an approximate description of  the world. (See my blog of 24 Jan. 2017 and The Philosophy of Niels Bohr, H. J. Folse, North Holland, 1985) Having different senses and effectors Asa H will always have somewhat different concepts than humans have. (Even different humans have somewhat different concepts, one from another.) I suspect, then, that man-machine communication in natural language may never be as good (accurate) as is natural language communication between humans. The Turing test may be exactly the wrong way to try to judge intelligence. I believe that a lot of the current work on artificial intelligence may be over emphasizing "conversational intelligence."

There will also be an impact on the quality of machine reading systems.

Thursday, January 26, 2017

Memories

Humans may have sensory memory (of duration of a few millisec), short term memory (duration of at most a few tens of sec), intermediate-term memory (several hours), and long term memory (>= a few tens of min). Asa H has a different memory scale on each successive hierarchical level that it learns.

Tuesday, January 24, 2017

Private language versus communicable knowledge

The network of concepts that are recorded as Asa H's casebases constitute Asa's knowledge and private language.  Even if we were able to assign "names" ("words") to each of these concepts (see my blog of 30 October 2015) they will not be identical going from one Asa agent to another and they will change over time.  The situation becomes even more uncertain if we try to relate Asa's thoughts to those of humans.
If an Asa agent begins life tabula rasa it is possible to read into it casebases from some other (previously trained) agent. In this case the private language developed by the trained agent also serves as a language of communication.  If all agents begin life with the same few foundational concepts (casebases) it is again possible to read some newly learned concepts from one agent to another.  But once agents have become specialists their various concepts will differ, one from another, and they no longer share a common private language. 
By naming each of (many of) an agent's concepts one forms a communicable (shared) language. But somewhat different concepts, held by different agents, will all come to share the same name.  Inter agent communication and publicly shared knowledge will be imperfect. This is important in thinking about some of Wittgenstein's work and the limitations of language.
Languages need not be composed of names/words only. Mathematics often serves as a language. Again, concepts may change over time. An example might be the concept of  "algebraic ideal." Concepts can also be forgotten while new concepts are created. An example might be "cohomology."

Saturday, January 21, 2017

Space travel for artificial intelligences

I have considered sending the human genome from planetary system to planetary system without the actual transport of matter. (Trans. Kan. Acad. Sci. 118, 2015, page 145 and my blogs of 11 May 2012 and 5 July 2015 ) How might one do the same sort of thing for artificial intelligences? My artificial intelligences A.s.a. H. and A.v.a. Have usually been written in QB64, C++, and VBA. How should one write an artificial intelligence that you want to be run on some arbitrary unknown computing system? Or, must you send the design for the machine as well?

Thursday, January 19, 2017

Asa's imagination and creativity

Asa and other creativity machines (see R. Jones, Trans. Kan. Acad. Sci., vol. 102, 1999, pg. 32) generate new ideas in the ways described in my blogs of 12 Feb and 23 Oct 2015.  But part of the creative process is the subsequent evaluation of new ideas. As Asa continues to interact with its environment concepts (ideas) that are not useful are strongly modified or are deleted altogether.

Monday, January 16, 2017

Why don't the best people go into management?

They're paid well but administrators spend their time doing scheduling, organizing, report writing, running staff meetings, recruiting, interviewing, making personnel decisions, office politics, etc. Not work for a first rate mind. And so, by default, we get our Trumps.

Sunday, January 15, 2017

An agent can know too much

A large knowledgebase slows down search.  A specialist agent will be slowed down by having knowledge of matters outside of its specialty.

Friday, January 13, 2017

Asa's model of the world

Asa H begins to form its most primitive concepts as a phenomenalist. Each concept vector (or case) begins as a bundle of sensory inputs (or motor outputs) in space and time. But averaging, interpolation, extrapolation, and other learning algorithms modify each of Asa's mental concepts. Asa's ultimate objective (and that of science) is to bring compact order to an ever growing range of experiences.

Thursday, January 12, 2017

Asa agents of varied knowledge and intelligence

The wide range of occupations of which a human society is composed are not all intellectually demanding. Similarly, not all specialist Asa H agents in a society of artificial intelligences will be equally intelligent.

Saturday, January 7, 2017

Different conceptualizations of reality

Human's are justified in rejecting psychic phenomena and concepts like telepathy, telekinesis, and dowsing.  My artificial intelligence Asa H can detect buried metal (see blog of 1 Jan 2017), communicate via WiFi, sense its location using GPS, exert magnetic forces on distant objects, sense the earth's magnetic field and is justified in forming concepts based on these.

Truth is different for math, physics, chemistry, and engineering

(Refer back to my blog of 1 Jan 2017.) Mathematics, the science of patterns, places a strong emphasis on the truth component "deduction from assumptions" and a much weaker emphasis on the component "usefulness."  Engineering, on the other hand, places a much stronger emphasis on "usefulness." Physics and chemistry both place emphasis on the truth component "agreement with observation" but the cold fusion episode shows us that physics places more emphasis on theory than chemistry does while chemistry places more emphasis on experiment ("agreement with observation") than physics does. The vector concept of truth is a bit different in each field and for each individual scientist. A theoretical physicist's valuation differs from that of an experimental physicist.

Tuesday, January 3, 2017

The body in the mind

My artificial intelligence Asa H's model of the world/reality (see my blogs of 1 October and 5 November 2015) is roughly divided into a model of the environment and a model of the self. (There is some overlap.) I believe that the model of the self (see my blogs of 21 July 2016 and 1 January 2017 "expanding consciousness") is something like what Jackson had in mind in his book The Body In the Mind (U. Chicago Press, 1990). Experiments with Asa H allow us to study how this kind of thinking might operate.

Sunday, January 1, 2017

Merging multiple sensory modalities in deep machine learning

I am studying David Harwath and Jim Glass's paper from CSAIL, "Deep Multimodal Semantic Embeddings for Speech and Images."  I discussed similar work in my 2011 paper "Experiments with Asa H" but rather than speech and images being the sole modalities I allowed for input from a number of Asa's various senses. (The paper is available on my website www.robert-w-jones.com under "book", chapter 1, page 13.)

Such merging of multiple modalities may explain why it has not been possible to discover a formal grammar that covers human language.

Case patches and interpolation

Rule patches (ripple-down-rules) were developed as a way to improve the accuracy/performance of production systems. (See Compton and Jansen, Proc. 2nd Australian Joint A. I. Conference, 1988, pg 292). It's possible to do the same sort of thing with case-based reasoning systems, provided one has logged a sufficient amount of experience. A patching case base takes as input the new input vector along with the nearest matching retrieved case vector and provides as output a prediction of the error in the expected output. One can use the error to attempt to correct/improve the output. This simply amounts to interpolation or case adaption. It can be used in case-based reasoners as well as A.s.a. H.

Senses

I've given Asa H a metal detector wand as an additional sensory input.

Different kinds of truth and the need for scientific pluralism

I have argued that truth is a vector quantity. (See my review of  Theories of Truth: a Critical Introduction, by R. Kirkham, at amazon.com, 25 September 2008.) There are various kinds of truth and these are each components of the truth vector. Things like:
      Valid deduction from true assumptions
       Satisfying a definition
       Agreement with observation
       Usefulness
       Coherence, consistency
       Etc.
There are then different kinds of science making use of the various different ways of assessing truth (Along these lines see B. Latour, An Inquiry into Modes of Existence, Harvard U. Press, 2013). This is a further argument for the need for scientific pluralism.

Beauty

Humans use the value/concept "beauty" in both mate and scientific theory selection.  My artificial intelligence Asa H and most other AIs do not currently have a concept of beauty, see, for example, my blog of 21 September 2010. Dorner's Psi cognitive architecture does incorporate a notion of beauty, modeled as an emotion. (D. Dorner, Bauplan fur eine Seele, Reinbeck: Rowohlt, 1999, pg 373) Perhaps I should try to give Asa H a concept of beauty. I have Bach's microPsi 2 software running in my lab. (See my blog of 2 March 2015.) Several groups have assembled software modules that assess the beauty of humans.  Even if an AI does not need such a value for its decision making it might still prove useful for man-machine communication and understanding. (Another example of reconceptualizing reality.)

Doing science can make you more intelligent

Models of cognition involve a value function, a dopamine circuit in humans. (See, for example, the model in my blog of 1 Sept. 2012) The goal of any intelligence is to maximize rewards. 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.  Doing science promotes and develops some of the values which we then apply elsewhere in life, things like evidence-based belief.

Expanding consciousness

Using mobile robots, simulators, and hand coding the network fragment that constitutes Asa H's concept of self (see 21 July 2016 blog for a listing) has now been grown, adding the vectors: smell=(MQ-2 sensor, MQ-3, MQ-4, MQ-5, MQ-6, MQ-7, MQ-8, MQ-9, MQ-135), taste=(pH, salinity), touch=(contact1, contact2,..., force1, force2,..., whisker1, whisker2, ...), see=(camera1, camera2, ..., light sensor, color sensor, IR sensor, ...), temperature=(temperature1, temperature2, ...), pain=(pain1, pain2, ...), sense=(smell, taste, humidity, touch, see, hear, pain, B field, E field/charge, sense far, sense near, motion sensors, sense weight, compass heading, accelerometers, wind, temperature, air pressure, GPS positions, solar power), image manipulation=(shift y, shift x, rotatecw, rotateccw, scale +, scale -, reflect, invert), think=(sort, load, save, simulation, search, deduction, interpolation, extrapolation, update, image manipulation, mutate, sensitivity analysis, forgetting, compression), leave=(sense near, move, sense far), battery charge=(battery1 charge, battery2 charge,...), act=(grasp, release, kick, lift, lower, walk, move, carry, turn, approach, leave, push, pull, twist, recharge, taste, electromagnet, fan, messaging, telemessaging), health=(battery charge, pain, temperature, hardware diagnostic, software diagnostic). The development of hardware and software diagnostics is an ongoing project. All together Asa's concept of its self has roughly doubled in size.