Brain Architecture for an Intelligent Stream of Consciousness

Abstract

This chapter pursues a search for deeper computational power within the brain, needed to explain not only higher intelligence but also common sense, truth judgment, intuition, and artistic appraisal. Toward this end what neurons might accomplish using natural pulse bursts is discussed.

Four types of pulse bursts are considered below. A burst with a single pulse is a limiting case. Single pulses are necessary for system operation; they are assumed to result from special synapses called weak synapses, although single pulses may also be generated with special neural circuitry. Regular synapses, used for interneural communications and for neural logic, are those which produce a burst of about ten pulses.

Biological explicit memory, to be analyzed in a later chapter, is characterized by persistent neural pulses that readily signal other neurons. Short-term memory (STM) neurons are modeled as delivering bursts up to hundreds of milliseconds. Long-term memory (LTM) neurons are prepared to deliver a controlled burst for an indefinite duration, and may actually turn on and off as needed.

A bottom-up brain architecture is presented based on what neurons actually do, supported by simulations of neurons as suggested in an appendix to this book. Major system components of brain architecture include (1) STM, which holds images from the senses and recalls from LTM; (2) subconscious associative LTM, which stores all that a person knows; and (3) a system of editors to adjudicate between conscious STM and subconscious LTM. The editors mentioned below are (1) a cue editor which uses inexact cues to find information quickly in long-term associative memory and (2) a recall editor which selects the highest priority image in a group of multiple recalls.

Finally there is discussion of brain decision making: (1) decisions based on past experience; (2) decisions that use a random variable; and (3) inspired decisions that are difficult to explain using ordinary logic. Quantum computations are considered for inspired decisions, as are simulated qubits, which are classical and do not require a quantum system. Recursive neurons, presented in a later chapter, can behave like qubits, such as supporting a probability space that holds true and false simultaneously. Perhaps more important, recursive neurons are capable of controlled toggling and so may be important to mental calculations.