Timing Devices
or
why brains are not computers

The belief that "brains are computers" is widely held. (See, e.g., the Encyclopedia of Computational Neuroscience). Timing devices challenge that belief in a concrete and practical way.

(You can download the full Timing Devices article in .pdf format. See also links near the bottom of the page.)

Essence of Timing Devices

1.  The purpose of timing devices designs

Timing devices are proposed electronic components that can be assembled into networks. An individual timing device embodies essential activity of a neuron; and a network of timing devices operates an engineered organism like a brain operates an animal. Some timing devices in such a network drive muscle-like fibers while other timing devices are sensitive to external stimuli and act as sensors. E.g., in a proposed engineered organism that would dwell in a water environment (shown in Figure 27 from the Timing Devices article), sensory units produce signals that control muscle-like twitches in motor units so as to direct the engineered organism with respect to a the source of stimulation, e.g., towards a light or a particular food-like molecule.

2.  The primal timing device

The simplest timing device, called the primal timing device, is a point of origin for development. Additional pieces of hardware (e.g., various "projections") and functions are added stepwise to the primal timing device for more complex operations. A development path connects the primal timing device with the general Quad Nets model that is more powerful and that incorporates a suggested account of consciousness. links.

The primal timing device in Figure 1.a has a "response clock" shown in the form of a round clock dial. The response clock resembles a stopwatch used in a sports contest and is controlled by two timing intervals: namely, δ, the responding period, and β, the refractory period. A timing interval is a period of time that governs the operations of a timing device and that is adjustable during operations. A timing interval or time period has a dimension of seconds, e.g., milliseconds.

Timing devices are interconnected by "projections." In the simplest realization, a projection is a wire that conducts electricity. A "projection from" extends out of a timing device and a "projection onto" attaches to a timing device through a "junction." A junction embodies the asymmetrical nature of a projection and resembles a synapse in a neuron. A projection from one timing device becomes a projection onto another timing device.


In operations, as shown in Figure 1.b, an input pulse through the projection onto at t=0 triggers the primal timing device and starts the response clock. After a passage of time equal to the responding period, at t=δ, the timing device produces a pulse through the projection from. The response clock continues to run until t=δ+β and then resets. During operations, the timing device is first in a ready condition, then a responding condition, then a refractory condition, and, finally, it returns to a ready condition. While the timing device is in the ready condition, the response clock is "stopped" at the start point and the ready condition is a state that continues until changed by an input pulse. While the timing device is in the responding condition or the refractory condition, a second input pulse has no effect.

3.  Operational control shown in a "4-cycle" of primal timing devices

The 4-cycle made up of four primal timing devices is shown in Figure 2.a. In the simplest form of the 4-cycle, all timing devices have a common δ and β. During operations, each primal timing device produces pulses in a steady stream, called a pulse train, shown in Figure 2.b. The period between any two pulses in such a pulse train is τ=4δ. Operation of the 4-cycle requires that β<3δ. If β>3δ, a timing device will not have returned to the ready condition when the next input pulse arrives, which will, therefore, have no effect.

Suppose a 4-cycle is producing pulse trains while β<3δ. Then suppose that β is gradually increased until pulsing ceases as β passes upward through β=3δ. That is, a continuous variation in β causes a discontinuous change in the activity of the assembly. The activity of the 4-cycle is controlled by means of the ratio of β to δ. The control is in the nature of switch. In larger assemblies, small variations in timing intervals can cause an engineered organism to switch quickly between global modes of activity, e.g., like a switch from feeding to flight from danger. See § 5 of the Timing Devices paper, setting forth the design of a timing devices assembly called a "pulse period selector" that operates like a band-pass filter in electronics, where switching can be controlled by a small variation in a timing interval.

The 4-cycle is not closely similar to a brain design but it illustrates features of the Quad Nets model, which is more similar to brains. A gradual variation in β and/or δ operates on a longer time scale than the pulsing timing interval based on δ. It is like pressing on the accelerator in a car: the engine is turning over at 15-100 cycles per second (90-6000 rpm) while the variation caused by pressing on the accelerator is occurring at the rate of about 1 cycle per second. In sum, timing device assemblies inherently operate on multiple interconnected time scales.

Operations in computers are temporally more rigid, based on the division between hardware and software. "On-the-fly" temporal variations are not inherent in computers but can only occur through high-level refinements.

The lack of inherent temporal adjustments in computers is illustrated by an extract from Biophysics of Computation: Information Processing in Single Neurons (1999) by Christof Koch, at page 470:
"Brains differ in some rather obvious ways from present-day computers: memory and computation are not separated as they are in all of our current machines; the nervous system operates without any systemwide clock and is built from stochastic elements. Finally, developing as well as mature brains are constantly reprogramming themselves, up or down regulating synaptic weights, modulating the degree of adaptation, shifting the character and frequency of central pattern generators, changing the time constants of integration, and so on. Conceptually, this amounts to the input changing the transition function governing how the machine switches from one state to the next. Put differently, a nervous system will act like a machine that changes its instruction set as a function of its input."
Koch's statement is a confusing collection of temporal adjustments that have no inherent basis in a computer model. In contrast, levels of temporal adjustment in the Quad Nets model are organized by the inherent design. On the longest time scale, there are adjustments of junctions that correspond to "regulating synaptic weights." The next shortest time scale is the "situational time scale," corresponding, e.g., to movements of a student from math class to physical education to lunch to English class to band practice. Each situation calls for a different distribution of blood flow (energy flow) into the various brain parts. The next shortest time scale involves variations in timing intervals that govern activities of a brain part, like the variations in β and/or δ that control the activities of the 4-cycle, described above. Finally, at the shortest time scale, are the intervals between pulses, e.g., τ=4δ in the 4-cycle.

4.  Timing devices networks are different from computers

All quantities involved in operations of timing devices, e.g. the timing intervals δ and β, are stated in terms of time and nothing else. Each timing device has its own individual timing intervals; and individual timing intervals can be separately varied, e.g., through sensory input (as in Figure 27, above). In contrast, devices used in computers ("finite state machines") are governed by an external clock and have no internal timing controls. While timing variations are central in timing devices, such variations are of little or no significance in the general theory of computers; and incorporation of timing controls in computerized systems is by way of a superstructure.

An input to a computer device (that is, a finite state machine) changes the state of the device; and the new state remains without change until another input. In contrast, an input to a timing device starts a sequence of changes that occur in accordance with the timing intervals.

Operations in timing device networks (and Quad Nets constructions), as illustrated in the 4-cycle, above, show an essential characteristic: a continuous variation in timing intervals causes a discontinuous change in operations. This is like the change that occurs when liquid water freezes as temperature passes continuously downward through 0º C. Such changes are called phase changes and they are studied in thermodynamics. The physical systems that are conformable to computer operations are studied in the physics of mechanics. (See, e.g., R. Penrose, The Emperor's New Mind, e.g., at 173 - "the Newtonian world is indeed computable.") Thermodynamics and mechanics are different branches of physics and involve different kinds of reasoning. Although some physisicts contend that mechanics "comprehends" thermodynamics, my view is to the contrary. Only very special phase changes, e.g., those that occur "quasi-statically," can be described by mechanics. The "quasi-static process" is inadequate "for engineers who wish to see engines run, not creep." [C. Truesdell and S. Bharatha, The Concepts and Logic of Classical Thermodynamics as a Theory of Heat Engines etc. (1977) at xii.] In my opinion, the same inadequacy afflicts artifical intelligence that is based on mechanical foundations. My "thermal device models of brains" are an alternative to the "mechanical device models of brains" embodied in computers.

The more general Quad Nets approach is grounded in analysis of the Ising Model that addresses phase changes in a simple formal system. Even the simplest Ising Model is extremely difficult to describe mathematically. Calculations of phase changes in advanced Ising Model forms have been shown to be NP-complete. (See Istrail, S., ~2000, "Statistical Mechanics, Three-Dimensionality and NP-completeness, I. Universality of Intractability for the Partition Function of the Ising Model Across Non-Planar Lattices," link.) Phase changes in Quad Nets constructions appear to be even more intractable to computer emulation. In the Timing Devices article, I present very simple designs that appear to present extreme difficulties for computer emulation. (I am not a computer scientist and will appreciate comments from knowledgeable persons on this proposition.)

5.  Networks of timing devices and Quad Net devices have advantages over computers in accounting for biological brains and human psychology.

I am unable to pursue the many possibilities suggested by these systems. However, the following propositions appear to me to be potentially fruitful in ways that (as far as I can see) are superior to computerized models. The "4/4 Musical Meter Generator" shown in the adjacent image can generate any 4/4 musical meter within broad operating constraints. No "computational system" is needed. The Generator (shown here as if "dissected" from a full brain) combines the more advanced Quad Nets system with the simpler Timing Devices system, but it is an operational design and potentially realizable. Timing devices make up the 4-cycle previously discussed that is at the top of the image and that provides the basic 4/4 beat (with timing adjustments available for each pulse in the cycle). Additional projections extend downward from the 4-cycle and operate according to the "equal-output rule" discussed in the Timing Devices paper, section 3: when a timing device with multiple projections from produces a pulse, it does so equally through all projections from that timing device.
 
The "assembly of Quad Net device parts" at the bottom of the image resembles the 4-cycle. The white round objects are "Toroidal Quad Nets" (TQN's) that are collective forms of a simple timing device. The blue arrow-like objects are collective forms of simple projections. During operations, each TQN periodically generates a pulse bundle (a collective form of a single pulse) and pulse bundles circulate in the assembly of Quad Net devices at the same tempo as the circulating pulse in the 4-cycle. The 4-cycle drives the assembly of Quad Net device parts through (1) the "intermediary timing devices" that enable further fine-tuning of timing variations, (2) the "one-many connections" that use the equal-ouput rule and (3) green and purple "controllers." (This assembly of Quad Net device parts is a straightforward adaptation of the Phase Transfer Controller fully described in the Quad Nets paper.)


An example of output from the 4/4 Musical Meter Generator is shown below. The "high-frequency" pulse bundles with more numerous pulses produce a more intense output than the "low-frequency" pulse bundles with less numerous pulses. A bit of timing variation is shown (in contrast to the "metronome" beat below the line) to illustrate the capacities of the assembly and in keeping with Levitin's observations that such variations are common in music produced by human musicians, in contrast to music produced by machines. Because the Quad Nets model is applicable to all brain parts, including those involved in emotions, timing of output signals can be varied according to a performer's emotions; such variations are foreign to "computational models" such as that espoused by Levitin.

Links (return to top of page).

The Timing Devices article for download.

Embodiment of Freedom (2007) (currently in development) -- a separate website setting forth high-level physical and psychological principles of my alternative approach to "artifical intelligence."

Quad Nets (2006)-- Quad Nets is the chief engineering and scientific presentation of my "thermal device models of brains." Timing devices discussed here are a simplified version of the Quad Nets model, like a RISC ("reduced instruction set computer") is a simplified version of a CISC ("complex instruction set computer")

Quad Nets website -- separately organized

web page on the full article titled Quad Nets: Material Foundations for Thermal Device Models of Brains

download Quad Nets article (.pdf format)

Your comments and suggestions are welcome. Please write to the adjacent email address (shown in an image to minimize spam).

You are also welcome to visit my professional web page:
Robert Kovsky, Attorney-at-Law (services provided for attorneys only)

Copyright © 2007 Robert Kovsky, revised version February 3, 2008