December 9, 2010

The Common Good

One ethical argument is that group interests should have priority over selfish interests. An investigation of ethics must consider this argument and develop metrics for the common good. No-one should assume that it is easy to define the common good. In political battles, clearly divergent if not contradictory ideas of the common good prevail and efforts to achieve consensus are difficult to impossible. The ethical implications are profound.

Michael Sandel asks What’s the Right Thing to Do? He teaches political philosophy at Harvard and offers the most popular course on campus -- Justice. One of his intellectual anchors is Jeremy Bentham who wrote Introduction to the Principles of Morals and Legislation in 1780. Bentham proposed a utilitarian test to evaluate the morality of any action: ask the question will my action produce the greatest amount of happiness for the greatest number of people? John Stuart Mill later argued that respect for individuals rights as "the most sacred and binding part of morality" is compatible with the idea that justice rests ultimately on utilitarian considerations. In simple terms, the two arguments compare individual interests with group interests.

Sandel also reviewed the philosophy of Immanuel Kant who argued that reason tells us what we ought to do, and when we obey our own reason, only then are we truly free. Kant’s ideas seem oddly unrealistic in the 21st century. Reason is in short supply. Every person assumes that he or she is more reasonable than others who disagree. There is no consensus about the “common good.” We know that some humans are bad and will harm others as a matter of course; their behavior will not be altered by rational argument or laws and must be constrained by force. Some of these bad people arrive in positions of authority and power. Some bad people are elected, even to the highest positions in government where they can do much harm without insight or remorse.

We know that the audience, the "public", is made up of different groups with vested interests that conflict. We know that everyone invents stories that support their own point of view. Everyone deceives others and there is no absolute truth. We know that the voting public contains individuals with different mental abilities and that most humans have distinct limitations on what they can and will understand.

Human destiny as a species still lies with the programs in the old brain that offer only limited empathy and understanding and insist on the priority of local group survival at any cost. Individuals can transcend the old programs by diligent learning and practice but individual effort and learning does not change the genome, so that there can be no enduring civility without the persistent and relentless initiation of new humans into a rational and compassionate world order. Whatever we value about civilized human existence - culture, knowledge, social justice, respect for human rights and dignity must be practiced anew and stored as modifications of each person's neocortex.

From The Good Person, Ethics and Morality by Stephen Gislason

Download eBook (PDF) from Persona Digital

December 1, 2010

Brain Circuits and Computer Metaphors

Wild Speculation

I have been reading and hearing too much wild speculation about smart computers and computer chips implanted in humans to replace damaged brain parts or to make people smarter. 99% of this speculation is not smart. I want to provide a basic understanding of brain "ciruits" and put computer metaphors in perspective. Here is an introduction from my Neuroscience Notes.

What About Circuits?

The word circuit is often used in neuroscience. Early metaphors compared brain connectivity with electrical wiring that appeared on streets and in buildings. Simple electrical devices such as switches, relays and telephone switch boards were often used as models of neuronal network components. An early description of an action potential travelling along an axon became "firing" and this inappropriate comparison is still with us. Later, electronic circuits became better models of how neurons may be connected together and neural networks were built with transistors and circuit boards. Any comparison of brain activity with electricity can be misleading, since the voltage fluctuations detected by electronic instruments are produced chemically in the brain and are transmitted by ion fluxes across cell membranes, quite different from electrons flowing in wires and transistors.

An important sub-discipline, electrophysiology, explored the wave activity of the brain using the equipment and techniques developed to monitor electronic circuits. A fundamental concept of brain activity was the dynamic balance between excitation and inhibition. Arousal, action and reaction were properties of excitation. Rest, recuperation and inactivity were properties of inhibitory systems. Loss of inhibition in the brain has serious consequences, from hyperactivity to seizures.

Electronic circuits evolved dramatically in the last half of the 20th century and became essential for much human productivity, communications and information storage. Two domains of circuitry developed – analogue and digital. The differences are important to understand.

Analogue Not Digital

In electronic terms, the brain is more like an analog system than a digital device. The difference is obvious to theorists and electronic engineers but may not mean much to other people. A tape recording is an analog record that stores a waveform analogous to sound waves or electronic waves generated by sound interacting with a microphone. A digital recording stores samples of the waveform as a series of bits. To play back a digital recording you have to convert the samples back to an analogue waveform that plays through sound-producing speakers.

Analog devices are more closely related to the phenomena they represent. Analog computers use circuits that compute in real time. You can arrange transistors to add or subtract two incoming signals - this can be done with a few transistors and the circuit can operate in real time 24 hours a day with instantaneous results.

You could add other simple circuits to notice and make decisions about the similarity or difference between the two signals. This is an efficient approach to real-time computing. You would construct an analog computer if you wanted a simple, efficient and reliable system that sensed, decided and acted in real time continuously over many years. This is the essence of living brains.

You can identify several functional circuits in the brain that can be represented by analogue electronics: oscillators, tuners, feature detectors, mixers, differentiators, integrators, amplifiers, filters, relays, compressors, exciters and feedback controls.

In contrast, a digital computer requires a more elaborate structure - the input signals have to be encoded into a binary stream and stored in memory along with program instructions. The data can later be directed to a CPU that combines the incoming data with a program using logic, arithmetic instruction and memory addresses to arrive at the result. Results of CPU manipulations are then stored in memory, decoded and displayed. The digital machine is more complex than an analogue computer, less reliable and consumes more space and energy than the analogue computer to accomplish the same task.

A brain takes the analog approach and accomplishes sophisticated computations in real time, efficiently, usually without storing any data. Early studies of neurons focused on the on-off characteristic of action potentials and a misleading comparison has been made with the transistor switch in digital circuits. An action potential does not have the meaning of a 0 or 1 digital switch since there is no evidence of binary coding in the brain. The most important distinction between digital computing device and the brain is that the digital device separates data and the procedures that operate on the data. A brain integrates data and procedures so that you can never separate data, programs, and hardware. There are no brain programs that resemble computer programs stored in a coded format since all the programming and all the data is built into neuronal networks. To change a brain program you have to grown and/or prune the connections among neurons. There are several properties of living systems that set them apart from non-living systems. The most fundamental properties of a living system are not possessed by any non-living system are a preprogrammed but adaptable metabolism, self-replication, self-reference, self-modification, spontaneous activity, growth and repair. Information processing in the brain is electro-chemical and involves a neuronal version of quantum mechanics. The quanta are packets of chemicals, neurotransmitters.

Koch suggested many years ago that the brain should be viewed as a hybrid computer, one that employs both digital pulses (between neurons) and analog computations (within them). I do not share Koch’s notion that there is any digital processing going on in the brain. A neuron’s action potential or “spike“ is an analog signal,; a series of action potentials can be compared to pulses in frequency modulated radio transmission. The prerequisite of digital computing is not the wave form but the binary nodes that decide the meaning of inputs. You would have to find neurons wired like switching transistors to make the logical decisions: and, or, not-or, not-and. You would also have to find memory storage in binary code, large arrays of two-state switches with a reader that can tell if each switch is on or off. I will be surprised if a single instance of digital computation is discovered in animal brains.

I will argue that animal brains interface analogue-like input and output circuits with processors that are unlike any electronic circuits we have invented and we do not understand the basic principles. We know that memories are not true analogues of experiences, but rather are compressed recordings of features extracted from the incoming data stream. The recordings are not discrete entities; they are distributed, overlapping and interactive.

While you can create a science fiction story of self-modifying computers, the engineering challenges are formidable. The first step in a practical system would be to develop self-modifying software that does not frustrate users with inappropriate decisions and does not self-destruct. The sense, decide, act and remember capabilities of insect brains are remarkable, but no man-made system can come close to realizing the genius of insects. You could argue that human intelligence is just an elaboration of ant intelligence, complete with dependence on group interactions. Or you could argue that humans and other animals are special creatures with unprecedented abilities.

The intelligence in digital machines is not native to the machines and was put there by humans. The advantage of investing your knowledge and skills in a computer program is that your hard-earned procedural skills are retained and can be used by hundreds of millions of other humans for years to come. Groups of smart, dedicated people can perfect programs over time so that users have all the advantages of accumulated human experience.

You can download a PDF copy of Neuroscience Notes by Stephen Gislason at Persona Digital Online.