Mechanical vs' Symbolic Computation: Two Contrasting Strategies for Information Processing

1

• Mechanical vs. Symbolic Computation Two
  Contrasting Strategies for Information Processing

Anthony F. Beavers The University of Evansville
afbeavers_at_gmail.com
2
Mechanical vs. Symbolic Computation
What is a computation?
... a procedure whereby tokens are manipulated
according to the specifications of a formal
language D1.
3
Symbolic Computation
Alan Turing 1912-1954
4
If a computation is D1, then examples include ...

• math problems,
• logic problems, or
• any other problems that are encoded (or
  represented) by tokens and solved by use of a
  formal language.

5
These computations are symbolic, not by virtue of
their implementation (which is mechanical), but
by the method used to encode and manipulate
information.
Symbolic computation is possible because of the
relationship between tokens, or representations,
that stand in for other things or
representations, and the formal specifications
for manipulating them.
6
The Symbolic Computational Strategy

• Represent information using some appropriate set
  of tokens.
• Write a formal specification (or program) for
  manipulating these tokens to produce a desired
  output.
• Never mind the details of the physical
  implementation, because they are unimportant.

Note that realized symbolic computations do
involve mechanics at the level of implementation.
But the power of symbolic computation relies on
programs and tokens.
7
So, mechanical computation is not D1

• Because in mechanical computation there is no
  formal specification or program, and
• Information processing is controlled by
  physically changing the machine (or getting the
  machine to change itself).

8
Why call it computation at all?
I dont know
Some try this for a definition A computation is
any nomological transformation of input into
output D2.
But not me, so lets start over.
9

• Mechanical vs. Computational Information
  Processing Two Strategies for Transforming
  Inputs to Outputs

Anthony F. Beavers The University of Evansville
afbeavers_at_gmail.com
10
Mechanical vs. Computational Information
Processing
What is information processing?
... a procedure for nomologically transforming
information from input into output. ... between
input and output???
11
Information Processing

• Examples include ...
• Symbolic (or Turing style) computation
• Network processing using ...
• Biological neural networks (BNNs)
• Artificial neural networks (ANNs)
• Dynamic associative networks (DANs)
• Agent-based modeling of Complex Adaptive Systems
  (CAS)

12
Information Processing
Its what goes on between information storage and
retrieval.
Some Considerations concerning Devices

• A book vs. an eDocument
• A record player vs. an MP3 player
• An analog telephone vs. a cell phone
• A network vs. a computer

13
Computation is a form of information processing
that uses a formal language and symbols to
program a mechanical device to transform
information from input into output.

• Has this use of program and symbol been a
  seduction in AI and cognitive science that has
  led us to believe that the software does the
  important work? (Note that in a programmed
  machine there is no software.)
• Can we do without it, and if so, what else is
  there?

14
Two Inventors at the Start of the Information
Revolution
Thomas Alva Edison (1847-1931)
Alexander Graham Bell (1847-1922)
(Turings dates were 1912-1954)
15
Thomas Alva Edison (1847-1931)

• Carbon Telephone Transmitter (1876)
• Phonograph (1877-1878)
• Light Bulb (1879)
• Wireless Telegraphy (1881)
• Power Plant (1882)

• Motion Picture Camera (1891)
• Alkaline Storage Battery (1900)
• Kinetophone (1912)
• Telescribe (1914)

Original Foil Phonograph 1877 Edison National
Historic Site
16
Alexander Graham Bell (1847-1922)

• Started experiments with the production of sound
  (1863)
• Opened his School of Vocal Physiology and
  Mechanics of Speech (Boston,1872)

• Telephone (1876)
• Photophone (1880)
• Wax Cylinder for Edison (1881)

Watson, Come here. I need you. (1876)
17
Neither Edison nor Bell were able to work with
information processing. They were able to work
with information, storage, transmission and
retrieval
mechanically.
The question on the table is whether we can do
information processing mechanically with
Edison/Bell-style information encoding rather
than Turing-style encoding.
18
Several Have Tried
(though perhaps without realizing it)
But what, then, do we mean by Edison/Bell-style
information encoding?
19
Edisons Phonograph (1878)
Edison called his sound-recording machine the
phonograph, which means literally "sound-writer."
It had a wooden cylinder with a thin sheet of
foil wrapped round it, and a sturdy needle with a
horn attached to it pressed against the foil.
Edison spoke into the horn and the sound energy
from his voice, funnelled and concentrated by the
horn, made the needle vibrate up and down. As
Edison cranked a handle, the cylinder rotated and
the needle cut a groove into the foil. Since the
needle was moving up and down, the depth of the
groove varied according to how loud or soft his
voice was in other words, the groove was a
recording of the sound of Edison's voice
translated into a mechanical form. To play back
the recorded sound, Edison simply ran the process
in reverse. He put the needle back at the start
of the groove and cranked the handle. Obediently,
the needle ran along the groove, jolting up and
down to follow the pattern it had cut previously.
As it moved about, it vibrated and the noise of
its vibrations was amplified by the horn,
recreating the sound of Edison's voicealbeit in
a very scratchy- fashion. http//www.explainthats
tuff.com/record-players.html
20
Edisons Phonograph (1878)
Smithsonian Museum of American History
21
Bells Telephone (1876)
Working in the transmitter room and trying to
free a reed that had been too tightly wound to
the pole of its electromagnet, Watson produced
atwang . Bell, who had been working in the
receiving room heard thetwang and came running.
Bell surmised the complex overtones and timbre of
the twang to be similar to those in the human
voice. He was now convinced that his vision of
sending speech over a wire was more than just a
dream. As Bell raced to perfect his telephone,
he was also writing up specifications to be filed
with the United States Patent Office in
Washington. On March 7, 1876, he was issued
patent number 174,465. Meanwhile, Bell had
discovered that a wire vibrated by the voice
while partially immersed in a conducting liquid,
like mercury, could be made to vary its
resistance and produce an undulating current. In
other words, human speech could be transmitted
over a wire. http//www.pbs.org/wgbh/amex/telepho
ne/peopleevents/mabell.html
Sketch of Phone System By Bell
22
Bells Telephone (1876)
http//www.people.hofstra.edu/geotrans/eng/ch2en/c
onc2en/bellteleph.html
23
Networks and Mechanical Information Processing

• Biological neural networks (BNNs)
• Artificial neural networks (ANNs)
• Dynamic associative networks (DANs)

The rest of this talk will be concerned with
dynamic associative networks.
24
Dynamic Associative Networks

• Dynamically structured
• No (static) thresholding
• Recurrent (in whole or in part)
• Multiply Realizable

• Learn by adding nodes and connections rather than
  setting weights
• Transformationally equivalent to standard
  artificial neural networks

25
Two Non-Dynamic Precursors

• IdentiNet -
• Identification and classification (from
  comparison of similarity and difference)
• Shows non-monotonic tendencies

• VisNet -
• Signal/Information transduction
• Shows that surface-level information arrangements
  are not indicative of what lies below
• Demonstrates the liquidity of information

26
What Do Such Networks Know?

• They know how to transform a specific set of
  inputs into an output
• They do not know that because without input
  they can do or know nothing

How Do They Know?

• By virtue of their structure or wiring schematic
• They are merely circuits

27
First Dynamic Modifications

• Dynamic Real-Time Network Construction
• Short Term / Longer Term Memory Control
• Experiential Learning on the Fly

• Example Bi-Modal Association Network

28
Second Dynamic Modifications

• Simultaneous Input Clustering
• Text Box Activation / Input File Loading
• Use of Systemic Inhibition

• Example Natural Language Processing (NLP)
• Works without any grammar parsing on an S/R
  activation model
• Language here is reduced to who, when, where and
  what
• Input File Appears on Next Slide

29
NLP Input File Example
Facts F1 Facts F2 Facts F3 Facts F4
Facts F5 Facts F6 Facts F7 Facts
F8 Facts F9 Facts F10 Facts F11
Facts F12 Facts F13 Facts F14 Facts
F15 Facts F16 Facts F17 Facts F18
Facts F19 Facts F20 Facts F21 Facts
F22 Facts F23 F1 Descartes born France
1596 F2 Descartes died Sweden 1650 F3
Descartes wrote LeMonde 1633 F4 Descartes
wrote Meditations 1637 F5 Spinoza born
Amsterdam 1632 F6 Spinoza died 1677 F7
Spinoza excommunicated 1656 F8 Spinoza wrote
PrinCarPhil 1663 F9 Spinoza wrote TheoPolTre
1670 F10 Leibniz born Leipzig 1646 F11
Leibniz died 1716 F12 Newton born England
1642 F13 Newton died 1727 F14 Locke born
England 1632 F15 Locke died England 1704
F16 Locke wrote Essay 1690 F17 Berkeley born
Ireland 1685 F18 Berkeley died England 1753
F19 Berkeley ordained priest 1710 F20
Hume born 1711 Edinburgh F21 Hume died
Edinburgh 1776 F22 Hume wrote Enquiry F23
Newton wrote Principia 1687 Who Descartes
Who Spinoza Who Leibniz Who Newton
Who Locke Who Berkeley Who Hume What
LeMonde What Meditations What PrinCarPhil
What TheoPolTre What Essay What
priest What Enquiry What Principia
When 1596 When 1650 When 1633 When
1637 When 1632 When 1677 When 1656
When 1663 When 1670 When 1646
When 1716 When 1642 When 1727 When
1704 When 1690 When 1685 When 1753
When 1710 When 1711 When 1776
When 1687 Where France Where Sweden
Where Amsterdam Where Leipzig Where
England Where Ireland Where Edinburgh
30
Third Dynamic Modifications

• Automatic Network Tuning Via Systemic Inhibition
• Simultaneous Recursion on Multiple Input Nodes

• Example Abstraction for Query Purposes
• Controlled Branching to Subnets Based on S/R
• Automatic Association Based on Likeness
• For this example, input Is propositionally rather
  than spatially encoded.

31
Other Examples and Projects

• Buffering and Branching of 8-bit ASCII Sequences
• Primary Sequential Memory Model
• Input and Recall of Seven Digit Sequences
• Gesture Recognition - Michael Zlatkovsky (Indiana
  University)
• Automatic Document Classification - Guy Wyant
  (University of Evansville)

32
Acknowledgements
Some of this research was conducted while on
fellowship from the National Endowment for the
Humanities. Any views, findings, conclusions or
recommendations expressed here do not necessarily
reflect those of the National Endowment for the
Humanities.
Dynamic Network Development Team
Tony Beavers
Michael Zlatkovsky
Guy Wyant
Project Director
Programmer
Programmer
Anthony F. Beavers, Ph.D. - Professor of
Philosophy Director of Cognitive Science The
University of Evansville - 1800 Lincoln Avenue -
Evansville, Indiana 47722 afbeavers_at_gmail.com -
http//faculty.evansville.edu/tb2/